97
" tI_ Revenue Forecasting Models and Tax Policy Analysis for Ghana August 2003 Sigma One Corporation " tI_ Revenue Forecasting Models and Tax Policy Analysis for Ghana August 2003 Sigma One Corporation

Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

  • Upload
    vokiet

  • View
    223

  • Download
    8

Embed Size (px)

Citation preview

Page 1: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

"

tI_

Revenue Forecasting Models and Tax Policy Analysis for Ghana

August 2003 Sigma One Corporation

"

tI_

Revenue Forecasting Models and Tax Policy Analysis for Ghana

August 2003 Sigma One Corporation

Page 2: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

WI

August 2003

Revenue Forecasting Models and Tax Policy Analy~is for Ghana

Submitted to:

U.S. Agency for International Development Mission to Ghana

for:

Trade and Investment Reform Program mRP) Improved Policy Reform and Financial Intermediation

USAID Contract Number: 641-C-OO-98-00229

by:

Duke Center For International Development Duke University, Durham, North Carolina

Under a Sub Contract from Sigma One Corporation

In fulfillment of the following milestones:

2.29 Increased efficiency of existing or proposed revenue generation activities

2.30 Improved and balanced policies for revenue generation 2.33 Recommend alternate ways to enhance revenue buoyancy 2.34 Propose the adoption of methodology to forecast tax revenues

relatives to a set of exogenously provided macroeconomic assumptions 2.35 Test the implementation of methodology adopted to estimate the

revenue impact of changes in tax policy using microeconomic data

Sigma One Corporation

..

WI

August 2003

Revenue Forecasting Models and Tax Policy Analy~is for Ghana

Submitted to:

U.S. Agency for International Development Mission to Ghana

for:

Trade and Investment Reform Program mRP) Improved Policy Reform and Financial Intermediation

USAID Contract Number: 641-C-OO-98-00229

by:

Duke Center For International Development Duke University, Durham, North Carolina

Under a Sub Contract from Sigma One Corporation

In fulfillment of the following milestones:

2.29 Increased efficiency of existing or proposed revenue generation activities

2.30 Improved and balanced policies for revenue generation 2.33 Recommend alternate ways to enhance revenue buoyancy 2.34 Propose the adoption of methodology to forecast tax revenues

relatives to a set of exogenously provided macroeconomic assumptions 2.35 Test the implementation of methodology adopted to estimate the

revenue impact of changes in tax policy using microeconomic data

Sigma One Corporation

Page 3: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

• REVENUE ESTIMATING MODELS:

AN OVERVIEW and MACRO-MODELS l

1. Introduction

Revenues and revenue forecasting are at the core of the budget and the budgeting process. An estimation of future revenues creates predictability in expenditure programs, reduces deficits, makes own-resource sustainability possible and helps avoid unexpected deficits and revenue crises.

The Government of Ghana does not have well-established capacity in revenue forecasting and one of the aims of this project is to bridge that gap. The following capacity building steps have been undertaken:

(I) A four-week training program for representatives of all organizations involved in tax analysis and revenue forecasting.

(2) Building of data bases in the revenue agencies: a. Using existing revenue data by tax type: all revenue agencies b. Using existing computer based tax administration data bases: Customs.

Excises and Preventive Services; and Value Added Tax (3) Conducting sample surveys where missing data to create new data bases: Inland

Revenue Services (4) Building revenue forecasting models: all revenue agencies (5) Planning for organizational and systems design and implementation of data systems

and models

This report describes the present status of the databases in the three revenue agencies and the different forecasting models that may be used for estimating revenues.

1.1 Revenue Growth and Structure In Ghana, revenues have ratcheted up at the beginning of every decade. The sum of revenues and grants has been increasing over time:

• 1983-86 6% to 15% of GOP • 1992-94: 15% to 20+% of GOP

• 2000-03: 20% to 25+% of GOP

The composition of revenues has also changed over time. Non-tax revenues and revenues from cocoa tax have followed a declining trend. After an initial growth phase. the import duty stayed around 3-4% of GOP. On the other hand, domestic excise duty declined from about 2% to less than 1% of GOP. Petroleum taxes peaked at 3.9% in 1993-94, declined to 1.7% in 200 I, and now is slightly higher than 4% of GOP. The revenues from income taxes have gradually grown from less than 2% to higher than 5% of GOP. The sales taxN AT grew gradually to 1.6% of GOP in 1998; VAT in 1999 increased revenues 10 3.8% and has grown

I Prepared for the Ministry of Finance & the Revenue Agencies Governing Board, Accra. Ghana by a Duke University Center for International Development team lead by Profs. G. Glendayand G.P. Shukla. lbe \\Corl was funded by the USAID Mission to Ghana through Sigma One Corporation Contract No. 641-C-OO-98·00~2'J over the period Octoher 2002 through Septemher 2003.

• REVENUE ESTIMATING MODELS:

AN OVERVIEW and MACRO-MODELS l

1. Introduction

Revenues and revenue forecasting are at the core of the budget and the budgeting process. An estimation of future revenues creates predictability in expenditure programs, reduces deficits, makes own-resource sustainability possible and helps avoid unexpected deficits and revenue crises.

The Government of Ghana does not have well-established capacity in revenue forecasting and one of the aims of this project is to bridge that gap. The following capacity building steps have been undertaken:

(I) A four-week training program for representatives of all organizations involved in tax analysis and revenue forecasting.

(2) Building of data bases in the revenue agencies: a. Using existing revenue data by tax type: all revenue agencies b. Using existing computer based tax administration data bases: Customs.

Excises and Preventive Services; and Value Added Tax (3) Conducting sample surveys where missing data to create new data bases: Inland

Revenue Services (4) Building revenue forecasting models: all revenue agencies (5) Planning for organizational and systems design and implementation of data systems

and models

This report describes the present status of the databases in the three revenue agencies and the different forecasting models that may be used for estimating revenues.

1.1 Revenue Growth and Structure In Ghana, revenues have ratcheted up at the beginning of every decade. The sum of revenues and grants has been increasing over time:

• 1983-86 6% to 15% of GOP • 1992-94: 15% to 20+% of GOP

• 2000-03: 20% to 25+% of GOP

The composition of revenues has also changed over time. Non-tax revenues and revenues from cocoa tax have followed a declining trend. After an initial growth phase. the import duty stayed around 3-4% of GOP. On the other hand, domestic excise duty declined from about 2% to less than 1% of GOP. Petroleum taxes peaked at 3.9% in 1993-94, declined to 1.7% in 200 I, and now is slightly higher than 4% of GOP. The revenues from income taxes have gradually grown from less than 2% to higher than 5% of GOP. The sales taxN AT grew gradually to 1.6% of GOP in 1998; VAT in 1999 increased revenues 10 3.8% and has grown

I Prepared for the Ministry of Finance & the Revenue Agencies Governing Board, Accra. Ghana by a Duke University Center for International Development team lead by Profs. G. Glendayand G.P. Shukla. lbe \\Corl was funded by the USAID Mission to Ghana through Sigma One Corporation Contract No. 641-C-OO-98·00~2'J over the period Octoher 2002 through Septemher 2003.

Page 4: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

to 5% by 2003. Grants have fluctuated around 2% over the years; since 2001 they have been on the higher side at 3.1 % to 6.9% and currently stand at 4.4% of GDP.

The status of tax revenues and their trend in the recent past (1999-02) and projections for 2003 and 2004 are shown in the table below.

1999 2000 2001 2002 2003 2004 Prov Proj Proj

A. Tax Revenue 14.9% 16.3% 17.2% 17.4% 20.0% 21.7% 1. Income & Property 4.5% 5.2% 5.6% 5.7% 5.5% 5.6%

PAYE 1.5% 1.8% 1.8% 1.9% 1.9% 2.0% Self-Employed 0.3% 0.3% 0.3% 0.3% 0.3% 0.4% Companies 2.2% 2.6% 2.5% 2.4% 2.3% 2.5% Others 0.4% 0.6% 1.0% 1.1% 0.9% 0.8%

2. Domestic Goods & Service 6.6% 7.4% 7.5% 7.6% 9.9% 11.1% Excise Duty 0.8% 0.8% 0.7% 0.7% 0.7% 0.7% Sales Tax

VAT total (net of refunds) 3.8% 4.7% 5.1% 4.7% 5.0% 5.8% Domestic V AT 1.6% 1.4% 1.3% 1.5% 1.6% 1.9% Import VAT 2.3% 3.3% 3.8% 3.3% 3.6% 4.1% VAT Refunds 0.1% 0.1% 0.2% 0.2% Petroleum Tax 2.0% 2.0% 1.7% 2.2% 4.2% 4.6%

3. International Trade 3.8% 3.6% 4.1% 4.1% 4.6% 4.9% Imports 2.6% 3.0% 3.3% 3.4% 3.7% 4.1%

Export - Cocoa 1.2% 0.7% 0.8% 0.7% 1.0% 0.9% B. Non-Tax Revenue 1.8% 2.6% 0.9% 0.5% 0.6% 0.7% Tax + Non-Tax Revenue 16.7% 18.9% 18.1% 17.9% 20.6% 22.4% C. Grants 1.7% 2.1% 6.9% 3.1% 4.4% 4.4% TOTAL REVENUE & GRANTS 18.3% 21.0% 25.0% 21.0% 25.0% 26.8%

1 Note: Tax revenues from Imports mclude more than regular Import duties. In 2002, taxes on Imports were composed of import duties (80.3%), other levies on imports (ECOW AS, exports development etc. were about 9.3%). other customs fees and charges (6.6%) and export duties.

1.2 Some Macro-Trends and Implications for Tax Revenues Over the past decade, the average growth rate and the standard deviation of real GDP and the rate of inflation have been as follows.

Real GDP Real GDP per capita Inflation (GDP deflator)

Average 4.7% 1.7%

28.8%

Std. dev 1.2% 1.0% 9.8%

These figures indicate that the real GDP growth has been steady, but the inflation rate has been rather high and quite variable. Since taxes fallon nominal values, inflation becomes a high risk variable for revenue forecasting. The situation at present remains unchanged. The GDP deflator for 2003 was originally estimated as 18.8% on the average for year; now the estimate has been revised to 27.6%, which is quite a sil,'IIificant increase. As the exchange rate movements vary to reflect a high inflation, this creates a significant variability in the exchange rate as welL This implies that forecasting of nominal revenues in Ghana will depend heavily on the accuracy of inflation and exchange rate forecasts.

2

-

..

..

..

..

..

..

to 5% by 2003. Grants have fluctuated around 2% over the years; since 2001 they have been on the higher side at 3.1 % to 6.9% and currently stand at 4.4% of GDP.

The status of tax revenues and their trend in the recent past (1999-02) and projections for 2003 and 2004 are shown in the table below.

1999 2000 2001 2002 2003 2004 Prov Proj Proj

A. Tax Revenue 14.9% 16.3% 17.2% 17.4% 20.0% 21.7% 1. Income & Property 4.5% 5.2% 5.6% 5.7% 5.5% 5.6%

PAYE 1.5% 1.8% 1.8% 1.9% 1.9% 2.0% Self-Employed 0.3% 0.3% 0.3% 0.3% 0.3% 0.4% Companies 2.2% 2.6% 2.5% 2.4% 2.3% 2.5% Others 0.4% 0.6% 1.0% 1.1% 0.9% 0.8%

2. Domestic Goods & Service 6.6% 7.4% 7.5% 7.6% 9.9% 11.1% Excise Duty 0.8% 0.8% 0.7% 0.7% 0.7% 0.7% Sales Tax

VAT total (net of refunds) 3.8% 4.7% 5.1% 4.7% 5.0% 5.8% Domestic V AT 1.6% 1.4% 1.3% 1.5% 1.6% 1.9% Import VAT 2.3% 3.3% 3.8% 3.3% 3.6% 4.1% VAT Refunds 0.1% 0.1% 0.2% 0.2% Petroleum Tax 2.0% 2.0% 1.7% 2.2% 4.2% 4.6%

3. International Trade 3.8% 3.6% 4.1% 4.1% 4.6% 4.9% Imports 2.6% 3.0% 3.3% 3.4% 3.7% 4.1%

Export - Cocoa 1.2% 0.7% 0.8% 0.7% 1.0% 0.9% B. Non-Tax Revenue 1.8% 2.6% 0.9% 0.5% 0.6% 0.7% Tax + Non-Tax Revenue 16.7% 18.9% 18.1% 17.9% 20.6% 22.4% C. Grants 1.7% 2.1% 6.9% 3.1% 4.4% 4.4% TOTAL REVENUE & GRANTS 18.3% 21.0% 25.0% 21.0% 25.0% 26.8%

1 Note: Tax revenues from Imports mclude more than regular Import duties. In 2002, taxes on Imports were composed of import duties (80.3%), other levies on imports (ECOW AS, exports development etc. were about 9.3%). other customs fees and charges (6.6%) and export duties.

1.2 Some Macro-Trends and Implications for Tax Revenues Over the past decade, the average growth rate and the standard deviation of real GDP and the rate of inflation have been as follows.

Real GDP Real GDP per capita Inflation (GDP deflator)

Average 4.7% 1.7%

28.8%

Std. dev 1.2% 1.0% 9.8%

These figures indicate that the real GDP growth has been steady, but the inflation rate has been rather high and quite variable. Since taxes fallon nominal values, inflation becomes a high risk variable for revenue forecasting. The situation at present remains unchanged. The GDP deflator for 2003 was originally estimated as 18.8% on the average for year; now the estimate has been revised to 27.6%, which is quite a sil,'IIificant increase. As the exchange rate movements vary to reflect a high inflation, this creates a significant variability in the exchange rate as welL This implies that forecasting of nominal revenues in Ghana will depend heavily on the accuracy of inflation and exchange rate forecasts.

2

-

..

..

..

..

..

..

Page 5: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-.,

.,

.,

1.3 Databases and Revenue Forecasting Models for Gbana As outlined above, the tax datasets in original form had several gaps and discrepancies and needed to be supplemented and cleaned, particularly for the income tax revenues. 1berefore a simultaneous exercise in collaboration with the Revenue Agencies Governing Board (RAGB) was undertaken for collecting samples from the IRS and its field offices. Additional computer-based information was also gathered from CEPS and VAT service research divisions. As the data became available, the comprehensive forecasting models for each kind of tax have been developed and made available to the Ministry of Finance and the RAGB.

2. Forecasting Models

For forecasting tax revenues, three approaches are feasible: (a) GDP based models (b) Monthly receipts models (c) Comprehensive or Micro-simulation models for each tax type

GDP based models: The model uses long-term tax elasticity of revenues with respect to growth in the tax base. For applying this model, three sets of data are required:

(1) Time series of tax revenues for ten years or more (2) Time series of GDP or tax base proxy for ten years or more (3) Series of discretionary changes made over the years and their impact on tax revenues.

This is important for getting an adjusted time series of tax revenues that only reflects the impact of growth in the tax base and not the impact of changes made in the ta:{ rate or tax base from time to time.

Once we have the adjusted time series of revenues and the time series of GOP. tax elasticity is estimated using regression. The elasticity is used to forecast the growth rate of revenues with the help of the growth rate of GOP. Often this k.ind of data is not available which also happens to be the case in Ghana. In such cases an alternative method can be adopted in which the growth rate in the tax base or its proxy is determined with the help of real economic growth and any other explanatory variables. The tax revenues are then taken to grow proportionately with the growth in the base.

These approaches are explained in greater detail below.

Monthly receipts models It is a simple. yet functional tool for making short-term month-to-month projections of revenues from major taxes. The monthly projections can then be used for mak.ing annual forecasts. The model requires primarily monthly tax collection data and the estimated growth rate of the tax base. The advantage is that it captures seasonal effects of tax collections. and with simple calibration. provides relatively accurate results.

The tax receipts forecasting model uses actual monthly receipts data and projected growth rate for GOP or other tax base proxies (e.g. private consumption or imports) to forecast tax collections. The basic approach is as follows. For forecasting revenues for any given month in a financial year. the starting point is the revenue collected in the same month of the preceding year. The rate of growth is taken as a weighted sum of the rate of increase in overall revenues up to the preceding months of the year under consideration compared to the

3

-.,

.,

.,

1.3 Databases and Revenue Forecasting Models for Gbana As outlined above, the tax datasets in original form had several gaps and discrepancies and needed to be supplemented and cleaned, particularly for the income tax revenues. 1berefore a simultaneous exercise in collaboration with the Revenue Agencies Governing Board (RAGB) was undertaken for collecting samples from the IRS and its field offices. Additional computer-based information was also gathered from CEPS and VAT service research divisions. As the data became available, the comprehensive forecasting models for each kind of tax have been developed and made available to the Ministry of Finance and the RAGB.

2. Forecasting Models

For forecasting tax revenues, three approaches are feasible: (a) GDP based models (b) Monthly receipts models (c) Comprehensive or Micro-simulation models for each tax type

GDP based models: The model uses long-term tax elasticity of revenues with respect to growth in the tax base. For applying this model, three sets of data are required:

(1) Time series of tax revenues for ten years or more (2) Time series of GDP or tax base proxy for ten years or more (3) Series of discretionary changes made over the years and their impact on tax revenues.

This is important for getting an adjusted time series of tax revenues that only reflects the impact of growth in the tax base and not the impact of changes made in the ta:{ rate or tax base from time to time.

Once we have the adjusted time series of revenues and the time series of GOP. tax elasticity is estimated using regression. The elasticity is used to forecast the growth rate of revenues with the help of the growth rate of GOP. Often this k.ind of data is not available which also happens to be the case in Ghana. In such cases an alternative method can be adopted in which the growth rate in the tax base or its proxy is determined with the help of real economic growth and any other explanatory variables. The tax revenues are then taken to grow proportionately with the growth in the base.

These approaches are explained in greater detail below.

Monthly receipts models It is a simple. yet functional tool for making short-term month-to-month projections of revenues from major taxes. The monthly projections can then be used for mak.ing annual forecasts. The model requires primarily monthly tax collection data and the estimated growth rate of the tax base. The advantage is that it captures seasonal effects of tax collections. and with simple calibration. provides relatively accurate results.

The tax receipts forecasting model uses actual monthly receipts data and projected growth rate for GOP or other tax base proxies (e.g. private consumption or imports) to forecast tax collections. The basic approach is as follows. For forecasting revenues for any given month in a financial year. the starting point is the revenue collected in the same month of the preceding year. The rate of growth is taken as a weighted sum of the rate of increase in overall revenues up to the preceding months of the year under consideration compared to the

3

Page 6: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

revenues in the same mouths of last year and the macroeconomic forecast for the tax base for remaining months of the year. Thus it is a simple way of projecting future revenues based on the monthly revenue collections of last year and the performance in the preceding months of the current year.

Monthly receipts models have been created for CRPS, V A TS and IRS monthly revenue receipts. The description of these models is provided separately.

Comprehensive models for different taxes For each major type of tax, more comprehensive models have been developed that not only help in enhancing the quality and precision of revenue forecasts but also enable impact analysis of proposed policy changes in the tax system. The following have been developed: micro-simulation models for PA YE and company income taxes based on sample surveys; micro simulation models for trade taxes based on computer-based information disaggregated at the harmonized code and customs processing code levels; and VAT models based on detailed V AT returns that (i) link domestic VAT to import V A T collections, and (ii) allow the estimation of domestic V AT cash receipts to be estimated from the accrued VAT revenues allowing for changes in VAT arrears, credit carry forwards and outstanding VAT refunds. These detailed micro-simulation models are described separately.

3. Macro- or GDP Based Estimating Models

Aggregate tax revenues of a major tax type (personal income tax, V AT, customs, etc) are typically forecast based on the expected growth in the lax base of the type of tax assuming that the rate structure stays the same (or there are no discretionary changes to the tax rate structnre.) For budgeting purposes, these forecasts form the base case revenues before discretionary revenue measures are considered. The results of these macro-models also enter into the monthly receipts forecasting models.

In addition to a static tax structure, these forecasts also assume that there are no major changes in the effectiveness of tax administration or tax compliance, and that the structure of the economy is not changing in a way that affect tax yields. For example, the composition of consumption and imports remains relatively constant such that there is no major shift towards or away from high tax rate goods or services. In general, these macro- or GDP-based forecasts are considered static forecasts in that they do not take into account any of the dynamic interactions that may be occurring between components of an economy as relative prices change over time. In most cases, such interactions are of a second order of magnitude or have offsetting effects and do not affect the results substantially. Care has to be taken to notice major structural changes in an economy that may be occurring and causing significant shifts in the tax base. For example, as will be noted below, Ghana has experienced a major structural change over the last decade towards higher trade shares that make the forecasting of trade-based collections more critical in the overall revenue forecasts.

Two approaches can be used to macro- or GDP-based forecasting. The first approach is to estimate the growth in revenues that arise with real economic growth (or with the growth in a macro-proxy to the tax base) without any change in the tax structure (base, rates or exemptions.) This estimates the tax elasticity, l]Ty' If the tax elasticity exceeds unity, then tax revenues tend to grow faster than the real economic growth, and conversely, if the tax elasticity is less than unity. To estimate the tax elasticity of a tax type requires a time series

4

-

-

..

..

..

..

..

-

..

-

revenues in the same mouths of last year and the macroeconomic forecast for the tax base for remaining months of the year. Thus it is a simple way of projecting future revenues based on the monthly revenue collections of last year and the performance in the preceding months of the current year.

Monthly receipts models have been created for CRPS, V A TS and IRS monthly revenue receipts. The description of these models is provided separately.

Comprehensive models for different taxes For each major type of tax, more comprehensive models have been developed that not only help in enhancing the quality and precision of revenue forecasts but also enable impact analysis of proposed policy changes in the tax system. The following have been developed: micro-simulation models for PA YE and company income taxes based on sample surveys; micro simulation models for trade taxes based on computer-based information disaggregated at the harmonized code and customs processing code levels; and VAT models based on detailed V AT returns that (i) link domestic VAT to import V A T collections, and (ii) allow the estimation of domestic V AT cash receipts to be estimated from the accrued VAT revenues allowing for changes in VAT arrears, credit carry forwards and outstanding VAT refunds. These detailed micro-simulation models are described separately.

3. Macro- or GDP Based Estimating Models

Aggregate tax revenues of a major tax type (personal income tax, V AT, customs, etc) are typically forecast based on the expected growth in the lax base of the type of tax assuming that the rate structure stays the same (or there are no discretionary changes to the tax rate structnre.) For budgeting purposes, these forecasts form the base case revenues before discretionary revenue measures are considered. The results of these macro-models also enter into the monthly receipts forecasting models.

In addition to a static tax structure, these forecasts also assume that there are no major changes in the effectiveness of tax administration or tax compliance, and that the structure of the economy is not changing in a way that affect tax yields. For example, the composition of consumption and imports remains relatively constant such that there is no major shift towards or away from high tax rate goods or services. In general, these macro- or GDP-based forecasts are considered static forecasts in that they do not take into account any of the dynamic interactions that may be occurring between components of an economy as relative prices change over time. In most cases, such interactions are of a second order of magnitude or have offsetting effects and do not affect the results substantially. Care has to be taken to notice major structural changes in an economy that may be occurring and causing significant shifts in the tax base. For example, as will be noted below, Ghana has experienced a major structural change over the last decade towards higher trade shares that make the forecasting of trade-based collections more critical in the overall revenue forecasts.

Two approaches can be used to macro- or GDP-based forecasting. The first approach is to estimate the growth in revenues that arise with real economic growth (or with the growth in a macro-proxy to the tax base) without any change in the tax structure (base, rates or exemptions.) This estimates the tax elasticity, l]Ty' If the tax elasticity exceeds unity, then tax revenues tend to grow faster than the real economic growth, and conversely, if the tax elasticity is less than unity. To estimate the tax elasticity of a tax type requires a time series

4

-

-

..

..

..

..

..

-

..

-

Page 7: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

..

-

of annual tax collections adjusted so that the tax structure is held constant over all the years. This means that the effects of discretionary tax changes in each year are adjusted for to keep the tax structure constant. Once the tax elasticity is known the revenues in the next year . R.+l. are forecast from the current year revenues, R.. if the real GDP or economic growth rate is gR, and the general rate of price inflation is 1t as:

R.+l = R., *(1 + TJTY ~)(1 + 1t)

Unfortunately, no lengthy time series of the tax impacts of discretionary measures over the years is available for Ghana tax revenues to construct the adjusted or constant-tax-structure revenue series. Hence, the second approach is followed to estimate aggregate tax type revenue growth.

The second approach estimates the growth rate in the tax base (or proxy for the tax base) in terms of real economic growth and any other explanatory variables that determine the growth in the tax base. Here. the elasticity of the tax base with respect to real economic growth. TlBY. and other variables (such as real price changes. TlBP. etc), are used to estimate the growth in the base, and tax revenues are then taken to grow proportionately to the base growth arising from the economic growth and changes in other explanatory variables. Hence. again revenues are forecast as follows .

R.+l = R.,*(l + TlOY ~ + TJ BP gp )(1 + 1t)

Once estimates of the elasticity of the tax base to real economic growth and real changes have been established, then forecasts of real economic growth (~). real price changes (gp) and general price inflation (1trare required. Ideally. basic macroeconomic variables (real economic growth rate. inflation rate, exchange rate. interest rates. etc) should be derived from macroeconomic planning and forecasting models and should be consistent with expert expectations of key economic managers and planners in the Ministry of Finance. Central Bank and other central planning agencies. Alternatively, real growth in GDP could be itself predicted by means of some simpler growth model as illustrated in Appendix A. but this would be poor substitute for a macro model that contained both the major market behavioral relationships and national accounting consistency for the economy. Given a set of forecasts for macro-variables. it is useful to have estimates of how the major tax bases are expected to grow. Estimates are made here of the relationship between consumption and GDP and between imports and GDP and the real price of imports.

3.1 Consumption

Consumption forms the base of the V A T and excise duties. Private plus public sector consumption in Ghana averaged 92.6% of GDP over 1986-200 1. It generally remained in the range of 89% to 97% except when it dropped in 1994-1996 to a temporary low of 78.8%. This reflected a peak in the investment rate in 1996 and a sharp acceleration of exports over 1994-1996 that clearly generated an unexpected growth in income. When real consumption (private plus public) is regressed on real GDP based on annual data from 1986 through 200 1. an elasticity of consumption with respect to GDP of 0.67 is found. as shown in Table 3.1.

5

.,

..

-

of annual tax collections adjusted so that the tax structure is held constant over all the years. This means that the effects of discretionary tax changes in each year are adjusted for to keep the tax structure constant. Once the tax elasticity is known the revenues in the next year . R.+l. are forecast from the current year revenues, R.. if the real GDP or economic growth rate is gR, and the general rate of price inflation is 1t as:

R.+l = R., *(1 + TJTY ~)(1 + 1t)

Unfortunately, no lengthy time series of the tax impacts of discretionary measures over the years is available for Ghana tax revenues to construct the adjusted or constant-tax-structure revenue series. Hence, the second approach is followed to estimate aggregate tax type revenue growth.

The second approach estimates the growth rate in the tax base (or proxy for the tax base) in terms of real economic growth and any other explanatory variables that determine the growth in the tax base. Here. the elasticity of the tax base with respect to real economic growth. TlBY. and other variables (such as real price changes. TlBP. etc), are used to estimate the growth in the base, and tax revenues are then taken to grow proportionately to the base growth arising from the economic growth and changes in other explanatory variables. Hence. again revenues are forecast as follows .

R.+l = R.,*(l + TlOY ~ + TJ BP gp )(1 + 1t)

Once estimates of the elasticity of the tax base to real economic growth and real changes have been established, then forecasts of real economic growth (~). real price changes (gp) and general price inflation (1trare required. Ideally. basic macroeconomic variables (real economic growth rate. inflation rate, exchange rate. interest rates. etc) should be derived from macroeconomic planning and forecasting models and should be consistent with expert expectations of key economic managers and planners in the Ministry of Finance. Central Bank and other central planning agencies. Alternatively, real growth in GDP could be itself predicted by means of some simpler growth model as illustrated in Appendix A. but this would be poor substitute for a macro model that contained both the major market behavioral relationships and national accounting consistency for the economy. Given a set of forecasts for macro-variables. it is useful to have estimates of how the major tax bases are expected to grow. Estimates are made here of the relationship between consumption and GDP and between imports and GDP and the real price of imports.

3.1 Consumption

Consumption forms the base of the V A T and excise duties. Private plus public sector consumption in Ghana averaged 92.6% of GDP over 1986-200 1. It generally remained in the range of 89% to 97% except when it dropped in 1994-1996 to a temporary low of 78.8%. This reflected a peak in the investment rate in 1996 and a sharp acceleration of exports over 1994-1996 that clearly generated an unexpected growth in income. When real consumption (private plus public) is regressed on real GDP based on annual data from 1986 through 200 1. an elasticity of consumption with respect to GDP of 0.67 is found. as shown in Table 3.1.

5

Page 8: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Table 3.1. Log Real Consumption

Variable Coefficient t·StatisticCoefficient t-Statistic

Constant 3.559 1.2 7.379 12.3 Log Real GDP 0.671 2.0 Lagged Log Real Export Values 0.287 3.6 AR(1) 0.491 1.9

R-squared 59.0% 50.1% Adjusted R-squared 52.1% 46.3% Durbin-Watson stat 1.66 1.04 F-statistic 8.62 13.07 Data 1986-2001

The growth in consumption being less than the growth in GOP reflects the increase in investment as a share of GOP over the period. A similar result is found when the consumption base is reduced by the public sector wage bill in order to better reflect the taxable value of public consumption. This result implies that the nominal accrued V A T revenues can be estimated to grow as follows with real GOP growth of ~ and inflation of n:

Rt+l = R" *(1 + 0.67 gR ) (I + n)

3.2 Imports

The share of GOP in Ghana has grown significantly over the period from the mid-1980s to the early 2000s as illustrated in the figure below. This reflects a major opening up of the economy and significant structural shifts. Imports have grown from around 20% of GOP to about 50% of GOP over this period as shown in the graph below. Growth in the exports of goods and service lagged behind growth in imports, but accelerated in the mid 1990s and again sharply in 2000 when the Cedi experienced a major devaluation.

The high growth and large relative size of trade flows have resulted in the estimation of the gowth in the import base being very important for revenue estimation. In recent years, excluding petroleum taxes, some 42% to 46% of total tax revenues have been collected on international trade as import duties, V A T, excise duties and other trade duties and levies.

6

-

..

..

..

..

..

..

..

..

Table 3.1. Log Real Consumption

Variable Coefficient t·StatisticCoefficient t-Statistic

Constant 3.559 1.2 7.379 12.3 Log Real GDP 0.671 2.0 Lagged Log Real Export Values 0.287 3.6 AR(1) 0.491 1.9

R-squared 59.0% 50.1% Adjusted R-squared 52.1% 46.3% Durbin-Watson stat 1.66 1.04 F-statistic 8.62 13.07 Data 1986-2001

The growth in consumption being less than the growth in GOP reflects the increase in investment as a share of GOP over the period. A similar result is found when the consumption base is reduced by the public sector wage bill in order to better reflect the taxable value of public consumption. This result implies that the nominal accrued V A T revenues can be estimated to grow as follows with real GOP growth of ~ and inflation of n:

Rt+l = R" *(1 + 0.67 gR ) (I + n)

3.2 Imports

The share of GOP in Ghana has grown significantly over the period from the mid-1980s to the early 2000s as illustrated in the figure below. This reflects a major opening up of the economy and significant structural shifts. Imports have grown from around 20% of GOP to about 50% of GOP over this period as shown in the graph below. Growth in the exports of goods and service lagged behind growth in imports, but accelerated in the mid 1990s and again sharply in 2000 when the Cedi experienced a major devaluation.

The high growth and large relative size of trade flows have resulted in the estimation of the gowth in the import base being very important for revenue estimation. In recent years, excluding petroleum taxes, some 42% to 46% of total tax revenues have been collected on international trade as import duties, V A T, excise duties and other trade duties and levies.

6

-

..

..

..

..

..

..

..

..

Page 9: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

1/11

1/11

1/11

1/11

Trade over GOP, Ghana,1986-2001

la%t-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~------·

a%L---__ --__ --__ --__ --__ --__ --__ --__ --__ --__ --__ --______________ ~

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

I_,mports over GOP ___ ,mports G&S over GOP _Export G&S C1Vfl< GOP'

Estimates of the import price and income elasticities of demand are made from reponed annual import values (c.i.f) for 1986-2001. These nominal values were deflated by the import price deflator to obtain the real import values. The real price of impons was taken as the ratio of the import price deflaor to the overall GDP price deflator. Regression results presented in Table 3.2, show that the elasticity of imports to real GDP (1fMy) is high at 2.3. (This is both the elasticity of import values and import quantity to GDP, assuming constan! prices.) The elasticity of import value to the impon price (/]Mp) is -{).69. This implies the ealsticity of import quantity to the import price (T/Qp) is highly elastic at -1.69. as "'fP = I+T/QP (see Appendix B.) Imports into Ghana are both highly price and income sensitive.

The price of imports depends upon both fluctuations in the world prices expressed in foregn exchange and the exchange rate. If most of the import price changes are arising from the real changes in the exchange rate. Using the real exchange rate from the IMF IFS data series. but expressing the exchange rate in Cedis per foreign currenct unit. the elasticiy of the import value to the real exchange rate is estimated as -0.55, which is slighly smaller in magnitude than the elasticity of -{).69 based on the full real import price. This shows that he exchange rate movements are accounting for most of the import price changes .

7

1/11

1/11

1/11

1/11

Trade over GOP, Ghana,1986-2001

la%t-~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~------·

a%L---__ --__ --__ --__ --__ --__ --__ --__ --__ --__ --__ --______________ ~

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

I_,mports over GOP ___ ,mports G&S over GOP _Export G&S C1Vfl< GOP'

Estimates of the import price and income elasticities of demand are made from reponed annual import values (c.i.f) for 1986-2001. These nominal values were deflated by the import price deflator to obtain the real import values. The real price of impons was taken as the ratio of the import price deflaor to the overall GDP price deflator. Regression results presented in Table 3.2, show that the elasticity of imports to real GDP (1fMy) is high at 2.3. (This is both the elasticity of import values and import quantity to GDP, assuming constan! prices.) The elasticity of import value to the impon price (/]Mp) is -{).69. This implies the ealsticity of import quantity to the import price (T/Qp) is highly elastic at -1.69. as "'fP = I+T/QP (see Appendix B.) Imports into Ghana are both highly price and income sensitive.

The price of imports depends upon both fluctuations in the world prices expressed in foregn exchange and the exchange rate. If most of the import price changes are arising from the real changes in the exchange rate. Using the real exchange rate from the IMF IFS data series. but expressing the exchange rate in Cedis per foreign currenct unit. the elasticiy of the import value to the real exchange rate is estimated as -0.55, which is slighly smaller in magnitude than the elasticity of -{).69 based on the full real import price. This shows that he exchange rate movements are accounting for most of the import price changes .

7

Page 10: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Table 3.2. Log Real Import Values

Variable Coefficient t-StatisticCoefficient t-Statistic

Constant -9.616 -10.3 -9.829 -8.5 Log Real GDP 2.314 15.2 2.260 12.5 Log Real Import Price -0.691 -4.8 Log Real Exch. Rate (CdlF$) -0.552 -3.6

R-squared 96.6% 95.3% Adjusted R-squared 96.0% 94.5% Durbin-Watson stat 2.25 2.19 F-statistic 182.69 130.76 Data 1986-2001

In forecasting import values, these can be done either by forecasting import quantities (Q) and domestic import prices (pd= pWE), or by forecasting the US dollar value of imports (QpW) and the exchange rate (E). Typically it is easier to forecast changes in the workd prices of major commodities and the exchange rate separately, rather than as the product (pd= pWE). Information on the US dollar value of imports is generally avaialble on a fairly current basis from trade statistics and balalnce of payments accounts, and exchang rate information is known up to date. By contrast, the import quantity and import price deflator is derived from national accounts data that is only available with about a one to two year lag. Note, however, that even where the import price is being split into components, the price elasticity of imports that should be applied to changes from either world price or exchange rate component should be still the overall price elasticity from the combined import price (pd). See Appendix B for furthr clarification on this point. Hence, the growth in the real value of imports (gM) can be given as:

gM = (gpw + gE + gpw*gE) 'IMP + gy'!MY = (gpw + gE + gpw*gE )(-0.69) + gy(2.3)

Note that the growth in world prices (gpw) is in real prices (adjusted for foreign currency inflation) and the real devaluation rate or real change in the exchange rate (gE) has been adjusted for both domestic and foreign currency inflation. Accordingly, the nominal growth in the import values is given by gM! 1 + ff) where ff is the rate of domestic price inflation.

8

-

..

...

..

..

..

..

..

Page 11: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Appendix A

Simple GDP forecast

The standard approach to forecasting GOP is based on investment-led growth models. Hence, models are estimated of real GOP in terms of real investment based on 1986 to 200 1 annual data. As the results in Table A show, however, there is a weak and insignificant relationship between real GOP and real investment in Ghana. This surprising relationship is also evident in the graph below that shows the real growth rate and the investment to GOP ratio over 1986 to 2001. Real growth appears to be invariant to the growth in investment rate over time. This may indicate that investment needs to be disaggregated to identify the productive and non-productive types of investment in the economy. By contrast, the Ghanaian economy has undergone a major structural change as it has opened up to international trade and the trade to GOP ratios have risen dramatically over the years. Imports as a share of GOP, for example, rose from 20% of GOP in the mid-1980s to over 50% of GOP in the early 2000s. Exports also rose in line with imports although a major trade deficit has been sustained over the period. When the one-lagged real exports values are used as an explanatory variable of growth, a strong and significant relationship is found. Table A shows that a 10% increase in real exports has resulted in a 4.1 % increase in real GOP. This suggests Ghana has been experiencing export-led growth. In practice, a more detailed macro-model is required for reliable GOP forecasts.

Table A. Log Real GOP

Variable

Constant Log Real Investment Lagged Log Real Export Values

R-squared Adjusted R-squared Durbin-Watson stat F-statistic

Data

9

Coefficient t-Statistic

5.512 0.037 0.417

97.2% 96.8%

1.61 209.87

1986-2001

15.6 0.6 14.0

Page 12: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Investment and Growth, Ghana, 1986-2001

~%~------------------------------------------------------.--------~~------------------------------~

~%+----------------------------------

~+-------------------------~---- -~~-------------~------~

15% +-----------------------..~~.___I-----

10%~~~~~------------------

1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

I-t--Investment over GOP ~Real growt~

10

-

...

...

...

...

..

Page 13: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

..

Appendix B Formal Derivation Of Growth In Real Import Values

All prices are real or constant prices. Derivation shows that import values as show the standard result from sales revenues facing a price elasticity of demand. namely. that the price elasticity of revenues equals (1 + price elasticity of quantity demanded) .

M = pM Q = import value base in D$

pM = p W E = clf import price = (world price in F $)* ( exchange rate in D$ IF $)

pd = pWE(l + T )= domestic import price. where T = ad valorem equivalent of all indirect taxes

Q = Q( pd. y ) = demand function in terms of pd and GDP( Y )

dM = pMdQ+dpMQ

= pM [( OQI/ipd )dpd +( OQIOY)dY J+dp"'Q

dM 1M =[( OQI/ipd X pd IQ)]( dpd I pd )+dp" I pM +( pdOQIOY K Y I M K dY IY)

dM 1M = 1JQp'( dpd I pd )+dp" I pM +( 1JMY X dY IY )

dM 1M = [1JQp' + \]( dp" I pM )+1JQp,dT 1(\ + T )+( 1JMy X dY IY )

dM 1M = 1J"pA dp" I pM )+ 1JQp,dT 1(\ + T )+( 1J"y X dY IY )

where:

1JQp' = import quantity elasticity with respect to import price

1J Mp' = [1J Qp' + \] = import value elasticity with respect to import price

dpd I pd =dp' I p' +dEI E+dT I(\+T )=dpM I pM +dT 1(I+T)

II

Page 14: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-

..

..

..

..

12 •

Page 15: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

....

..

..

COMPREHENSIVE FORECASTING MODELS FOR

INCOME TAXES2

I. PA YE MICROSIMULA TION MODEL 1. Introduction

Micro simulation models are widely used as analytical tools to estimate tax revenues and evaluate the impact of tax changes. The main advantage of micro simulation modeling lies in its capacity to make an impact analysis of policy proposals, especially their distributional effects on particular groups of taxpayers breaking them down by income, gender, age, or tax jurisdiction under the income tax system. For instance, these models allow the evaluation of tax benefits from proposed policy changes to low-, medium-, or high-income families. They also allow the evaluation of potential impacts of the changes in the tax structures on diverse groups, such as the elderly or families with children.

The salient feature of a micro simulation model is its capacity to assess the effects of a variety of policy options. Examples include changes in the definition of income. modifications in the size and scope of deductibles, revenue effects due to modifications in the tax rate structure, and the effect of population growth or changes in income. Thus these models provide useful handles for assessing the efficacy of policy proposals in achieving the stated policy objectives .

The micro simulation models also help in assessing the revenue impact of such policy proposals and making a more accurate overall revenue forecast for the future years.

2. Pay-As-You-Eam mlcro-slmulatlon model

The pay-as-you-eam (PA YE) personal income tax micro simulation model has been developed for Ghana using the Microsoft Excel spreadsheet software ("Ghana

Microsimultion Model - PAYE Non-AG.xls") based on randomly selected samples collected from the Inland Revenue Service district offices all over the country.

Just as in many other developing countries, the major difficulty in the development of sophisticated micro simulation models in Ghana lies in the lack of detailed and accurate taxpayer and household information on individual taxpayer basis. In Ghana. the IRS lacks a computerized system of data collection and data maintenance and all the records are still maintained through files. As a result, in order to get sufficient representative samples from the total population and development of micro-simulation models involved a demanding data collection process through field surveys. It cannot be over-emphasized that the quality of micro simulation results is highly dependent on the quality of the data used by the model.

2 Prepared for the Ministry of Finance & the Revenue Agencies Governing Board. Accra. Ghana by a Duke University Center for International Development team lead by Profs. Graham Glenday and G.P. Shukla with the assistance of Rubi Sugana, Anthony Doku and Carlos Delia Torres. The work was funded by the USAID Mission to Ghana through Sigma One Corporation Contract No. 641·C·OO-98-00229 over the period October 2002 through September 2003.

13

PREVIOUS PAGE BLANK

Page 16: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

3. Components of the PA YE micro simulation model

The Microsoft Excel-based ("Ghana Microsimultion Model- PAYE Non-AG.xls") non-public service PA YE personal income tax micro simulation model that has been developed for Ghana has four main components (Figure 1):

1. The Typical Taxpayer Personallncome Tax Calculator Model

2. The Personal Income Tax and Growth Parameters

3. The Taxpayer Personal Information and Ill(:ome Data (Sample Database)

4. The Aggregate Impact Distribution Output

The level of information provided under these four components determines the degree of analysis that can be undertaken.

Figure 1 - PA YE Income Tax Micro simulation Model Structure

Personal Income Tax Parameters

Current Tax I Proposed Regime Changes

!]. Taxpayer Q Personal Tax Calculator

Information

!]. Personal Income Tax

Liability

Current Tax Proposed Regime Changes

In addition to the above components. the Excel file is also equipped with a simple macro module (computer programs), which is used to automate the simulation of income tax calculation for thousands of sample taxpayers stored in the database.

3.1 The Typical Taxpayer Personal Income Tax Calculator Model

The Typical Taxpayer Model calculates the tax liability of a typical individual following the logic as stated in the income tax laws. A typical taxpayer may include a young student, an elderly retired person, a couple with children, a single person, or a disabled person. This

14

..

-

..

..

..

..

..

Page 17: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.. model simulates the impact of proposed policy changes on the tax liability of the individual taxpayer.

The underlying idea of a typical tax calculator model is to understand the application of tax laws on individuals under the existing tax regime and to compare it with the tax liability of the same individual under a proposed tax policy change. This model may become rather complicated as the current or proposed tax structures become more complex. The impact of the tax policy change proposal on the individual is the difference between the individual's current tax liability and the tax liability if the proposed changes were to be implemented.

The implementation of the personal income tax calculator is modeled in the "Calculator" worksheet of the Excel file, as also shown in Annex 2. In this model, the chargeable income is obtained by subtracting personal relief/allowances from gross assessable income from all sources minus exemptions. Tax liability is calculated by applying the chargeable income to the marginal tax rate schedule. Finally, the tax due is calculated by subtracting tax credits from total tax liability.

The formulas in the tax calculator model are linked to the personal income tax parameters. Hence, to evaluate the impact of a tax policy change 00 the individual tax liability (e.g. changing the income brackets and tax rates), one needs to modify the value of income bracket and tax rate parameters only in the "Parameters" worksheet under the "Option" column without having to modify the calculator formulas.

Table I details the calculation of income tax liabilities based on a progressive tax rate structure as implemented in the model. In this example, a proposal was made to change the tax rates and their corresponding income brackets. The first four columns on the left calculate the tax liability under the current tax structure, while the next four columns calculate the tax liability for the same individual under the proposed tax structure.

Table 1 - Detailed Tax Liability Calculator Model (in tlwusands oicedis) BASE OPTIOH

lCUrrwlll S!I -Tax Uability before Credit _. -.... ....... ...... - - - -5.000 5.000 "" 0 •. 000 6.000 "" 0

3.000 8.000 ""'- 600 2.500 8.500 " .. 375

aooo 11,000 "'" 900 3.500 12.000 25 ... 875

14,101 lJ' .... 5.641 13.101 LP 35" .'" 25,101 7,141 25.101 ..... ·1.305

The formulas behind the cells that show income tax rates and their corresponding brackets are linked to the income bracket and tax rate parameters in the "Parameters" worksheet. Simulating the proposed policy changes, therefore, is done by changing the values in the table of parameters.

Using the example above, a taxpayer with a total chargeable income of ¢25, IO I would benefit from the proposed tax changes. The taxpayer's tax liability would decrease by It 1.305 as shown in the last column should the income brackets and their corresponding tax rates have changed according to the proposal.

When modeling the tax calculator, it is important to understand the difference between tax credits and deductions. In a progressive tax system, a deduction will result in a tax

15

Page 18: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

reduction based upon the marginal tax rate of the taxpayer. This results in a higher tax reduction for a high-income taxpayer compared to a lower income taxpayer for the same amount of deduction. Tax credits, on the other hand, provide the same relief regardless of the income of the individual. In a typical tax system, credits lower the tax payable unit by unit (i.e., dollar by dollar), while deductions reduce it by an amount equivalent to each taxpayer's marginal tax rate. It is for this reason that many countries have gradually switched to a credit system from a deduction system.

Using this typical taxpayer model, various scenarios of tax policy change proposals - such as changes in the definition of income, modifications in the size and scope of deductibles, and rates structure - can be simulated. Hence, the impact on the tax liability of an individual taxpayer because of the policy changes can be evaluated.

One simple example would be to apply a 25% flat tax rate scenario into the tax system. For simplicity, the parameters that have to be changed in this example are only those related to the tax rate schedule. The tax liability calculation under this scenario for the same taxpayer as in the previous case is shown in Table 2.

Table 2 - Tax Liability Calculation under Flat Tax Rate Structure (in thousands of cedis)

BASE !Currentl

Tax Uabillty before Credit Income .. Income .. B~cket Bracket Ra'a Bracket Bracket Rate.

5.000 5.000 0% 0 .,000 6,000 25%

3,000 6,000 20% 600 2,500 8,500 25%

3,000 11,000 30% 900 3,500 12,000 25%

14,101 UP 40% 5,641 13,101 UP 25%

25,101 7,141 25,101

OPTION !PfODOsedl

Impact

',500

625 875

3,275

6,275

Under this scenario, the taxpayer is still better off compared with that under the existing tax structure, although the tax benefit is not as much as that under the previous proposal.

To automate the simulation of the tax liability calculation for each individual taxpayer in the sample database, a computer program (macro) has been developed, This program will read from the database the personal information and income data of each sample taxpayer one record at a time, calculate hislher the tax liabilities under the current and proposed tax structure using the formulas written in the "Calculator" worksheet, and put the calculation results back in the aged database in column 23 and 24 (Tax Payable - Base and Tax Payable - Option)

To run the program, simply press the "Run Simulation" button which appears in the "Calculator" worksheet.

3.2 The Personal Income Tax and Growth Parameters

The personal income tax and growth parameters contain detailed information regarding statutory tax parameters - such as income brackets, tax rates, the amount of allowable deductions and other modifiable policy parameters - as described in the current tax regulations, and their corresponding values as proposed in the policy change proposals.

16

..

..

..

..

..

..

Page 19: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

fI

Annex 3 shows the personal income tax parameters used in the model as presented in the "Parameter" worksheet. The "Parameter" worksheet also contains growth factors which are used to "age" the sample database from the sample year into the observation year .

As mentioned earlier, the personal income tax and growth parameters are used as a "look-up" table from which the tax calculator picks up the appropriate tax rates, allowable deductions, and tax credits based on the personal information of the individual taxpayer in order to assess his/her total tax liability. This component is intentionally separated from the Excel formulas in the tax calculator model to provide a greater flexibility in the process of running the various tax policy simulation exercises.

3.3 The Taxpayer Personal Information and Income Data

The sample database for this model was constructed using the data collected from the randomly selected samples of taxpayers registered in the IRS district offices across Ghana. It is important to note, that to maintain the integrity of the model, the samples must be selected randomly. The sampling design and the outcome from the field survey exercise are given in detail in Annex I.

The taxpayer personal information contains required data to determine the taxpayer income bracket and its marginal tax rate, and to identify whether the taxpayer is entitled to a certain relief/deductions or other preferential tax treatments (e.g. lower tax rates). This information includes taxpayer identification, age, marital status, number of children in school, wages and salaries, investment income, etc. This information is stored in the "Database" worksheet in the Excel file.

To facilitate the aggregate impact distribution analysis (i.e. to analyze which groups in the society would be the winners and losers from the proposed tax policy changes). the personal information includes some categorizing data. These categorizing data include age group. income class, gender, and District Office or region of residency. This type of categorization of data allows the impact analysis of the tax policy changes on a certain group. The results of this analysis will be invaluable to guide the government in making necessary preparation. such as taxpayer education campaigns, before implementing the policy change proposals. To the extent possible, it will be useful to identify whether or not the taxpayer is among the

politically sensitive groups.

Since most of this personal information would be collected from historical data. some necessary adjustments need to be made to update the personal information into the observation year.

Sample Weight

Sample weight is an important concept in data sampling. It is used so that conclusions drawn from the analysis based on the samples can be scaled up to the total population. The sample weight indicates the number of taxpayers, or potential taxpayers, in the population with similar characteristics that is represented by the sample. For example, in the stratified sampling,3 if the total population with similar characteristics in a given stratum is known. say

1 Stratified sampling is done by subdividing the 10la1 populatioo into partirular groups 0'" straIa. aDd each unit 0( the ~btioo ~ auig.ned 10 a unique stratum according 10 its specific nature. lbe purpose of stratifical:ioo is to increast' the efficitocy of ~ by dn"'1dlng :I

17

Page 20: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

1,000, and the number of sample which was randomly drawn from that stratum is 4, the weight for each sample is 250. Hence, if the tax liability of one of the samples is ¢7,OOO, then the total tax liability for the total population which is represented by the sample is estimated at ¢I,750,OOO (250 x ¢7,OOO).

heterogeneous universe is such a way that (a) there is some degree ofhomo~:eneily; and (b) a somewhat marked difference is possible between strata.

The strata may be created more than one level deep that is one stratum may be subdivided further into several strata. After the grouping is done, the sample within a stratum is created using either systematic or random sampling. The number of items taken from each group may be in proportion to their relative weights. The weights are used to scale the sample to represent the population. Given the fact that this is a method with an equal probability of selection of items, it is also known as proportionate stratified sampling.

For example, a total of one hWldred tax returns are divided into four strata. lFor various reasons, the weights of the strata are determined to be 10,20,30. and 40%, respectively. A total of 30 samples are to be drawn for detail tax verification. Thus. the desired proportional sample can be obtained as follows:

From the first strata: 30 x 10% 3 From the second strata: 30 x 20% 6 From the third strata: 30 x 30% 9 From the fourth strata: 30 x 40% 12

This proportional stratification is simple and satisfactory if there is no great difference in dispersion from stratum to stratum.

Alternatively, sampling fractions can be set at different rates for each stratum. The use of different disproportionate stratified sampling leads to an unequal probability of selection. Disproportionate stratified sampling requires additional work on the part of the researcher, as the responses have to be weighted for the analysis. However. the payoff in tenns of increased precision may justify the additional work. The use of proportionate stratification ensures the adequate representation of relatively small. but especially relevant groups, in the sample. In contrast, disproportionate stratification permits the analysis of particular members of a given stratum.

18

...

...

II

II

II

Page 21: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

II

-

.,

Table 3 - Sample Weight and Weighted Impact

'- - - -- ... - - ,-- ,-- - , ....... -,.:: -=- -m - - - -, • • • , - - - - - - -- ~ -- -- -- -- --- - L'!a._ _.- ~ ~ . - - . - - Lft;I_~ - ~ ~ ~ ~ - --- • ,- - - n_ u_ - ~ - - -- -- -- --- -- - - - - -- -UD __ - ,- ~ - -- -,~- - - - -- --- -.~- -.- -- -- - -- - .~- -.- - - -- --- -.~- -.- ~- ~ -- ---. -.~- -.- ~ ~ - -_. -.~- -.- .- .- -- --_. -.~- -.- - - - -- -.~- -.- ~ ~ -- --_. - .~- -.- - - - -- --- - -- -- --- ---_. - .~- -.- - - -.- --- -~- -.- - - ~- ~-- - .~ ..... -.- --- --- - -- - - -- -- -- --Table 3 shows partial data of the first twenty samples of taxpayer records from the taxpayer information database. Column 2 contains the sample weight for every record in the database. Column 28 contains the total impact of the tax policy change proposal on the whole population represented by each sample. The weighted impacts in Column 28 are multiplication of sample weights (Column 2) times impacts on the individual tax payable (Column 25).

The determination of a sample weight sometimes can be complex. After the sample is drawn. or surveyed, the distribution of sample with certain characteristics must be compared against the whole population. If the sample distribution is skewed (too many or too lillie sample with certain characteristics), the sample weight must be corrected, or some samples must be dropped.

The sample weights used in this model are estimated by comparing the total actual tax collections by regions in 2002 with total tax liability calculated using the information in the sample database. This calculation is shown in the "Weigh'" worksheet.

DoJoAgeing

Most input data, like the one used in Ghana, for tax policy simulations comes from the past -usually the latest available historical data lags behind the time period into which the analysis has to be made, while estimates are needed for future years. Because of this, data "ageing" and updating of the latest available data to future periods are required before running the simulations.

The process of data ageing can be completed using simple, static growth factors; or using more complex, longer-term dynamic methods, which take into consideration behavioral response due to changes in tax policies. For most developing countries - as also in case of Ghana - the static method is usually sufficient to be adopted in the early stage of micro simulation model development to evaluate short-term effects of tax policy changes. The static method uses information from macroeconomic and demographic forecasts. such as real GDP, inflation rate, population, and labor force growths, to determine the growth factors used in the data ageing process.

19

Page 22: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

The process of data ageing using the static method is typically done in two stages, with some interactions that may impose iterative procedures:

a. Projection of Populations

Depending on the structure of the existing tax system and proposed tax policy changes, the projection of populations may include the estimation of growth/reduction in the labor force, increase in home ownership, pensioner and children population, urban/rural population, number of married couples, gender and age distribution, etc.

b. Projection of Income Levels

The projection of income levels includes the estimation of growth in wages and salaries, self-employment income, income from savings or investments, and mortgage interest.

Table 4 - Examples of Growth Factors Used in the Model

Population 2% Wage 12% Investment NIA Consumer Price 10%

The nominal growth factors presented in Table 4 are suggested for use in the model to age the database to the observation year. Concerned government agencies or research institutes may be able to provide more precise estimates for these factors.

The different factors are used to grow different sets of infonnation in the database. The population growth rate is used to transfonn the population base data to represent a future year. In this case, the population growth factor is used to adjust the sample weight with the assumption that the population growth is unifonn across the board and that the distribution does not change.

Table 5 - Unadjusted Taxpayer Information Database

' ....... W.l/OMIend Now ..... ..... ,.

Benellt, Benefn. ,.,.,

othetReIlel" r.xPayable W .... ~ ..... Ind. Seelor -- --" "'- 80~ <I<Illon

Vehicle 00 ... (>educllon Deduction. (B ... ) , , , • " " " " " ~ " " ,- - ~ - ---- ~ -- -~- -- -~- D"U""g

T.,. __

- ~ LTO_.occRA _I.RNNfCE 124.111._ 21,713.(181 29,137,913 . - ~. lTOINUlTl) ~~ 11,1 •• 1'117 -,- ~m - n. --~ -~~ 8.Q1,oIGl ,~

-~ = - ~ - ~~ ."'''' "..,., 1,_'. - 0.' - -~. ..... ~ m_ -- ~. LTO·.ICCRA -~. 11,_._ l,12!1,SOO - ~. em ..... ~~ -.~ '''.-'-- ~. em ..... IIINNG .. QUMlRYIHQ '_.141 4:l,513.15C - ~. em ..... _I.QUoIfIRYIHG 1.',., .... ~ _.

~. LTO(!"IoI\I) _ .. OUNIRYINO '_',1187 3,Il00,000 1,""".000 51.1:.,3511

20

-

..

..

...

..

..

..

..

..

-

Page 23: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

,.

..

Table 6 - Aged Taxpayer Information Database

'-- ...... -- -- - ,- - - ,- -- ,--D - - '-.;;,;; - -. .= - -, , • • " " .. .. " ~ " " - - - - ~ - -- -- -- -- - --- ~ ~'IO.-", _.- -, ....... - • • ~ • ~ - ~ LlO .... " -~ ~ -- - • -~ - --- • - .- ~ - - -- ~ - -~ ~ - - --- ~ - - -- • .,- -- ~ LlO·_ - ---- - • ,-- ~ = ...... - -- • • ~ • -- ~ ,~ ...... _.- -- - • --- ~ ~ ...... _.- - • • - -- ~ ,~ ... _.- ~ - - •. -

Using the growth factors in Table 4 above - as also shown in cells C48:G51 in "Parameter" worksheet, the weights in the unadjusted/original taxpayer infonnation database. stoted in the "Database" worksheet (fable 5), is multiplied by 1.02 (or 2% growth) to get the weights in the aged database as presented in "Aged Data" worksheet (fable 6).

Similarly, the wage and pension incomes for a future year are estimated by multiplying the existing wage and pension by the wage growth factor (12%), Income from investment activities is adjusted by the investment growth factor. Other incomes and expenditures are adjusted by projected Consumer Price Index.

3.4 Aggregate Impact Distribution Analysis

The aggregate impact distribution analysis can be done by summarizing the total impacts of discretionary changes on individual tax liabilities by taxpayer category (e.g. age group. income class, gender, and District Office or region of residency) using the powerful E,cd tool: Pivot Table'.

Annex 4 shows results from aggregate impacts by district office and income group of replacing all personal relief/allowances by a basic tax credit of 300,000 cedis. It can h" '<'Cn from this table that the lower income groups will benefit more than those of the hIgher

income groups, and the total PA YE income tax revenues would decline by approximately 60 billion cedis due to the discretionary changes .

4. Results validation

Any micro-simulation model is likely to contain errors of estimation caused either h, hiJses in the sampling process, poor quality of the input data or a failure of some of the COI1lI","~nts of the underlying model. Once the models are designed and the analysis is carried out. a validation system is needed to check on the likely margin of error.

There are basically three validation techniques for obtaining infonnation about the quality of micro-simulation outputs:

4 Detailed descriptions regarding Pivot Table and its use can be found in the Microsoft Excel User's Manual or on-line' hdp

21

Page 24: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

a. External Validity Studies: Modeling output is compared with external sources to ensure objectivity.

b. Sensitivity Analysis: Models are evaluated under different alternative of input assumptions to study the effects of changes in the specific components of the model. These analyses enable the assessment of relative effect of changes in each component on the overall outcome.

c. Computer Statistical Techniques to measure the variance of model estimates.

Continuous model validation is an integrated part of the micro-simulation processes and enhances the reliability of the model output. Validation serves to overcome the natural limitations faced in micro-simulation modeling. One major limitation is the possible effects of policy changes on the behavior of taxpayer units affected by the policy. These responses introduce certain elements of uncertainty that would alter the model outcomes.

II. PAYE PUBLIC EMPLOYEES

The approach and structure used in the development of non-public service PA YE micro simulation model can also be used to develop a micro simulation model for the public service PA YE income tax. It was, however, not attempted dllring the course of this exercise for two reasons. First, the data pertaining to the public service employees maintained by the Accountant General's office does not have any information on allowable deductions and exemptions. It is also not clear from the database the amount of annual tax liability paid by each registered taxpayer. Also, there is no information about the age, marital status, family size etc. and this makes it difficult to do an impact analysis by categories. The data concerning certified employee deductions is maintained separately from the main payroll information.

Second, the distribution of income and income tax paid by the public employees (Excel file "Output Accounting General.xls") shows that bulk of the tax revenues are paid by employees with annual income between Cedis 4,400,000 and 28,000,000. The number of these employees is precisely known and therefore the revenue forecasting can be easily done without going through the sampling process. The revenue impacts due to income bracket and tax rate changes can also be estimated using the data on income distribution provided. However, as the database of the AG employees becomes complete in future and more information is gathered about the public employees after suitably revising the tax returns or amendments are made to the structure of personal deductions and/or the IRS introduces the system of regular filing of the returns by the employees, the same micro-simulation approach may be adopted for this category as well.

III. SELF-EMPLOYED

The data collection exercise through the field survey aimed at collecting data for the self­employed group as well. However, against a targetedl sample size of 1,497 (which is about

22

...

..

..

..

..

...

..

-

Page 25: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

.,

..

..

2.3% of the whole population), only 350 samples could be collected which is less than 25% of the planned sample size. This poor data yield is a result of several factors:

I. The database of taxpayers kept by the IRS Headquarters is not up-to-date and therefore the sample selection which was entirely based on database provided by the IRS did not meet its purpose.

2. A good number of files in the sample were missing and were either not found or were in a dormant state in most district offices.

3. Most of the files lacked financial statement and tax assessment. The lack of standard and structured tax returns has made it impossible to reconstruct and analyze the tax assessment.

4. Many files had not been examined since 1998 and the tax assessment was clearly quite low .

5. Inadequate number of qualified staff in field offices contributed to the problems in data collection.

As a result of these practical difficulties, the data collection for this category of taxpayers turned out to be inadequate and less than satisfactory and therefore basing the revenue forecast on this type of data is clearly not advisable. It is suggested that the IRS introduces structured tax returns in a standard format, not only to enhance the quality of the tax analysis and revenue projections, but also to improve the overall tax administration and compliance .

IV. COMPANIES INCOME TAX

The data collection, cleaning and inputting of data for the corporate sector is still in the process. So far, about 1,750 company data collection forms have been collected and entered into the computer. Each individual form was designed to capture financial statements and ta.'(

assessment of one company for one fiscal year. The company data collected was from the last five fiscal years (1997-2002).

Using the sample data already collected, the Duke Team has conducted preliminary data analysis and generated summary statistics (Excel file "Company Data Analysis.x1s") based on industrial sector, assets size, and chargeable income. It makes sense, however, to continue and complete the conduct of meaningful data analysis after the whole data collection process is completed. As the RAGB completes this process and goes onto the next stage of anaI}sis. the Duke team will be extending all the necessary support in this process .

23

Page 26: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Apnex 1 Sampling And Data Collection Methodology For Building PA YE Micro simulation Model

Target Number of Samples: 2,500 - 3,000 Taxpayers

This note describes the sampling methodology for the construction of an individual income tax (pay-as-you earn, PA YE) micro simulation model for Ghana. A multi-stage, purposive random sampling technique is used to select samples of individual taxpayers from the Internal Revenue Service (IRS) filing system. The proposed approach takes into consideration the manual nature of the IRS data collecltion and database system.

I. FIRST STAGE SAMPLING - Define Zones (PUIrposive)

Following divisions used in the Ghana Living Standard Survey (GLSS), Ghana is divided into three agro-ecological zones: the coastal plain, the middle semi-equatorial forest, and the northern savannah. Five administrative regions are located exclusively in a single zone: Greater Accra in the Coastal zone; Ashanti in the Forest zone; and Northern, Upper West, and Upper East regions in the Savannah zone. Three regions cut across two zones: Western and Central regions are partly in the Coastal zone and partly in the Forest zone; and Brong­Ahafo is partly in the Forest zone and partly in the Savannah zone. Eastern and Volta regions are located in all three ecological zones.

Table 1. Ghana Agro.Ecological Zones

Coastal Zone Greater Accra Region Forest Zone Ashanti Region Savannah Zone Northern, Upper West, and

Upper East Regions Coastal-Forest Zone Western andl Central Regions Forest-Savannah Zone Brong-Ahafo Region Coastal-Forest-Savannah Eastern and Volta Regions Zone

Economic activities in Ghana are concentrated mainly in Coastal and Forest zones. This situation is also reflected in the distribution of PA YE 1:axpayers registered in the IRS. PA YE taxpayers in Greater Accra and Ashanti regions make up for more than 90 % of the total taxpayers. For the purpose of this sampling exercise, the reported number of PA YE taxpayers per District Office is adjusted to reflect a more realistic average tax yield per taxpayer. The average tax yield per taxpayer at the LTO District Office is used as the basis for calculating the estimated number of taxpayers in each District Office (see Appendix I).

24

..

...

...

..

-

Page 27: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

"

Table 2. PA YE Taxpayer Distribution by Region

Agro-Ecological Zone Region Est. No. of % Taxpayers

COASTAL (79.36%) LTO 9\3,087 5752%

Greater Accra 346,746 21.84%

FOREST (10.19%) Ashanti 161,815 10.19%

SAVANNAH Northern 6,301 0.17% (0.64%) Upper-East 2,640 0.08%

Upper-West 1,274 0.40%

MIXED Western 113,853 7.17%

(9.81%) Eastern 15,058 0.95%

Central 14,069 0.89%

Brong-Ahafo 7,881 0.50%

Volta 4,838 0.30%

TOTAL 1,587,562 100.00%

II. SECOND STAGE SAMPLING - Select District Offices (Purposive)

The next stage in the sample selection process is to select IRS District Offices in each zone/region from where taxpayer information will be collected. Since the number of taxpayer at the LTO-PA YE District is so large relative to that of other districts, LTO-PA YE District is treated separately in the sample selection (see Table 3). The District Offices are selected to represent urban and semi urban areas, and - to the extend possible - the distribution of taxpayers in the selected District Offices matches with that of the national distribution across zones (see Table 4).

III. THIRD STAGE SAMPLING - Select Employees (Random)

In the District Office, employees' information are filed and grouped together in one folder by their respective employer. Employers' files are sorted alphabetically from A to Z in the filing cabinet. Within an employer file (IT FORM 51), data on monthly income. reliefs/deductions. and income tax paid for every employee is recorded and numbered sequentially. The data is sorted by the employee's rank/grade in the company.

Employees are selected as samples irrespective of their employer. They will be selected from the employee main files of each employer as if they are all employed by one employer and the employees are sequentially numbered from I for the first employee of the first employer through the last employee of the last employer.

25

Page 28: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.. Table 3. PA YE Taxpayer Distribution by Region, Excluding L TO

Agro-Ecological Zone Region Est. No. of % Taxpayers

COASTAL (51.41 %) Greater Accra 346,746 51.41%

FOREST (23.99%) Ashanti 161,815 23.99%

SAVANNAH Northern 6,301 0.93% (1.51%) Upper-East 2,640 0.39% ...

Upper-West 1,274 0.19%

MIXED Western 113,853 16.88% (23.08%) Eastern 15,058 2.23%

Central 14,069 2.09%

Brong-Ahafo 7,881 1.17% • Volta 4,838 0.72%

TOTAL 674,475 100.00%

.. Table 4. PA YE Taxpayer Distribution in Selected Districts

Zone Region District OffiCI! Est. No. of % Taxpayers

COASTAL Greater Accra Adabraka 24,344 5.49% (51.17%) Makola 11,793 2.66% ..

Terna 64,666 14.58%

Osu 65,117 14.68%

Legon 53,530 12.07%

Teshie/Nungua 7,588 1.71% .. FOREST Ashanti Obuasi 107,966 24.33% (24.33%)

SAVANNAH Northern Tamale 6,226 1.40% (1.40%)

MIXED Western Tarkwa 64,201 14.47% ..

(23.09%) Sefwi-Wiawso 12,093 2.73%

Eastern Koforidua 6,934 1.56%

Asarnankese 4,718 1.06%

Central Cape-Coast 6,991 1.58%

Brong-Ahafo Sunyani 4,073 0.92%

Volta Ho 3,429 0.77%

TOTAL 443,669 100.00%

...

26 -

Page 29: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

*'

..

Table 5. Number of Samples per District Office

District Office Est. No. of No. of %

Est. Sample Random Taxpayers Samples* Wei2bt Number**

LTO-PAYE**' 603,087 500 16.74% 1,206 436

Adabraka 24,344 110 3.68% 339 129

Makola 11,793 106 3.55% 169 79

Tema 64,666 292 9.78% 339 184

Osu 65,117 294 9.85% 339 290

Legon 53,530 241 ! 8.07% 339 261

TeshielNungua 7,588 137 4.59% 85 i 10 I ·Obuasi 107,966 ! 243 , 8.14% I 665 110 "

Tamale 6,226 I 112 3.75% . 91 72

Tarkwa 64,201 I 289 9.68% 337 178

Sefwi-Wiawso 12,093 : 136 4.55% 135 31

Koforidua 6,9341 125 4.19% 84 24 -

Asamankese 4,718 : 106 3.55% : 67 10

Cape-Coast 6,991 I 126 4.22% 84 10

Sunyani 4,073. 92 3.08% i 67 2

Ho 3,429· 77 2.58% ' 67 3 • , .

TOTAL! 2,986 ,100.0% • Excluding employees with incoflU' greater than or equal 10 if JO(j.(XXJ,OOO u'hich art' not randomly uiuled .

Income tax data from all employees with income greater than or equal to It J((),OOO.OOO thaI art' nol selected randomJy will also be collected

** Random number is used to select the first sample from the employu main file based 011 the emplnyu number in the file.

*** Excluding civil servanls and other go\-'emmenl employees .

To select samples, each selected District Office will be given a random number 10 piCK the first employee from the employee main file, and another number to pick su"scquent employees (see Table 5). If the total number of employees from all employers regist·.·r~d III

the District Office is 10,000 and 200 samples will be selected, the District Office" III he given a random number between 1 and 50 to pick the first sample. Thereafter, subsequent samples are to be selected for every 50th employee in the employee main file. The subsequent number is equal to the total number of employees from all employers registered in the District Office divided by the total number of expected samples for that particular District Office. For example, if the random number generated for the District Office is 38 and the first employer has only 33 employees, the fifth employee from the second employer "ill he selected as the first sample provided that the second employer has more than 5 employ"",. If

27

Page 30: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

the second and third employers have IS and 42 employees respectively, the 40th employee from the third employer will be selected as the second sample and so forth.

In addition to the samples selected following the above procedures, the data of all taxpayers with annual assessable income of greater than or equal to ¢ 100,000,000 will need to be collected. Nevertheless, some taxpayers with income greater the ¢ 100,000,000 may well be selected following the random selection described above.

Steps for selecting samples from District Offices:

1. Fill out the "Count of PA YE Taxpayers and List of Samples per Employer" form for each employer in the District Office (see Appendix 2). This form should be filled in with data from all IRS District Offices in Ghana whether or not samples are going to be selected from the District Office. Filling out this form can be done simultaneously by more than one person.

2. After completing step 1, calculate the total number of employees for the whole District Office.

3. Calculate the "Subsequent Number" by dividing the total number of employees as calculated in (2) by the total number of expected samples for the District Office. For example, if the total number of employees is 6,000 and total number of expected samples is 100, then the "Subsequent Number" equals 6,000/100 = 60.

4. Using the Random Number assigned to the District Office (Table 5), select the first employee sample. For example, the Random Number assigned to Makola District is 79. If the first and second employers in Makola District have 48 and 60 employees respectively, the employee number 31 registered in the most recent employee main file of the second employer will be selected as the first sample. The employee number 31 is equal to the employee number 79 (the assigned Random Number) if all employees in the District Office were treated as employed by one company and numbered sequentially.

Write down the original employee number for that particular employer under the "Selected Samples" column of the "Count of PAYE Taxpayers and List of Samples per Employer" form (see Appendix 2).

5. Select employee number of the next sample using the "Subsequent Number." If the "Subsequent Number is 169 and the third, fourth, and fifth employers have 40, 80, and 26 employees respectively, the employee number 20 from the fifth employer will be selected as the second sample (see Appendix 3).

Write down the original employee number for that particular employer under the "Selected Samples" column of the "Count of PA YE Taxpayers and List of Samples per Employer" form.

6. Repeat step 5 till the last employer in the District Office. Once the filling out of the "Number of Employees per Employer" form is completed, the PA YE taxpayer data collection can be done simultaneously by more than one person in the District Office using the PA YE Data Collection Form (see Appendix 4).

28

...

...

..

...

..

..

..

Page 31: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

"

IV. FIELD DATA COLLECTION

1. Non-PublicServicePAYE

The PA YE data collection was done directly in the selected District Offices. A team consisting of officers from the RAGB Office and IRS Headquarters visited each District Office, trained the District Office staff, and supervised the data collection process. Taxpayer personal and income information were collected from the employee and employer files for the year 2002 and transcribed onto a specially designed form (Appendix 4).

Table 6. Target Number of Samples and Actual Yields

DISTRICT TARGET ACTUAL SAMPLES YIELDS

TOTALLTO SOO 465 93.0% LTO-ACCRA 268 LTO-KUMASI

LTO(MULn) 148

LTO(PMU) 48

ADABRAKA 110 71 64.5% ASAMANKESE 106 57 53.8% CAPE-COAST 292 0.0% HO 294 11 3.7% KOFORIDUA 241 0.0% LEGON 137 30 21.9% MAKOLA 243 40 16.5% OBUASI 112 23 20.5% OSU 289 0.0% SEFWI WIA WSO 136 126 92.6% SUNYANI 125 57 45.6% TAMALE 106 92 86.8% TARKWA 126 0.0% TEMA 92 107 116.3% TESHIDNUNGUA 77 88 114.3%

TOTAL 2,486 702 28.2%

Table 6 presents the target number of samples and the actual sample yields collected from the field. The poor performance of the fieldwork was due to the following reasons: I. Poor Filing System - a good number of the files in the sample could either not be traced

or were found to be dormant in most of the district offices.

2. Insufficient Data Quality - information kept in the files are not sufficient to support the recalculation of the tax liability of the taxpayer. or to identify whether the taxpayer is entitled for certain deductions.

3. Non-Standard Format - information are not kept in a standard tax returns-type format that allows data transcriber to follow the tax liability calculation.

29

Page 32: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-2. Public Service PA YE from the Accounting General Office

In addition to collecting PA YE data from the District Offices, an effort was also done to collect public service PA YE data from the Accounting General Office. The data collected "-

from the Accounting General office was not sampled, but for the whole population. Table 7 present the summary statistics of collected public service PA YE data.

Table 7. Summary Statistics of the Public Service PA YE Data in Cedis ..

0 92 0 0 0 11,046,269

I 1,852 54,474,363 182,416 54,656.779 7,102,258 12.99%

2 1,337 92,736,344 536,550 93,272,894 12,340,762 13.23%

3 855 116,695,276 1,349,937 118,045,213 15,867,245 13.44%

4 1,476 389,612,291 4,573,628 394,185,919 18,946,705 4.81%

5 18,002 5,105,155,869 985,290,607 6,090,446,476 313,503,114 5.15%

6 29,764 9,992,870,373 2,372,369,717 12,365,240,090 733,279,321 5,93%

7 60,089 30,615,240,457 1,145,141,509 31,760,381,966 2,242,987,268 7.06%

8 50,912 42,688,306,497 2,079,699,488 44,768,005,985 4,195,289,899 9,37%

9 54,135 55,365,955,915 1,954,140,448 57,320,096,363 5,843,612,309 10.19% .. 10 34,165 44,193,394,580 2,539,435,073 46,732,829,653 5,010,448,308 10.72%

11 26,775 40,494,456,700 2,497,542,434 42,991,999,134 4,835,278,157 11.25%

12 11,577 20,061,433,962 1,229,238,472 21,290,672,434 2,473,973,333 11.62%

13 3,884 7,506,554,589 646,870,569 8,153,425,158 965,128,056 11.84%

14 836 1,702,117,137 374,229,372 2,076,346,509 258,042,677 12.43%

15 750 1,840,559,435 227,258,997 2,067,818,432 280,613,354 13.57% • 16 376 994,826,024 198,541,322 1,193,367.346 164,934,983 13.82%

17 230 614,568,998 180,032,487 794,601,485 118,912,892 14.97%

18 71 184,152,683 85,855,108 270,007,791 38,521,551 14.27%

19 175 370,801,092 371 ,970,312 742,771,404 121,839,748 16.40%

20 65 151,566,434 159,798,332 311,364,766 52,639,532 16.91%

21 30 72,889,728 81,992,662 154,882,390 28,499,106 18.40% .. 22 30 69,875,833 103,474,734 173,350,567 34,271,795 19.77%

23 29 68,063,848 116,508,235 184,572,083 32,793,376 17.77%

24 53 133,978,126 265,831,108 399,809.234 79,445,930 19.87%

Total 297,560 262,880,286,554 17,621,863,517 280,502,150,071 27,889,317,948 9.94%

Source: Accounting General Office. June 2003 .. Public service PA YE information gathered from the Accounting General Office includes monthly basic salary, monthly benefits received, and monthly tax liability. What is not clear from the Accounting General database is the information about the actual tax payment made by each taxpayer and for which months. The Accounting General database also does not provide information on allowable deductions and exemptions.

30 -

Page 33: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

APPENDIX 1

..

..

..

.,

.,

Page 34: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

APPENDIX 2

COUNT OF PAYE TAXPAYERS AND LIST OF SAMPLES PER EMPLOYER (This form has to be filled in for all IRS District Offices)

DISTRICT OFFICE:

Employer File Total Number List of Employee Numbers

No. Selected asSllmples Number of Employees

(onlv for selected District Office)

:

i

I

-

..

..

..

..

Page 35: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

" EXAlPLf FOR SflECTMG SAlFUS OF PAVE TAXPA'lfRS

..... ~ '''0.- ..... ....... ..... ~ - 5_4' ..... ...... ...... S .... ....... ...... ..... ~ ..... ..... • • • ...;. -;. • ...;. ..; . • ...;. _ . • , , , 2 ,. B2 • " '63 : • " ". "" , 2 2. 2 35 8J • " '" • ~ ~-, 3 3 2 -,; .. • 17 165 : • 1 • • 2 ., lIS • " 166 • .. /,., "" 1 , , 2 .,

" 7 • .. '67 : • B!!l!II'" 1M 1 • • 2 .. fSl • :!l "" • 21 26 1

1 7 7 2 '" '" • 21 I," iii • 11 19l 2 1 • • 2 " '" 1 • 22 171l ~ • Z3 25' J 1 • • 2 " ., 11 • 2J 171 • " 252 • 1 11l 11l 2 C .. 1 • " In ~ • " 253 , 1 11 11 2 .. 92 1 • 25 173 • >; "" 1 12 12 2 .. .,

" • >; '" : 1 13 13 2 .. .. 1 • 71 17' , " " 2 " 95 1 • ,. 176 : 1 15 " 2 .. .. 17 • ,. .n , ,. ,. 2 .. ., , • 3) 170 ,: 1 17 17 2 '" .. ;. • 31 179 1 ,. ,. 2 " .. • 32 '00 :~ 1 .. .. 2 52 'Ill ~ • 3J I •• 1 21l 21l 2 53

,,,, • ,. 182 :~ , " " 2 .. '02 Z • 35 103

1 22 22 2 ,. '03 ~. • -,; '" :: 1 2J Z3 2 ,. , .. • 37 '115 1 " " 2 51 '''' ~ • ., '0; 1117 , 25 25 2 .. 'iii • P 'fSl :~ 1 16 16 2 59 1117 2ll • .. "" , " 71 2 OJ ,(11 ,. • " '''' 11 1 ,. 2ll J 1 "" 3) • "

,., 111 , ,. ,. 3 2 111l J1 • C '91 " , 31 31 3 , 111 32 • .. '92 11 , J1 31 3 • 112 ~ • .. '" '" ,

" 32 3 • '" • .. , .. '" , 3J 3J , • '" 35 • " '95 "' 1 ,. ,. 3 1 '" -,; • .. , .. 11 , 35 35 3 • "' 37 • 6 ,., 11 , -,; -,; 3 • '" ., • '" "" :~ , 31 31 , ,. 11. P • " ,., ... , ., ., , 11 "' '" • 52 l1Il '" , ." ." ,

" 121l " • 53 2111 122 1 .,

'" , 13 '" " • .. ;m 123 , " " 3 " m " • ,. 2!l3 '" , " " 3 " 123 .. • '"

,.. :~ 1 " " 3 " '" .. • 51 >lS I .. .. 3 17 125 .. • SO "" IN , .. .. 3 •• ,,.

" • '" 2117 .'" 1 .. .. 3 19 m .. • 60 21lO '''' 1 " "

, 21l "" .. • 61 21l!l 131

• 1 .. «I , " '''' 50 • 62 2'. m

2 1 .. J 22 '3) " • 63 211 '" 2 2 '" 3 Z3 131 52 • .. m 1D 2 ,

" , " '32 53 • 65 '"

,,. 2 • 52 , 25 1D .. • " ". 135

2 • 53 , 16 13' ,. • 67 215 1-'; , • .. , " 135 56 • .. '" 131 , 1 ,. 3 2ll 1:1i 51 • '" '"

,., , • 56 , ,. 131 511 • 10 ". , .. , • 51 , 31 ,., 59 • 11 ". ,'" • 2 ,. ,. J J1 , .. 60 • n 220 '" 2 11 59 3 " '''' 61 • T3 221 '" , " 60 , J3 '" 62 • " m I"

2 13 61 3 ,. '" 63 • " m , ..

2 " 62 , 35 '" " • 16 m '6 2 " 63 , :Ii , .. 65 • n 225 , .. 2 ,. .. 3 J1 '45 66 • 70 216 ." 2 17 65 , ., , .. 61 • ,., 227 '«1 2 ,. 66 3 .. U1 50 • 00 22ll I., 2 19 61 ,

'" 1«1 .. • 1 229 ISO 2 :!l .. • • , .. 10 • , 23J 15' ,

" '" • 2 ISO 11 • 3 '" '52 '" 2 22 10 • 3 '" n • • 2J2 153

2 " 11 • • 152 T3 • • 23J '''' 2 " n • 5 153 " , • '" 155

2 25 T3 • • '" " 5 1 23S 156 , ,. " • 1 155 16 • • 236 157

2 " 7S or- n •• I • 156 n • • 2J7 '58 , '" 16 • 151 70 • I. ,.,

"" , ,. • I. 158 79 5 11 ,." ,m , 31 • 11 '59 00 • " "" ••• , ~ " • " '60 •• • 13 '" 162

I

, 00 1 • 13 '61 B2 5 " '" '63 , J3 .. 2 • " '62 03 5 15 '" ."

"

Page 36: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

ANNEX 2

2002 PAVE SAMPLE DATA COLLECTION FORM

District Office:

Employer Industrial Sector: Employer File No.: Employee No.:

Personal Information Gender

Age

Marital Status

Number of Children

Number of OId·Age Dependents

Qualified for Disabled Relief?

Sum of life Insurance Policy

Annual Assessable Income Wages and Salaries

Allowances + Bonuses

Benefits in Kind:

- Acoomodation

- Vehicle

- Others

- Total Benefits in Kind

ITotal Assessable Income

Personal Reliefs I Deductions IAnnual Deductions Claimed I

Tax Paid IAnnual Tax Paid

1. Male

1. Single

1. Yes

If sample is not selected randomly (selected as an addition because assessible income = 100 million cedis), check this box 0

2. Female

2. Married 3. Widowed

2. No

I

I

...

-

..

..

..

..

...

..

Page 37: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.. ANNEX 2

PERSONAL INCOME TAX CALCULATOR MODEL .. BASE 0PT10H

(cwo.,n l!:! 9 -_'n_ T_ID RNDGHH1.17

""""" TESHlEINI.IOGUA Industrial Sedof PROFESSIONAL SERVICES Age 0 Sex -;.....-) ..

." Marital StaIuo CM-_ S-SO>gIa; w.w_ 0

Ntmber of Children - 0 Nwnber of OId-Age Dependents 0

Disabled Penon - (l.Y~o-No) 0 lie Instnnee Sum InsI.nd 0

Asse able Income Wages and SaBies 9.52V.495 0 Oirector's Fee 0 0 -- 0 0 Benefits • Aoootrw.1OdatitA I 0 0 Benefits • VeNde 0 0 _.Othe<s 0 0 _Income 0 0 Pension Income 0 0 Gross_ o 0

Total Gross A. b!e Income ......... ...... ... 0

• Exemption Inlefest Income 0 0 0 _Income 0 0 0 Gross_ o 0 0

T_ Exemption 0 0 0

Gross Income Minus Exempllon ......... ......... 0

.. _ Relief I Allowances I Deductions Standard Personal Tax Reief 0 0 0 ---... 0 0 0 C-..,-... 0 0 0 OklAge-... 0 0 0 __ AIowanoe

0 0 0 Disablement Reief (% of klcome) 0 0 0 ute Insurance Deduction ~,O 0 0 0 Social SecOOty Deduction " ~~ 476,475 476,475 0 .. Other Personal Refiefs I Oed. .. 0 0 0

Total Personal Allowances 471,475 471,475 0

Taxable Income 1.053,021 "053.112'1 0

Tax lJabjllly before CredH Income .. Income .. _ ... ....... - ....... ....... -1,200,000 1.200,000 0% 0 1,200,000 1,200.000 ... 0

" 1.200,000 2,400,(X)(} 5% 60,000 1.200.000 2,400.000 5'" 00.000 3.000,000 5,400.000 '0% 300,000 3.000,000 5,400,000 .... 300,000 3,653,021 24,000,000 '5% 547,953 3,653,021 24.000.000 15'" 547,953

0 48,000,000 20% 0 0 48,000.000 - 0 0 UP """ 0 0 UP 30% 0

9,053,021 907,1153 9,053,021 907,1153 0

TaxCredH Basic: Tax Ctemt 0 300,000 300,000

Tax Llablilly after CredH 907,1153 907,_ __ 000

..

.,

Page 38: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.. ANNEX 3

PERSONAL INCOME TAX PARAMETERS

BASE OPTION ... (Current) (Proposed) Change?

Personal Relief I Allowances Standard Personal Tax Relief 0 0 No Marriage Allowance 300,000 0 Yes Children Allowance

Education Allowance per Child 240,000 0 Yes Maximum Number of Children 3 3 No ...

Old Age Allowance Minimum Age 65 65 No Amount 500,000 0 Yes

Aged Dependent Allowance Alnount per Dependem 200,000 0 Yes Maximum Number of Dependents 2 2 No

Disablement Relief ('Yo of Income) 25% 25% No Social Security Allowance 5% 5% No .. ute Insurance Premium

Max. Allowable Deduction (% of Amount Insured) 10% 10% No Max. Allowable Deduction (% of Income) 10% 10% No

Maximum Allowable Tax Saving 999,999,999 999,999,999 No

Deduction I Exemption Overtime Income No Pension Income No • Dividends No Interest Income No

Tax Rates Standard PAVE Rates Bracket Rates Bracket Rates

First Income Bracket 1,200,000 0% 1,200,000 0% No Second Income Bracket 2,400,000 5% 2,400,000 5% No Third Income Bracket 5,400,000 10% 5,400,000 10% No .. Fourth Income Bracket 24,000,000 15% 24,000,000 15% No Fifth Income Bracket 48,000,000 20% 48,000,000 20% No Sixth Income Bracket UP 30%, UP 30% No

ProfitlPerfonnance Bonus Rate First Income Bracket Second Income Bracket

Credit .. Basic Tax Credit 0 300,000 Yes

Growth Factors

Population Growth (Taxfilers Only) 1.020 Wages & Other Income 1.120 Investment 1.000 .. Consumer Price Index 1.100

..

Page 39: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

"

ANNEX 4 IMPACl' ANALYSIS: Replacing Personal RelieflDeductions with Basic Tax Credit of ¢300.000

37

•.

Page 40: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

GHANA MONTHLY TAX RECEIPT FORECASTING MODELs

I. Introduction

The monthly tax receipt forecasting model is a simple yet functional tool to project short- and medium·term revenues from major taxes. This model requires primarily monthly tax collection data, which is in most cases readily available within the tax agency. This model captures seasonal effect of tax collections, and -- with simple calibration techniques -- it provides relatively accurate results that make the model attractive for developing countries. This model is particularly useful for developing countries where consolidated information on actual collection may be delayed by more than one month. For monitoring purposes, the monthly collections in any specific month, the projected revenues for the remainder of the year, and the estimated surplus or deficit by type of taxes can be extremely valuable information to the Ministry of Finance and the tax agency.

The tax receipt forecasting model uses actual monthly receipts data and projected GDP growth - or other tax base proxies (e.g. private consumption or imports) growth - to forecast collection. This model allows the analyst to account for changes in real GDP, the price level, effective average tax rates, and any behavioral effects that may be associated with changes in tax rates. Applications of this model for Ghana have been developed using Microsoft Excel (i.e. "IRS Monthly Receipts Projections.xls", "CEPS Monthly Receipts Projections.xls", and "VATS Monthly Receipts Projections.xls").

II. Methodology

At any given month, annual tax receipts for the fiscal year can be expressed as the sum of two parts: (1) actual revenues collected up to the month for which receipts data are available; and (2) forecasted receipts for each of the remaining months of the fiscal year. To project the second part of monthly receipts, the model takes into consideration the actual growth of the year-to-date tax collections as compared with that of the same period in the previous fiscal year and the projected growth of tax base proxies (e.g. GDP, private consumption, imports, etc.).

The general form of the monthly tax receipt of fiscal year, y, is as follows:

where:

T, = ~ 1'.., + ;~ Tiy • B,,_l . (1 + 8)· ( ~~':-l) q

Ty : Annual tax receipts for the fiscal year y.

Ta.y: Actual monthly tax receipts in fiscal year y. where tax collection data is available.

T~y.l: Actual monthly tax receipts in fiscal year y-l.

5 Prepared for the Ministry of Finance & the Revenue Agencies Governing Board, Accra, Ghana by a Duke University Center for International Development tearn lead by Profs. G. Glenday and G.P. Shukla with assistance from Rubi Sugana. The work was funded by the USAID Mission to Ghana through Sigma One Corpnration Contract No. 641-C-00-98-00229 over the period October 2002 through September 2003.

38

(1)

'"

..

..

..

..

Page 41: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

.-

..

..

8 .. ,..,: Actual monthly tax base in fiscal year y-l.

m

Ii

li., :

t;..,:

Number of months up 10 which actual tax receipts data in fiscal year y is available .

Growth faclor.

Projecled effective average tax rale in month i of fiscal year y defined as the ratio of tax revenues 10 the tax base.

Actual effective average tax rale in month i of fiscal year y-l defined as the ratio of tax revenues 10 the tax base .

I] : User defined elasticity of the tax base with respecl 10 changes in the tax rale'.

Since last year's tax base is equal to the actual tax collection divided by the effective average tax rate, equation (I) can be rewritten as follows:

T = i: I:, + .I: T, . (I:.,-, J. (1 + b) (~J' , ~ -' T T

1=.tII ... ,.-1 •. ,.-1

(2)

or

( J' ..

T =i:T +.I:T ,·(1+0)' ~ ,. .-=1 fl., """" ... '1- T

/:.., 4.7-1

From equation (3), we find that at any given month, i, in the current fiscal year, y, the general form of the monthly tax receipt forecasting model is as follows:

T '0'

T = '., I:., '(1+0){ ::,r (.j I

If the actual tax receipts data for month i is available, then Ti., is equal to the actual tax receipts. Otherwise, Ti.y is projected using the actual last year's receipts for the same month multiplied by a growth factor and adjusted by a user-defined elasticity, TJ to captu,," behavioral effect if there were policy changes during the fiscal year which modified the effective average tax rate.

The growth factor, Ii, can be measured by a weighted average of two growth factors:

(i). the actual growth of year-ta-date receipts in the current fiscal year over that of the same months in the previous fiscal year; and

(ii).the expected growth rate of tax base proxies (e.g. GDP) in the current fiscal year.

The weight for the first growth factor is the fraction of the number of months in a fiscal year during which the taxes were actually collected. The weight for the second growth factor is

6 The elasticity of the tax base to the tax rate,l}, is related to the price elasticity of demand for the market quantity forming the tax base, /}Qp, as follows: ",,(I:1"/}Qp)~(Itt). Typically, "will have a small ,-.Joe close to zero.

39

Page 42: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

the remaining fraction. If the weight for one factor is (l, then the weight for the other factor is (I-a). Therefore,

where:

8=a'g+(I-a). a' m a_' [

fTa.Y - fTa.Y-,]

g

LTa •y_, Qo::l

Expected growth of a tax base proxy (e.g. GOP, private consumption, imports, etc.) for the current fiscal year7

(5)

The second term in the above equation represents the growth rate of actual monthly receipts this fiscal year over that of the same period in the prev:ious fiscal year.

The weight a can be calculated as (l-mI12), which varies linearly from 0 (when no tax receipts data is available for any months in the current fiscal year) to I (when tax receipts data is available for all of the months in the current fiscal year).

The term (r.,y ITa,y.j )l+ry in equation (4) represents the ratio between projected current year's effective average tax rate over last year actual effective tax rate for the same month, Effective tax rate for a particular month is defined as the actual tax receipts divided by the effective tax base for that month. If there were no tax policy changes, which affected both tax rate and tax base during the fiscal year, this ratio would be equal to I,

If there were some policy changes (e,g, new rate structures or exemptions introduced) take place in the current month, the effective average tax rate ratio should be estimated using other source information through micro simulation or structural adjustment techniques, Alternatively, this ratio can be calibrated by comparing the actual collection against projection results,

Initially, the user-defined elasticity of the tax base with respect to changes in the tax rate, Tj,

is suggested to be set to zero. Under this condition, it is assumed that a one-percent increase, or decrease, in the effective average tax rate, the tax revenue would increase, or decrease, by the same rate. Overtime, as more actual tax receipts data becomes available, this elasticity -as well as the tax base proxy growth rate - should be calibrated by comparing the actual and projected tax receipts figures,

For illustrative purposes, let's assume that the effective average tax rate in July 1999 is lowered by 5% from that of 1998, because of the introduction new tax exemptions. Let's also assume that the elasticity of the tax base with respect to the tax rate, 77, is ",{},05, This implies that the growth factor, (1M), in equation (4) would be: adjusted downward by approximately 5%, since:

(,,;,y / )1+" = (0,95)'-{)oS ~ 0.95 /ra,y_t

7 The GDP component of the growth factor is generally the ratio of nominal GDP in year y to nominal GDP in year y-l, because the tax bases for most of the taxes are likely to change both with changes in the price level and with changes in real GOP. In the case of excise taxes, however, the GDP component of the growth factor is the ratio of realGDP in year y to real GDP in year y-l. Since the excise taxes are imposed on a per unit basis, the excise tax bases are likely to change only with changes in real GDP.

40

..

..

..

..

'.

..

..

Page 43: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

.,

The effective growth factor would become (1+0) times 0.95. This adjusted growth factor multiplied by the actual monthly receipts in the previous fiscal year generates the forecast revenues for the corresponding months of the current fiscal year.

The elasticity of the tax base with respect to the tax rate, .~an also be estimated from the following relationship with the price elasticity of demand and the tax rate:

1]= (1+ 7]ap )*('ti(I+1:».

where 7]ap = price elasticity of demand

1: = average effective tax rate

For example if 7]ap = -1.4 and 1: = 0.15, then 11 = -{J.05 .

Estimating the average effective tax rates and elasticity of the tax base with respect to the tax rate may not be practical when detailed breakdown of tax bases by tax rate is nO! availal>le. An alternative approach is to adopt revenue impacts due to discretionary changes and efficiency gain estimated using other models, e.g. micro simulation model.

III. Application to Gbana

Three monthly receipts forecasting models (one for IRS, one for CEPS and one for VAn have been developed using Microsoft Excel ("IRS Monthly Receipts Projections.xls". ··CEPS Monthly Receipts Projections.xls", and "VAT Monthly Receipts Projections.xls") .

3.1 Actual Tax Receipts Data

The actual tax receipts data is stored in the "Actual" worksheet. Every month. new actual tax collection information should be entered in the appropriate column (month) and row Ita., type) .

Below the actual collection data table in this worksheet (e.g. rows 53-54 in the IRS model) there are flags which are used to identify which months' data to be included in the fonnulas found in other worksheets. These flags are used in the formula to calculate the tOlaI actual collection of previous years up to the same month.

3.2 Macro Parameters

The "Macro" worksheet contains nominal tax base proxy (e.g. GOP) growth projections. 11te actual and projected real growth of tax base proxies and inflation rates are obtain~d from relevant ministries and other government agencies. The nominal growth is calculated lblOg

the following formula:

g=r+7C+r·n (6)

where: g Expected growth of a tax base proxy (i.e. nominal GDP growth).

r Real growth of tax base proxy.

11 Change in price level (inflation).

41

Page 44: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

3.3 Table of Parameters

The "Parameter" worksheet contains the estimated average tax base elasticity with respect to tax rate, and the estimated average effective tax rate ratio between current year and last year (see equation 4). When no data is available to estimate the figure, the elasticity of tax base with respect to changes in tax rate is set to zero. In this case, the tax base is assumed not to response to changes in the effective tax rate.

3.4 Growth Factors

The implementation of equation (5) to estimate the growth factor based on the actual tax collection growth and the projected growth of tax base proxy is shown in the "Growth" worksheet. The first table in this worksheet, titled "Estimation of Growth Factors" contains the growth factors estimated using equation (5). These factors are calculated based on the average actual tax collection growth for the year-to-month for the last 3 years, the projected growth of the tax base proxy, and the growth factor weight based on the number of months where actual collection data is available.

The second table in this worksheet, titled "Actual Collection Year-to-Month," calculates the total year-to-month actual tax collection for the months where actual tax collection data is available for the current year. The third table, titled "Actual Tax Collection Growth Year-to­Month" calculates the percent change of the actual year-to-month tax collection calculated in the second table.

Cell C5 in the "Growth" worksheet limits the actual growth component used in equation (5) to estimate the growth factor. Therefore, if the actual growth is greater than the maximum set in cell C5, i.e. 60%, the model will use the maximum percentage as the actual growth component. This is an attempt to smooth out the effect of wild fluctuations in the month-to­month collection.

The growth factor weight, a, is calculated based on the number of months where actual tax collection data is available (cell J5 in the "Growth" worksheet). For example, if the number of months where the actual collection data is available is equal to 2 (i.e. actual collection data is available for the months of January and February), then a is equal to (1 - 2112), or 0.83.

3.5 Baseline Projections

The "Baseline" worksheet contains the baseline projections following equation (4). The projections for the remaining months where the actual collection data is not available are estimated before discretionary changes, assuming that all figures for average tax rate elasticity in the table of parameters are set to zero, and the average effective tax rate ratios are set to 1.

3.6 Discretionary Changes

The "Measures" worksheet calculates the adjustment factors used to amend the baseline projections due to estimated impacts of discretionary changes and efficiency gains. Both revenue impacts and efficiency gains inputted into this worksheet are treated as exogenous variables; they are estimated outside this model. This is an alternative approach to estimating the average tax rate elasticity and average effective tax rate ratio. The estimation of revenue impacts due to discretionary changes may be done using tax micro simulation or other calculator models.

42

-

..

..

..

..

..

..

Page 45: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

The area shaded with light yellow color in the "Measures" worksheet is used to store the absolute values of the estimated revenue impacts due to discretionary changes for each type of taxes and every tax year. The area shaded with light green color to the right of the estimated revenue impacts table is used to capture the estimated efficiency gain in percentage terms. The total impact of these two measures on the annual baseline projections is calculated in the table titled "% Revenue Impacts on Baseline Revenue Projections" in this worksheet.

The table that is used to calculate the annual baseline projections is below the table used for calculating the percent revenue impact. The formula used to calculate the total revenue impacts in percentage terms is as follows:

%Revlmpact = x 100% + %EfficrencyGam [ EstimatedAnnuaJlmpact ] ..

AnnualBaselineProjection (7)

3.7 Projections with Discretionary Changes and Efficiency Gains

The percent of revenue impacts due to discretionary changes and efficiency gain calculated in the "Measures" worksheet is used in the "Projection" worksheet to project the monthly tax revenue after taking into account the estimated discretionary changes and efficiency gains. The projected monthly revenue is calculated using the following formula:

ProjectedRevenue = BaselineProjection x (I + %Revenuelmpact) 18)

3.8 Variance

The "Variance" worksheet calculates the difference, in percentage terms, between the total year-to-month projected and actual tax collections using the following formula:

where:

['iTP.J -'iTa.r]

Variance = a=' 'i T:~' (9)

0=1

Tp.y: Projected monthly tax receipts in fiscal year y, where tax collection data is available.

T~y: Actual monthly tax receipts in fiscal year y, where tax collection data is available.

3.9 Adjusted Projections

Using the Variance calculated in the "Variance" worksheet, the adjusted projections for the remaining months where no actual collection data is available are calculated in the "AdjProjection" worksheet using the following formula:

AdjustedProjection = ProjectedRevenuex(1 + Variance) (10)

IV. Steps in Using Monthly Receipts Model

I. Update the actual monthly collection data in the "Actuaf' worksheet. 2. Update the nominal growth projections of the tax base proxies in the "Macro" worksheet. 3. Estimate impacts on the tax revenues due tax policy changes and efficiency

improvements, and update the "Measures" worksheet with the estimates. 4. Review the projections.

43

Page 46: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

-

..

..

..

..

44

Page 47: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

"

.,

.,

..

.,

VAT analysis and revenue estimation based on micro-simulation dataS

Introduction

This report covers the analysis and revenue forecasting procedures developed for the VAT

based on the detailed micro-simulation data available from the V A TS and CEI'S databases.

The focus is put on two important areas

I. The relationship between V AT collected on imports by CEPS and VAT collected from domestic traders by VATS. Example of issues: When there are fluctuations or structural shifts in imports causing changes V AT collected on imports, what is the net change in total VAT revenues given that a significant share of V AT becomes deducted as input V A T by domestic businesses?

11. The relationship between the VAT accruing from new economic activity and the actual revenue collections given the complex relationship caused by V A T arrears, credit carry forwards and outstanding refunds. £mmple of issues: When the economy grows expanding the V A T base will the revenue receipts from VAT increase at the same rate?

VAT data sources

1. Import VAT: Detailed VAT liabilities and collections are available for 2000-

2002 based on import databases derived from ASYCUDA and GCMS

systems maintained by CEPS. These data sources are described in detail

under the CEPS tax models.

2. Domestic VAT: Information on domestic VAT operations is maintained on

the VIPS system by VATS. This allowed the following data bases to be

constructed:

a. Annual return information: Based on the VAT monthly VAT returns of registered VAT payers annual data based on the January-December returns were created for 1999-2001. This annual data gives data by taxpayer identifying the industrial sector (ISIC) and local of the VAT office plus

• Prepared for the Ministry of Finance & the Revenue Agencies Governing Board. Accra. Ghana by a Duke University Center for International Development team lead by Profs. Graham Glenday and G.P. Shukla with the assistance of Rubi Sugana, and Carlos Della Torres. The work was funded by the US AID Mission to Ghana through Sigma One Corporation Contract No. 6-I1-C-OO-98-00229 over the period October 2002 through September 2003

PREVIOUS PAGE BLANK 45

Page 48: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

i. Output supplies by taxable, exporVzero rated, exempt and remitted

ii. Output tax iii. Domestic input values and input VAT iv. Import input values and input VAT v. Deductible input VAT (total input VAT adjusted for exempt

outputs) vi. Net VAT accrued in year (positive amount gives VAT payable or

negative amount gives amount due to taxpayer as refund or credit)

This data gives a summary of accrued VAT from new economic activity in a year. It also gives a useful breakdown of the input VAT deductions arising from imports and domestic supplies.

b. Annual accrued tax and credits: Annual data based on monthly returns from December of prior year to November of current year for 1999-2001 which gives aggregate of months where taxpayer has positive VAT liability (box 13 on VAT return) and months where taxpayer has negative VAT liability or VAT credit (box 14 on VAT return). This data gives gross accrued positive and negative VAT amounts by a taxpayer in a year that can be compared with the VAT collections, refunds and credits absorbed in a year.

c. Annual reassessments, adjustments, penalties and interest: Annual data for 1999-2001 that give the additional VAT liabilities of a taxpayer over and above that declared in VAT returns. VAT 63 gives the reassessment and VAT 33 gives the adjustments to VAT payable.

d. Annual tax refund claims and payments: Annual data for 1999-2001 that give the amounts of refunds claimed by taxpayers eligible to claim settlement of VAT credits as refunds (V AT 37) and the refund payments made in a year.

Databases still under development by VATS: Work is ongoing to complete the above databases for 2002. In addition, databases are being constructed that give the VAT arrears and outstanding credit position of each VAT payer at the end of each year.

Estimation of VAT revenues based on import and domestic VAT returns and collection data

46

-

..

...

..

..

..

..

Page 49: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

..

The VAT is ideally structured as a tax on consumption. This means that the V AT revenue estimates depend primarily on estimating the size and composition of the consumption base relative to the VAT rate structure (including zero-rated and exempt items) and ideally it should be independent of whether the consumption goods are imported or domestically produced. In practice, the analysis and estimation of the VAT that separates out the impoll and domestic V AT collections is impoIlant for a number of reasons:

• Separate revenue targets are required for the monitoring of collections of impoll and domestic VAT .

• In the shOll-run, changes in exchange rates, world prices and other impoll conditions can cause changes in the mix of V AT derived from imporu versus domestic suppliers.

• Different production and marketing channels can affect whether VAT can effectively be collected. For example, assuming impoll compliance is reasonably high, it is easier to collect V AT at the border on many consumer items than if they are produced and traded by small local businesses. In the case of imporu there is no minimum size transaction that is tax exempt, whereas small businesses may be outside of the V AT net because of the high costs of collecting tax effectively from them.

• The detailed information available for impolls helps distinguish between the sources of non-compliance or inefficiency in the V AT system that may arise from the customs system separately from those in the domestic V AT system. Aside from customs smuggling, under valuation and other fraud, customs V AT collections on imporu can be delayed by imporu entering bonded warehouses before they are brought into home use and made subject to duty and V AT.

• Impoll information typically gives details of various impoll exemptions and also goods entering V AT free into expoll processing zones. Exempt and zero-rated imporu entering final consumption and exempt business inputs form paIl of the adjustments to GDP or consumer demand that need to be made in estimating the effective V AT base. At a minimum, use of the actual impoll and domestic \" A T information can strengthen the use of the National Accounts or Input-Output approaches.

The detailed information available on imporu, and the duty and V A T charged on these imporu provides a strong stalling point for estimating VAT. In countries with relatively high impoll shares, 25% of GDP and higher, impoll VAT collections typically form Sin or more of the total V AT collections. While it has to be recognized that a significant share (>f impoll VAT becomes deductible against domestic output V AT, still import V AT is likely 10

contain about 25% or more of the final VAT accruals. In the case of Ghana, for example. in 1999 through 2001, impoll VAT has formed 70% of lotal VAT. Hence. a significant share. probably higher than 25% of the final V AT base is likely contained in the impoll V AT.

VAT revenues effectively corne from two bases: final consumption by the private and government sectors, and purchases of inputs by V AT exempt businesses. The non­deductibility of exempt business inputs effectively increases the VAT base beyond thaI afforded final consumption where these exempt businesses, such as exempt financial institutions or farmers, produce intermediate inputs into other businesses producing final

47

Page 50: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

VATable goods and services. Essentially, the VAT on the inputs to these exempt businesses results in double taxation or tax cascading within the VAT system.

The issue in analyzing import and domestic V AT is the same: basically there is a need to identify the shares of import VAT and domestic output VAT that are (i) VAT on final consumption, (ii) VAT on exempt business inputs, and (iii) input V AT that will be deducted or refunded by a domestic business. Assume that a share yof the import VAT (Rimp) is

deductible as input VAT and a share 8 of output VAT on domestic supplies (R:;""::"') is

deductible as input VAT, then the normal total accrued! VAT (R") from business activity in a year, which is composed of domestic VAT (Rdom) and import VAT (Rimp), can be restated as follows:

R" = Rdom + Rimp (1)

= [ R output _ R illput _ R !"put ] dom darn unp + Rimp (2)

= [ Routput - is Routput - yRimp 1 + Rimp (3) dam dam

= R;;;:::;"'(1 - 8) + Rimp ( 1- y) (4)

Now, R:;""::"' (1 - 8) is the portion of output VAT that is the VAT on final consumption and

on exempt business inputs. These are final, non-deductible and non-refundable VAT collections. Similarly, Rimp( 1- y) is the portion of import VAT that is composed of VAT on final consumption and exempt business inputs. Therefore,

R"= Rfinal + R ebi + Rlinal + R.ebi dom dom Imp Imp (5)

= Rfinal + R ebi (6)

The issue now is how to estimate the relationships above for the accrued VAT revenues. First, assuming all VAT returns and declarations are accurate, Rdom can be found from the domestic V AT returns as accumulation of all newly declared output V AT payable minus all input VAT deductible (R = RO"Ip", _ Riop", _ Riop",)

dom dom dom unp·

Here the full amount of all input deductions is recognized whether or not it is actually absorbed as a deduction or refunded immediately in order to measure the full accrued input V AT deductions, and hence, the true value of final consumption and exempt business inputs subject to V AT in the year.

In (2), the share of output VAT that becomes deductible is found from the V AT returns as

8 = R;;;;;' I R;;;:::;"' , and the share of import VAT that becomes deductible is found from the

import VAT and domestic VAT returns as y = Ri'::;"' I Rimp . Relationship (4) is merely a

reorganization of (3) to group the domestic output VAT and import VAT amounts that will be final taxes.

48

-

-

...

..

...

...

.,

...

...

..

-

Page 51: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

,.

'"

..

..

..

Relationship (6) is the same as can be found from national accounting dala and other sources. The V AT base is fonned by taxable final consumption and exempt business inputs. Whether or not the V AT revenues from final consumption and exempt business inputs can be decomposed into the amounts coming from import VAT and domestic output VAT as in (5). or alternatively whether the amounts of final import VAT and domestic output VAT in (4) can be decomposed into final consumption and exempt business inputs. poses a challenge. The decomposition of import V AT should be feasible within limits depending upon the information available imports. Domestic V AT returns typically provide limited information about the nature and origin of inputs and the destination of outputs. but some information would be available about the nature of the business and its output.

If import dala includes information about the type of goods and their final use classifications. and the type of importer. then more of less strong inferences can be drawn about whether the import VAT will be deductible or not, and where it is not deductible. whether it is going into final consumption or an exempt business. The type of goods can be broadly classified as capilal and transport equipment, industrial inputs. primary goods for industry or consumptions. fuels. consumption goods. etc. These can be used to assign probabilities of their destinations and deductibility. For example. most industrial machinery and capital equipment, and raw and intermediate inputs would have probabilities near 100% of being deductible. If farmers are V AT exempt, then agricultural inputs would be largely non­deductible. In addition. information about

the type of importer (V AT registered or not and sector) can be used to identify certain exempt business. For example. imports of financial institutions would be exempt business inputs. Where exempt or zero-rated imports are goods entering final consumption such as pharmaceuticals. then the import data provides information to assist with the adjustments to the GDP-based VAT base estimates. Once probabilities (or shares) of imports entering various VAT categories have been assigned. these probabilities could then be adjusted at the micro data level such that the aggregate results are consistent with (I) through (6). For example. the shares of all import V A T assigned to being deductible from domestic output should add up to )'Rimp.

Further consistency checks can be applied between the different data sources. For example.

if the adjustments to national accounting data have yielded an estimate of R'" for the V A T

system as a whole. and a separate estimate of R:!;, has been made from the import data. then

R<bi can be found as (R,b.. R'b.) <10m ""I' •

In summary. the information from the actual V AT collections on imports and domestic supply can significantly enrich the estimation of the V AT base by allowing estimates of the accrued V AT be made directly and providing detail on the composition of imports and domestic supplies that will generate input V AT deductions. or enter final consumption or be purchased by exempt businesses.

49

Page 52: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Estimates of VAT Revenues based on Import and Domestic VAT Data, 1999-2001

Based on domestic VAT returns for 1999 through 2001 and import VAT collections, the accrued VAT can be analyzed in terms of relationships (1) through (4) above.

FIrst, the VAT returns give the basic structure of the accrued domestic VAT derived from the output V AT minus the domestic and import V A T claimed as input V AT, where

Rdom = (7)

Table 1: Accrued domestic VAT: Output and input VAT

Domestic VAT Share of Import VAT Share of Accrued Gross output VAT from claimed as input output claimed as input output domestic VAT

domestic supply VAT in year VAT VAT in year VAT in year

A B 0 C CIA D = A-B-C

Cedi millions

1999 965,068 422,491 43.8% 285,809 29.6% 256,769 2000 1,784,061 696,407 39.0% 733,541 41.1% 354,113 2001 2,290,548 945,072 41.3% 806,637 35.2% 538,840

Source: VAT returns

The total accrued V AT in a year comes from domestic and import VAT, as follows:

= (8)

Table 2: Accrued Total VAT: Domestic and import VAT

1999 2000 2001

Share of Accrued accrued

domestic VAT total VAT Cedi millions

256,769 354,113 538,840

33.1% 28.4% 28.5%

Import VAT

518,687 892,371

1,352,986

Share of accrued

total VAT Accrued total VAT

66.9% 71.6% 71.5%

775,456 1,246,484 1,891,825

Note that the accrued domestic V AT forms about 30% of accrued total VAT. Another way of stating this is that domestic value added contributes to about 30% of the total value of final consumption with the balance coming from imports.

Alternatively, from (7) and (8) accrued total VAT can be expressed in terms of the VAT collected on final domestic supplies plus imports that are final as follows:

na __ [ Routp"' " Routp"' D 1 R 1\ dom - U dom - r J[\.imp + imp

50

-

..

-

..

..

..

..

-

Page 53: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

..

..

"

=

=

R;:;:-(1- 15) + Rimp( 1- r)

[R.:::::'+ R:!l + [Rt:;' + R;!;l

(9)

(10)

Table 3: Total accrued VAT: Domestic and import VAT

Share of final VATabie Share of final Share of

supplies from VATabie import VAT Final accrued domestic Final import supplies from thaI is final Total accrued domestic VAT suppliers VAT imports (l-y) VAT

Cedi millions

1999 529,286 70.3% 223,886 29.7% 43.2% 753,172 2000 1,085,964 85.5% 183,650 14.5% 20.0% 1,269,614 2001 1,342,068 70.0% 574,176 30.0% 41.7% 1,916,244

From Table 3, the share of import VAT thaI becomes a VAT input deduction, y, is nearly 60% of the import VAT. This means that only about 40% of the import VAT is final VAT. Given that import VAT collections are about 70% of VAT collections, import V AT then fonns about 30% of final V AT collections (about 40% of 70%), whereas domestic V A T collections make up the remaining 70% of final VAT collections .

In 2001, import VAT, which was deductible, fotmed 58.3% of import V AT or Cd 806.637 million (colunm C in Table I). This amount can be reconciled approximately with the pattern of import V AT across different Broad Economic Code (BEC) categories of imports. Broad Economic Codes assign goods in HS Code categories to different classes of capital. intetmediate and final consumption goods. Based on these BEC categories, a probability range of the shares of import VAT that becomes deductible can be assigned and the resultant amount of deductible import VAT compared with the known total amount of import VAT deducted and collected in 2001. These are illustrated in Table 4. For example, import V AT collected on intermediate processed industrial supplies (BEC of 220) is assigned a probability in the 85% to 95% of being deducted, whereas non-durable consumer goods (BEC of 630) are assigned 0 to 5% probability of deduction. The suggested probability ranges give a 58% probability of deduction, the same as the known aggregate deduction of import V A T compared to reported import VAT collected. As discussed above. more detailed analysis of the types of importers to determine their V A T registration status could be used to refine these estimates of the probabilities of VAT deductibility.

51

Page 54: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Table 4. Estimates of Input VAT deductions of Import VAT collected in 2001

Probability of input Expected value of BEC Description Classification VAT deduction input VAT deduction

Low High Low High estimate estimate estimate estimate

Cd millions Cd millions 111 Primary food and beverages,

mainly for industry Intermediate 80% 90% 37,966 42,712 112 Primary food and beverages,

mainly for household consumption Consumption 20% 30% 1,610 2,415 121 Processed food and beverages,

mainly for industry Intermediate 85% 95% 36,005 40,241 122 Processed food and beverages,

mainly for household consumption Consumption 20% 30% 35,891 53,837 210 Primary industrial supplies n.e.s. Intermediate 85% 95% 36,307 40,578 220 Processed industrial supplies n.e.s. Intermediate 85% 95% 442,475 494,530 310 Fuels and lubricants, primary Intermediate 50% 60% 22,408 26,889 321 Fuels and lubricants, motor spirit 10% 20% - -322 Fuels and lubricants, other Intermediate 50% 60% 8,061 9,674 410 Capital goods (except transport) Capital 60% 70% 56,898 66,381 420 Capital goods - parts and

accessories Intermediate 60% 70% 30,391 35,456 510 Passenger cars Passenger Cars 5% 10% 9,600 19,200 521 Industrial transport Capital 80% 90% 66,472 74,781 522 Non-industrial transport Consumption 10% 20% 1,319 2,639 530 Transport equipment - parts and

accessories Intermediate 50% 60% 37,717 45,261 610 Durable consumer goods Consumption 15% 25% 7,914 13,189 620 Semi-durable consumer goods Consumption 5% 10% 2,428 4,856 630 Non-durable consumer goods Consumption 0% 5% - 2,599 700 Goods n.e.s. Other 10% 20% 198 396

Total of import categories in ASYCUDA database 53% 62% 833,660 975,635

Import VAT collections in 2001

Input deductions of import VAT by registered axpayers in 2001 58% i 806,637

Given the detailed infonnation available on import VAT and on import input VAT deductions, this approach can be used to estimate the impact on net VAT revenues of fluctuations in import VAT collections either in aggregate (in 2001, 58% of import VAT would be forgone as input VAT deductions) or based on shifts in the composition of imports which would change the share

of import VAT that is deductible. Given the importance of imports in the Ghanaian economy (about 50% of GDP) and the sensitivity of imports to changes in economic conditions, tracking and predicting the total value and composition of import V AT is important to forecasting the net accrued VAT in a year.

52

Import VAT in 2001

Cd millions

47,458

8,051

42,359

179,457 42,714

520,558 44,815

-16,123

94,830

50,651 192,000 83,090 13,193

75,435 52,757 48,558 52,206

1,979

1,566,235

1,383,722

..

-

..

-

..

..

..

..

Page 55: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

Forecasting VAT Revenue Collections from Accrued V AT Revenues

In any month a domestic VAT payer either accrues an additional VAT liability (output V AT exceeds input VAT deductions) or a V AT credit (input VAT deductions exceed output V An. In the case of a positive V AT payable either the tax is paid on time or tax arrears arise that need to be settled at some future date. In the case of a V AT credit. two outcomes are possible: either the taxpayer claims a refund where eligible (essentially only regular exporters)9 or has to carry forward the credits to be offset against future VAT liabilities. It is evident, therefore, that the revenues from any accrued V AT liability and credits depends upon the whether or not taxes are collected, credits absorbed and refunds paid over the remainder of the year (or alternatively stated, whether or not arrears, outstanding credits and outstanding refunds build up over the year.) Hence, the forecasting of VAT receipts depends not only on the forecasting of VATabie activity, but also on the changes in arrears, credit carry forwards and outstanding refunds that. in Part. depend upon administrative policy and performance.

VAT Arrears

For most domestic taxes based on assessments of tax (income tax, V A T, etc). there is the persistent problem of the collection of arrears. Changing degrees of success in collection tax arrears (or debt collection) can have significant influence on revenue performance in a particular year. When forecasting revenues, this phenomenon of arrears collections mises additional problems over and above forecasting future tax accruals. The behavior of future revenue accruals depends on different factors from the behavior of arrears collections. Assuming a constant tax structure, future revenue accruals depend upon the growth in the ta.ll base (which depends upon the nature of economic growth), the compliance of the taxpayers. and administrative efficiency of the revenue agency. Collection of outstanding arrears. however, depends primarily on the administrative efficiency. The state of the economy can also influence the willingness and capability of taxpayers to pay arrears when they are ,u hjecl to enforcement actions. This means that ideally to estimate future collections. tall aCdU31 s and arrears collections should be estimated separately.

Excluding the impact of V AT credits and refunds, the relationship between revenue collections and accrued VAT in a year is given by:

=

where Re,

R',

M,

R',- M (11)

=

= =

Revenue collections in year t Accrued revenues in year t (based on returns for December of year (t-I) through November of year I) Change in slock of tax arrears over year I Accrued revenues in year I

+ Additional reassessments. adjustments. penalties and interest in year I (VAT 67 and 33)

'Refunds are payable (i) wbere exports exceed 70% of lotal supplies in !be accounling period. or (ii) "IK-r< a V AT credit has been outstanding for 3 months or more. Refunds are payable within 30 days or applicatll.Jn Form V AT 35 is used to claim a refund. Refunds are charged against the expenditure accounts of the Government. In practice, only exports apply are granted refunds.

53

Page 56: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Collection of accrued revenues in year t Collection of am,ars outstanding at beginning of year t

Another perspective on (11) is to recognize that typically information on tax revenue structure and performance derived from tax returns reflects the accrued taxes (Rat) relating to the period in question and the prevailing economic conditions. The translation of these tax liabilities into tax collections depends upon the timing and enforcement of collections that follow on the filing of a tax return. The use of tax returns is often critical to get the details of the taxpayers and tax bases necessary for understanding the future behavior of the tax performance. It is therefore important to be able to translate tax return information into tax collection performance.

VAT credits and refunds

The forecasting of VAT collections is further complicated by the generation of V AT credits when input VAT exceeds output VAT. As noted above, VAT credits will only reduce VAT collections as fast as they are refunded or carried forwarded and absorbed by future VAT

Table 5, Gross V AT payable and V AT credits, net accrued VAT and

1999 2000 2001

net VAT collections, 1999-2001

Gross VAT payable

Box 13

Gross VAT credits over Gross VAT

Gross VAT credit payable

Box 14 Cedi millions

401,104 (145,010) 36% 564,904 (241,166) 43% 802,925 (275,361) 34%

accruals.

Net accrued V AT

Box 13 + Box 14

256,094 323,737 527,564

Net VAT collections

Annual collection net of refund

payments Cedi millions

244,819 349,682 536,070

Analysis of the V AT returns show that the frequency and size of V AT credits make the proper treatment of V AT credits a significant problem. Table 5 shows that for domestic accrued V A T, the total amount of V AT credits generated in a year are generally over one­third of the gross VAT payable in a year.

Furthermore, Table 6 shows that most firms experience some months that have V AT credits. In 2001, only 0.6% of the firms with VAT returns had positive accrued VAT payable in every month of the year. Most firms (88.5%) one negative or accrued VAT credit month in 2001 even though overall they generated positive had at least accrued VAT.

54

-

-

..

..

..

..

Page 57: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

'"

.,

The relationship between VAT collections and accruals in (11) is now extended to include the changes in the stock of VAT credit carry forwards and changes in outstanding refunds as follows:

Re• = Ra• _ M + ~c. + ~ef. (12)

where M = Change in stock of tax arrears over year t

= Accrued positive VAT in year t (V AT return, Box 13)

+ Additional reassessments, adjustments, penalties and interest in year t (V AT 33 and 63) V AT credits carried forward that are absotbed in year t Collection of accrued revenues in year t Collection of arrears outstanding at beginning of year t

~C. = Change in the stock of outstanding tax credits over year t

= Accrued gross VAT credi ts in year t (V AT return. Box 14) V AT refunds claimed in year t (V AT 35) V AT credits carried forward that are absorbed in year t

~ef. = Change in the stock of outstanding tax refunds over year t

= V AT refunds claimed in year t (V AT 35) V AT refunds paid in year t

From (12), it is evident that revenue collections will increase relative to accrued revenue in a year (i) if collections of current accruals and arrears are increased reducing the stock of arrears, (ii) VAT credits build up over the year, and (iii) if refund claims remain unpaid over the year. Clearly, however, future tax collections will be reduced as the credit carry forwards and outstanding refunds will have to be paid off in the future - the government is building up a contingent liability.

The stock and flow relationships in (12) can be studied for 1999.2000 and 2001 for the domestic V AT accruals and collections. Constructing the stock and flow accounts for the

Table 6. Distribution of firms with positive and negative accrued VAT In 2001

Number of firms Distribution Firms with positive accrued VAT in every month 112 0.6% Rrms with net positive accrued VAT (cred~s in some months) 15.371 88.5% Rrms w~h net negative accrued VAT (positive VAT in some months) 1,883 10.8% Firms w~h negative accrued VAT in every month 12 0.1% Total 17,378 100.0% V AT are simplified given the V AT was a new tax in 1999 making all opening balances zero.

Table 7 gives the stock and flow accounts for the domestic V AT for 1999-200 1. First. the accrued VAT and credits are found from the VAT returns (Box 13 and Box 14. respectively) for each year. The difference between these figures gives the new accrued VAT from the

55

Page 58: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

economic activity in the year. To this amount are added the reassessments, adjustments, penalties and interest to get R\ Next, this amount is compared with the actual domestic V AT collections in the year reduced by the amount of refunds paid to get Ret. The net collections are then compared with the accrued VAT in the year, or (Ret - Rat). In each of the three years, the collections fall short of the accruals, bUit in each year the short fall decreases in both absolute and relative terms (from about 25% in 1999 to less than 5% of the accrued VAT in 2001.) From (12), the collection short fall is filled by the increase arrears and/or the decrease in the stock of credits and refunds. In Table 7, the build up of the stocks of VAT arrears, outstanding VAT refunds, and VAT credit cany forwards over 1999 through 2001 are calculated based on the data on the flows of positive and negative V AT accruals, VAT collections, and refund claims and payments. No data has as yet been obtained on the absorption rate of credit carry forwards over the years, but given the accounting identity in (12), changes in assumptions about the absorption rate of credit carry forwards are always consistent with the collection short fall. An increase in the rate of credit absorption causes the stock of credits to decrease, but also causes the stock of arrears to crease, leaving the net arrears unaltered. In table 7, it was assumed that 90% of any opening balance of credits would be absorbed over the next year and remaining the balance would be fully absorbed in the following year. In any case, the stocks of net arrears, net of the stocks of credit carry forwards, is invariant to the assumption of the credit absorption pattern. These net arrears are presented in table 7: they rise from Cd 94 billion in 1999 to Cd 190 billion in 2001, but as share of the gross accrued VAT, the net arrears rise only moderately from 29% to 34%.

Table 7 also gives the build up of outstanding refund payments over the years. These rise from Cd 8 billion in 1999 to Cd 20 billion in 2001. When these stocks of refunds are also netted out of the arrears along with the credits, the net arrears rise from 26% to 30% of the gross accrued V AT. This implies that by the end of 2001, the net arrears were equivalent to about 3 to 4 months of VAT. When it is recognized that much of the arrears are incurred by different taxpayers from those with credit carry forwards and/or

56

-

-

..

..

...

...

..

..

..

Page 59: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

.,

Table 7. Domestic VAT stock and flow accounts for 1999-2001 1999 2000 2001

Flows Stock Flows Stock Flows Stock Cedi millions Cedi millions Cedi millions

New net output V AT or gross positive accrued VAT (Box 13) 401,104 564,904 802,925 New credits or gross accrued VAT credits (Box 14) (145,010) (241,166) (275,361)

Accrued VAT from activity In year 256,094 323,737 527,564 Reassessments, adjustments, pena/1ies and interest 75,314 84,672 33,659

Gross accrued VAT 331,408 408,409 561.223

Domestic VAT collections 274,528 385,158 574,042 Refunds paid (29,709) (35,476) (37,972)

Net VAT revenues (cash flow) 244,819 349,682 536,070 Net VA T revenues - Gross accrued I VAT (86,589) (58,727) (25,152)

Refunds outstanding 12.9931 Refund opening balance 0 7,926

Refund claims Refund payments Refund closing balance

Credit carry forwards V AT credit carry forward opening balance New Credits Credits absorbed and refunds claimed VAT credit carry forward closing balance Arrears VAT arrears opening balance New taxes and reassessments, etc V AT collections and credits absorbed

VAT arrears closing balance

VAT arrears net of cred" carry forward VAT arrears net of cred" carry forwards and refunds

Change in net arrears

Growth in net VAT Growth in gross accrued VAT Growth in change in net arrears

I 37,635

I (29,709) I

7,926

145,010 (92.635)

52,375

476,418 (329,528)

146,890

94,515

86,589

86,589

40,543 44.931 (35,476) (37,972)

12,993

0 52.375 241,166 275.361

(205,543) (264,931)

87,999

0 146,890 649,576 836.583

(550.158) (794,042)

246307

158,309

145,316

58,727 25,152

42.8% 53.3"k 23.2% 37.4%

-32.2% -57.2"/0

outstanding refunds, then the arrears situation looks much more unattractive. By the end of 200 1, the gross arrears were about Cd 289 billion or 52% of gross accrued V AT in the year or about 6 months VAT. This has important implications for V AT debt collection administration. Further analysis should be undertaken of the composition of the stock of

57

I I !

19,952

87,999

98,428

246.307

288,849

190,421

170,469

Page 60: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

arrears to determine the size and feasibility of the arrears collection. It is also clear that the sizes of the stocks of arrears, credits and refunds are large enough relative to the annual flows that the changes in these stocks should be explicitly dealt with in forecasting domestic VAT revenues.

From the initial data available on the V AT arrears and credit balance of individual VAT payers, it is evident that many taxpayers have both VAT arrears and VAT credit carry forwards. While a firm can absorb credit carry forwards by future positive V AT accruals, the opposite is not the case. If a firm builds up arrears and then generates V AT credits, these subsequent credits cannot be offset against the arrears. This situation needs further analysis and possible changes in V A T policy to allow reconciliation of VAT credits against VAT arrears. This will remove the need to take debt collection actions against such VAT payers.

. Forecasting VAT revenues

A common short-cut in revenue forecasting is to base the forecasts of revenue collections in the next period on the current level of revenue collections by applying a growth rate that reflects the growth in the tax base (g) based on expected growth in the economy, or

(13)

If (12) is substituted into (13), and the credits and refund balances are netted out of the

arrears, then

Rat *(1 +g) - ~NAt *(1 +g) (14)

where ~NAt = ~At - ~Ct - meft = change in net arrears over year t

This implies that arrears collections are expected to grow at the same rate as the underlying tax base over the next year. In many situations, this is a reasonable assumption, if compliance and enforcement behavior stays the same, and if the arrears are not large relative to the current taxes. This approach is not reasonable where arrears are relatively large and have been either increasing or decreasing rapidly in recent years. For example, where arrears have built up to unacceptably high levels in recent years, then changes in tax administration hold the potential for high revenues collection from debt collection efforts. It is then important to estimate the future revenues from reduced arrears separately from the taxes arising from new economic activity that generates the revenue accruals in the future. Alternatively, if more effective debt collection administration is introduced, the collection of arrears may be accelerated for a number of years before it settles at some lower stock value.

Table 7 compares the rates of growth in net collections with the rates of growth in accrued VAT and changes in net arrears in 2000 and in 200 I. It is evident that the net collections grow considerably faster than the accrued V AT because the growth rate in net arrears declined or decelerated over 2000 and 2001 after a rapid build up in 1999. For example, in

58

...

-

..

..

..

..

..

..

..

Page 61: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

2001 net collections increased by 42.S% whereas accrued VAT grew by only 23.2% as the change in net arrears declined by 32.2%. 10 The implications are that forecasting the V A T revenues from one year to the next requires more infonnation than the growth in the V AT base. It also requires projections of the changes in the V AT arrears. VAT credit carry forwards and outstanding refunds. A systematic approach can be taken to this type of projection. This presented next.

An alternative expression for RCt in (12) is to recognize that the revenue collections in a year

come from the share of new accrued VAT in the year that is actually collected (aR't) and the share of net arrears at the beginning of the year that are collected over the year (,8NA..\) or

I Ret = aRat + fJNA..\ (15)

By contrast, the stock of net arrears increases over year 110 the extent thai the accrued revenues are not collected [(I-a)Ratl and decreases 10 the extenl thai the outstanding nel arrears al the beginning of the period (,8NAt.\) are collected. Hence. the change in the nel arrears over year I can be expressed as:

ANA. = NA. - NAt.\ = (I-a)Rat - fJNA t.\ (16)

Now. to check how fast the change in net arrears is growing from one year 10 the nexl. thai is whether it is growing faster or slower than lhe accrued V A T. the difference between the change in nel arrears over year 1+ I can be compared 10 thai over year I:

ANA.+\ - ANA. = (I-a)(Rat+\ - R't) - fJ (NAt - NAt.tl (17)

Now ifR't is growing al rale. g. and substiluling (16) into (17) gives

ANA.+\ = (I-a)gR't - (l-fJ) ANA. (18)

From (13) and (14). for RCt also 10 grow al rale of g. then

ANAt+\ = ANAt*(I+g) (19)

Substituting (19) inlo (IS). the steady-state conditions for the revenue coUections to grow

at the same rate, g, as the accrued VAT are found 10 be:

10 The growth rate in revenue collections can be found from the growth rates in accrued V A T and change in net

arrears by weighting the growth in accrued V AT by the ratio of accrued arrears to V A T revenue collections to

2000 and weighting the growth in the change in arrears by the ratio of the change in net arrears 10 \' AT R"\cnue collections in 2000.

59

Page 62: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

!\NA, = (l-a)gI(g+jJ) Ra, = (l-a)fJgI(ag+/1J Re

,

R\ = (ag+/1J/(g+jJ) Ra,

(20)

(21)

Note that Re, + !\NA, = Ra, as expected from (12). Typically, for a< I, Ret < Rat and !\NA,>O. This means that for revenue collections to glOW steadily at the same rate as the new VATabie economic activity, the net arrears will continue to grow. The increase in net arrears will be smaller, the closer a and fJ are to one, and the slower the nominal growth in the VAT base, g.

In addition, it is of interest to check some extreme conditions. If there are no stocks of arrears, then a= 1, and !\NA, = 0 and Re

, = R\ such that the VAT collections will always grow as fast as the accrued VAT. Similarly, if g=O, then !\NA, = 0 and Re

, = Ra,. If Ra

, is growing at rate, g, and no arrears are collected or fJ= 01, then Re, = oRa, = (l+g)R\, and !\NA, = (1-a)Ra, = (1 +g)t.NA,.,

To estimate (15), (20) or (21), estimates of fJare required. The net arrears are reduced over a year, t, by the collection of outstanding arrears, the absorption of the credits being carried forward, and the payment of refunds outstanding at the beginning of the year. This can be expressed as

or

where:

fJNA,_, = fJAA,_, - fJc C,_, - '&crRef,_,

fJ = (fJAAt-, - fJc C,_, - '&crRef,_,)/NA,_, (22)

fJA = share of arrears, A,_" collected over year t

f3c = share of credits carried forward, C,_" absorbed over year t

'&ef = share of outstanding refunds" Ref,_" paid over year t

From the above, it is likely to be more the exception than the rule that, if the accrued VAT revenues are expected to grow at rate, g, then the revenues collected will also grow at the same rate. Note from Table 7, for example, that over 1999-2001, the domestic VAT revenues grew significantly differently from the accrued VAT revenues. In general, if R". is expected to grow at rate, g, then the revenues can be forecast from (15) and (22)

(23)

Here, Ra, and NAt should be known from current VAT accounts so that the issue is

estimating a and fJ. These parameters need to be estimated from past administrative experience as well as planned collection and refund actions for the coming year. For purposes of illustration, a hypothetical set of values consistent with forecasting 200 1 net VAT collection from 2000 V AT information in Table 7 could be

60

..

..

..

..

..

...

Page 63: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

Growth rate of nominal accrued VAT, g = Collection rate of accrued V AT, a = Share of net arrears, NAt.], collected over year t, fl =

37.4%

80%

60%

Share of arrears, At.I, collected over year t, flA = 69%

Share of credits carried forward, C..I, absorbed over year t, f1c = 80%

Share of outstanding refunds, Ref,.], paid over year t, /kd = 95%

An alternative approach is to go back to (12) and (14) to express future collections in terms of

current collections, which was the original question posed:

or

RCt+1 _ Re, = (R"+I - R',) - (~At+1 -~At)

R'\+I = Re, + gR', - gANA~A,

RCt+ i = Re

, (I + g*w, - gmA*Wb)

where: w, = R',! Re,

and from (17) gmA = (I-a)g R',I ~At - fl

At steady state growth conditions:

w, = (g+/1)/(ag+/1) (>1)

(24)

Wb = (I-a)gI(ag+/1) (generally, but not necessarily <I)

gmA = g

w, - Wb = I

Inspection of expression (24) shows that for a given expected growth rate in R', of g that all right-hand side terms are know in terms of current year accounts except for gmA. which depends upon a and fl. While at steady state growth conditions, growth in change in net arrears, gmA = g, in general gmA can be positive or negative. For example. Table 7 shows that in Ghana the growth rate of the change in net arrears was large negative both from 1999 to 2000 and from 2000 to 2001. This implies large differences between the growth in collections and accrued VAT arise. Hence. forecasting VAT collections from current year collections using (24) requires the same information content as (23). namely the share of new accrued V AT that is expected 10 be collected (a) and the share of outstanding nel arrears that are expected to be collected (/1). These clearly depend primarily upon the capacity and

61

Page 64: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

planned behavior of the V AT agency. To some extent they will also depend on the economic growth conditions. For example, if the economy is growing rapidly, the outstanding credits will be absorbed more rapidly and taxpayers can be expected to be more willing to pay both current VAT and arrears. The opposite can be expected to be the case in a slow growth economy.

Tax credits creation and absorption can also be cyclically sensitive. For the VAT, at the start of an economic growth phase, large investment in equipment and inventories can generate high credit flows and a build up socks of V AT credits, which are then absorbed over the following growth period. This causes a slower growth in V AT revenues. In an economic slow down, investment in equipment and inventories declines such that credit flows decline but also causing VAT revenues to be sustained as goods are sold out of inventories without offsetting deductions. This causes a counter cyclical behavior in V AT collections.

Next steps

To continue the work to build on the V AT analysis and forecasting procedures developed here based on micro-simulation data would include:

1. The updating of the VAT data files through 2002, and the extension of the analysis through 2002.

2. The direct estimation of a and Rab f3 and NA" f3A and At, f3c and Cb and /kef and

Ref,

3.

4.

5.

6.

for t = 1999 to 2002 from the VAT accounts of domestic VAT payers. Net arrears should be reported monthly in tenns of its three components: gross arrears, credit carry forwards, and outstanding refunds. Develop policies to allow for offset of credit carry forwards against VAT liabilities and possibly other tax liabilities. Review the impact of the disincentives to investment caused by prevalence of credit carry forwards rather than refunding Ihese taxes more rapidly. Refine estimates of deductibility of import V AT in tenns of VAT registration status and industrial sector of importers once the Taxpayer Identification Number (TIN) system is fully operational and TINs are accurately recorded on all customs transactions.

62

-

-

-

..

..

..

..

..

Page 65: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

,.

..

,.

..

.,

GHANA TRADE TAX CALCULATOR MODEL 11

I. Introduction

In the past, capturing and processing a large amount of data for policy analysis required a large mainframe computer and a complex organization to support it that only rich countries could afford to have. 1berefore, policy decisions in developing countries were often made without a thorough data analysis. This situation has now changed as more affordable and powerful yet easy to use personal computers are more accessible to almost any government agencies of any size. The completion of computerization of customs administration in Ghana has proved that technology is becoming more available to public services in developing countries.

The Trade Tax Calculator Model is a micro simulation model that uses large amount of actual raw trade data collected at the entry ports through the Asycuda systems or Ghana Customs Management Systems (GCMS). This model is used to analyze the impacts of proposed policy changes on projected revenues from trade taxes.

To simulate the calculation of trade taxes, the Trade Tax Calculator Model requires detailed import values by harmonized system (HS) code, country of origin. and customs procedure code (CPC). The CPC is used to identify exempt imports. imports with preferential tariff rates. and imports which subject to certain type of taxes only.

The model has been developed for Ghana using a combination Microsoft Access database (i.e. "Ghana Trade Calc· Database.mdb") and Excel spreadsheets (i.e. "Ghana Trade Calc· Policy.xls" and "Ghana Trade Calc· Output.xls").

II. Methodology

The value of year y imports of commodity i in domestic currency can be expressed as follows:

where: Mi.y: Value of imports of commodity i in fiscal year y.

Q<y: Quantity of imports of commodity i in fiscal year y .

p..; : World price of commodity i in fiscal year y in foreign currency (F$).

It' : Nominal exchange rate (D$IF$).

The nominal exchange rate:

E" = E' x (I + gpd Y(I + gpf )

where: It : Real exchange rate - relative to domestic and foreign inflation (D$IF$).

II Prepared for the Ministry of Finance & the Revenue Agencies Governing Board. Accra. Ghana by a Duke University Center for International Development team lead by Profs. Graham Glenday and G.P. Shukla ,.-ilh!he assislance of Rubi Sugana. The work was funded by !he USAID Mission to Ghana Ihrough Sigma One Corporation Contract No. 641·C-OO-98-00229 over lhe period Ocloher 2002 wough Septemher 2003

63

Page 66: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

gpd: Domestic currency inflation rate, or percent change in import price index.

gP: Foreign currency inflation rate.

Projected value of year y+ I import of commodity i:

M,.y+! = M", x (I + gM,")

where: gM,·: Growth in nominal import value commodity i = gM; X(I+ gpd)

gM;: Growth in real import value of commodity i.

Tax revenue in year y+ I from commodity i:

where:

R"y+! = M i,,+! xTj xx,(j)

Tj : Tariff, or tax rate, for tax type j

Xi(j) : Exemption rule for tax j apply to import commodity i. The value of Xi(j) is between 0-1. If the import is fully exempt from a certain tax, x'{j) is equal to O. If the import is fully taxable, Xi(j) is equal to I.

Without tariff or other tax rates change:

gM; = (7J,':aDP X g ~DP )+ [(I + 7J,':. )x (gP,'" + gE' )j

7J,':aDP Import quantity elasticity of commodity i with respect to real GDP

growth,

7J,':. : Import quantity elasticity of commodity i with respect to real import

pnces.

g ~DP: Real GDP growth rate,

gP;W" Growth in real world prices (F$) of commodity i.

gE' Growth in real exchange rate.

With tariff or other tax rates change:

gM; = (7J,':aDP x g~DP)+ [(I + 7J,~p)x (gp;W'" + gE' )j+ (q,~P xgP;T)

where: gp;T Growth in price because of change in tax rates (total for all tax types).

tlT, /(1 + T,)

The growth in real exchange rate can be estimated using the following fonnula:

where:

gE' = [(I + gE·)X I + gp~ ]-1 l+gP

gE" Growth in nominal exchange rate.

64

...

(3)

(4)

(5) ..

(6)

..

..

(7) ..

.. (8)

..

Page 67: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

III. Application for Ghana

TIle prototype of the Trade Tax Calculator Model that has been developed for Ghana has three main components:

- The Trade Database (i.e. "Ghana Trade Calc - Database.mdb").

- TIle Tax Policy Table (i.e. "Ghana Trade Calc - Policy.x/s"), which contains detailed information on tariff rates, exemption rules, commodity price and income elasticities. and projected macroeconomic measures (real GDP and real exchange rate growth).

- TIle Model Output (i.e. "Ghana Trade Calc - Output.xls"), which presents the summary outputs of a policy simulation based on the current and proposed tax structures.

Fig. 1. Trade Tax Calculator Model Components

3.1 The Trade Database

TRADE DATABASE

The Trade Database is the core of the Trade Tax Calculator Model. The database is a Microsoft Access file ("Ghana Trade Calc - Database.mdb"), which consists of 6 tables:

- BEC - This table is a lookup table. which contains codes of the broad economic categories and their descriptions.

- Country - This table is a lookup table. which contains country codes and their names.

- Currency - This Currency table is a lookup table. which contains currency codes and their descriptions.

- Exempt - This table contains customs procedure codes (CPC). CPC descriptions. and their corresponding exemption rules. This table is a mirror of the "£tempt" worksheet in the Excel file of the Tax Policy Table ("Ghana Trade Calc -Policy.xls").

- Tariff-This table contains HS Codes and their corresponding tariff rates and growth factors. New HS-Code in this table must be updated regularly. This table is a mirror of the "Tariff" worksheet in the Excel file of the Tax Policy Table ("Ghana Trade Calc - Policy.xls").

65

Page 68: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Trade - The Trade table contains commodity trade transaction data extracted from both the Asycuda systems or GCMS. This table must be updated with valid and complete import transactions from either of these two systems annually. Descriptions about the data required for this table is presented in Appendix I.

In addition to these tables, the Trade Database also contains database queries, which are used to calculate the projected import values and trade taxes, and generate the model outputs. The structure of these database queries follows mathematical formulas described in the methodology section. The database schema and detailed data dictionary of these tables are presented in Appendix I.

3.2 The Tax Policy Table

The Tax Policy Table contains the current and new, proposed tariff rates for each commodity, exemption rules, as well as the price and income elasticities for each commodity. At present, the spreadsheet is structured to capture rates for import duty, import VAT, special tax, import excise, processing fee, ECOWAS levy, export development levy, and vehicle examination fee. There are approximately 6,000 commodities sorted by its harmonized code listed in this worksheet. This worksheet also contains projected percentage change in gross domestic product (GDP), exchange rate, and import price index, as well as the world price of each commodity.

The actual tariff rates, growth factors, and exemption rules used in the model are stored in the "Tariff' and "Exempt' tables in the Access database. However, in order to enhance the user interface and facilitate the updating procedures of these tables the Excel worksheets are created. The "Tariff' and "Exempt" worksheets in the Excel file "Ghana Trade Calc -Policy.xls" are the mirror of their corresponding tables in the Access database.

Before using the Tax Policy Table, it is required to set the correct Access database name and the complete folder name where the database resides in the "Setting" worksheet of the "Ghana Trade Calc - Policy.xls". Otherwise, the program will display an error message when the "Update" button is pressed. The data year must also be entered using a 4-digit format (e.g. "2002"). This is to tell the program which data year to be retrieved if there are more than one year data available in the database.

- Tariff Worksheet

To simplify the updating of tariff rates for 6,000 or so commodities, the "Tariff' worksheet has been set up with the auto filter turned on. Since the tariff rate for commodities at the 6-to-8 digit level harmonized code is most probably the same, the auto filter is very handy in making the updating. To update the tariff rates for dairy products (HS code starts with "0402"), the filter for column "HS-CODE12" (column "B") can be set to "04" by clicking the down-arrow sign and select "04." To further refine the selection, the filter for column "HS­CODE34" (column "C") can be set to "02."

Modifications to the "Tariff' worksheet will not take effect until the "Tariff' table in the Access database is updated. To update the "Tariff' table, the "Update Tariff' button must be pressed every time an update is made.

- Exempt Worksheet

The "Exempt" worksheet contains adjustment factors that reduce the amount of taxes. The adjustments are based on the CPC of imports. The adjustment factors rage 0 to 1 for each type of taxes are provided for both current and proposed tax structures. If the factor is set to

66

..

..

ow

Page 69: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

0, the imports are fully exempt and no tax will be collected. If the factor is set to 0.5, only 50% of the statutory tax rate will be charged on the imports. These adjustment factors are the implementation of equation (5) as described in the Method%gy section.

Similar to that of the 'Tariff' worksheet, modifications to the "Exempt" worksheet will DOl

take effect until the .. Update Exempt" button must is pressed.

67

Page 70: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

3.3 The Model Output

The Model Output ("Ghana Trade Calc - Output.xls") retrieves data from the Trade Database by executing the queries stored in the Access file and summarizes the outputs for each tax type by different categories, both under the current and proposed tax structures. The breakdown categories are tax rates, broad economic categories (BEC), harmonized section, customs procedure codes (CPC), and country of origin. The output for each category is stored in different worksheet, namely "Summary," "BEC," "HS Section," "CPC," and "Country."

To generate or update the output after modifying the Tax Policy Table, the spreadsheets must be recalculated, either individually each time or all together in one run by pressing the appropriate buttons provided in every worksheet. Please note that the Trade Database can be very large that makes the processing time for each individual worksheet quite long, between 3-10 minutes depending on the computer speed, available physical memory, and hard disk space. Similar to that for the Tax Policy Table, the database name and the complete folder name where the database resides must be correctly set in the "Setting" worksheet prior to pressing the buttons.

68

..

..

..

..

..

..

..

Page 71: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

TRADE DATABASE DATA DICTIONARY

TARIFF Table • Data Max Field

No. Field Name Type Size

Descriptiom

HS-CODE Text 10 Harmonized Code 2 HS-DESCR Text 255 Harmonized CodeJCommodity Descriptions 3 HS-SECT Text 2 Harmonized Section 4 BEC Text 3 BECCode 5 CHRGDP Number 5.2 Projected Percent Change in Real GDP 6 CHWRLDP Number 5.2 Proj. % Change in Real World Commodity Price

7 CHXCGRT Number 5.2 Projected Percent Change in Real Exchange Rat< 8 INC-ELAST Number 5.2 Imp. Qty Elasticity wi Respect to Real GDP Growth

9 PRC-ELAST Number 5.2 Imp. Qty Elasticity wi Respect to Real Impon Prices .. 10 CHlMPIDX Number 5.2 Projected Percent Change in Import Price Index II DUTY-C Number 5.2 Import Duty Rate - Current 12 DUTY-P Number 5.2 Import Duty Rate - Proposed 13 VAT-C Number 5.2 VAT Rate - Current 14 VAT-P Number 5.2 V A T Rate - Proposed 15 SPCTAX-C Number 5.2 Special Tax Rate - Current

• 16 SPCTAX-P Number 5.2 Special Tax Rate - Proposed 17 EXCISE-C Number 5.2 Import Excise Tax Rate - Current 18 EXCISE-P Number 5.2 Import Excise Tax Rate - Proposed 19 PROCFEE-C Number 5.2 Processing Fee - CwreOl

20 PROCFEE-P Number 5.2 Processing Fee - Proposed 21 ECOLVY-C Number 5.2 ECOW AS Levy - Current

• 22 ECOLVY-P Number 5.2 ECOW AS Levy - Proposed 23 EXPDEV-C Number 5.2 Export Development Levy - Current 24 EXPDEV-P Number 5.2 Export Development Levy - Proposed 25 VEHEXM-C Number 5.2 Vehicle Examination Fee . Current 26 VEHEXM-P Number 5.2 Vehicle Examination Fee - Proposed 27 OTH-C Number 5.2 Other Tax Rate - Current 28 OTH-P Number 5.2 Other Tax Rate - Proposed 29 GROWTH-C Number 5.2 Estimated Gro"1h Rate wlo Rate Changes 30 GROWTH-P Number 5.2 Estimated Growth Rate wi Rate Changes

..

69

Page 72: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-TRADE DATABASE DATA DICTIONARY

EXEMPT Table

No. Field Name Data Max Field

Descriptions Type Size

I CPC-CODE Text \0 CPC 2 CPC-DESCR Text 255 CPC Descriptions 3 DTYXMT-C Number 5.2 Adj. Factor for Import Duty Rate - Current 4 DTYXMT-P Number 5.2 Adj. Factor for Import Duty Rate - Proposed 5 VATXMT-C Number 5.2 Adj. Factor for V AT Rate - Current 6 VATXMT-P Number 5.2 Adj. Factor for V A T Rate - Proposed 7 SPCXMT-C Number 5.2 Adj. Factor for Special Tax Rate - Current 8 SPCXMT-P Number 5.2 Adj. Factor for Special Tax Rate - Proposed 9 EXCXMT-C Number 5.2 Adj. Factor for Import Excise Tax Rate - Current \0 EXCXMT-P Number 5.2 Adj. Factor for Import Excise Tax Rate - Proposed .. 11 PROCXMT-C Number 5.2 Adj. Factor for Processing Fee - Current 12 PROCXMT-P Number 5.2 Adj. Factor for Processing Fee - Proposed \3 ECOXMT-C Number 5.2 Adj. Factor for ECOW AS Levy - Current 14 ECOXMT-P Number 5.2 Adj. Factor for ECOWAS Levy - Proposed 15 EXPXMT-C Number 5.2 Adj. Factor for Export Development Levy - Current 16 EXPXMT-P Number 5.2 Adj. Factor for Export Development Levy - Proposed • 17 VEHXMT-C Number 5.2 Adj. Factor for Vehicle Examination Fee - Current 18 VEHXMT-P Number 5.2 Adj. Factor for Vehicle Examination Fee - Proposed

19 OTHXMT-C Number 5.2 Adj. Factor for Other Tax Rate - Current 20 OTHXMT-P Number 5.2 Adj. Factor for Other Tax Rate - Proposed

..

..

..

..

..

70

Page 73: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

TRADE DATABASE DATA DICflONARY

TRADE Table

No. FJeld Name Data Max FJeld

Descriptions Type Size

1 RECORDID Auto Record ID 2 OfFICE Text 4 Customs PortlDistrict Office Code 3 YEAR Number 4 Fiscal Year (format YYYY, e.g. '2003') .. 4 MONTH Number 2 Month of Transactions 5 CPC Text 5 CPCCode 6 HS-CODE Text \0 HSCode 7 COUNTRY Text 3 Country of OriginlDeslination Code 8 CURRENCY Text 3 Currency Code 9 NETTWT Number 15.2 Net Weight 10 FOB_FX Number 15.2 FOB Value in Foreign Currency 11 FOB_DOM Number 15.2 FOB Value in Cedis 12 INSURANCE Number 15.2 Insurance in Cedis 13 fREIGHT Number 15.2 Freight in Cedis 14 OTHERS Number 15.2 Other Costs in Cedis 15 IMPDUTY Number 15.2 Import Duty Amount (Cedis)

WI 16 IMPVAT Number 15.2 Import VAT Amount (Cedis) 17 SPECTAX Number 15.2 Special Tax Amount (Cedis) 18 IMPEXCISE Number 15.2 Import Excise Duty Amount (Cedis)

19 PROCFEE Number 15.2 Processing Fee (Cedis) 20 ECOLVY Number 15.2 ECOW AS Levy Amount (Cedis) 21 EXPLVY Number 15.2 Export Dev. Levy Amount (Cedis) .. 22 EXAMFEE Number 15.2 Examination Fee (Cedis) 23 FLAG Text Aag to Exclude Data in the Analysis

..

"

..

71

Page 74: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-TRADE DATABASE DATA DICTIONARY

BECTable Data MaxField -No. Field Name Type Size

Descriptions

I BEC Text 3 BECCode 2 BEC·DESCR Text 255 BEC Descliptions 3 BEC-CLASS Text 30 BEC Classification -

COUNTRY Table

No. Field Name Data MaxField

Descriptions Type Size ...

CTRYCODE Text 3 Country Code 2 CTRYNAME Text 30 Country Name

CPCTabie ..

No. Field Name Data Maxfield

Descriptions Type Size

CPC Text 3 Customs Procedure Code 2 CPC-DESCR Text 255 CPC DeSCIiptions •

CURRENCY Table

No. Field Name Data Maxfield

Descriptions Type Size •

I CURCODE Text 3 Currency Code 2 CUR-DESCR Text 255 Currency Descriptions

...

72 -

Page 75: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

'"

WI

Tax Policy Analysis in Ghana: A Plan for Capacity BuildingJ2

1. Introduction

This study examines the existing capacity for analyzing tax policy issues and estimating tax revenues in the Ministry of Finance, Revenue Agencies Governing Board (RAG B) and the various revenue agencies of the government of Ghana and proposes a plan for improvement in the current situation. The report may be divided into three parts.

The first part is introductory and presents the general background (section 2), and the scope of the study (section 3). The second part examines the present structure of the Ministry of Finance (section 4), and goes on to explore the budget process and the existing status of and the capacity for fiscal policy analysiS in the Ministry of Finance, RAGB and the various revenue agencies (section 5) and finally proposes steps for filling the institutional gaps (section 6). The third part deals with the functions of the Tax Policy Unit and the subunits within RAGB and revenue agencies (section 7, 8 and 9), examines issues of coordination and organization (section 10) and finally describes the support services needed for smooth and effective functioning of the whole system (section 11).

A. Background and Scope of the Study

2. Background

2.1 Trends in Tax Revenues

Government revenues in Ghana as a percentage of GOP have been gradually going up during the last decade. However, total expenditures as a percentage of GOP have also increased leaving a deficit in all the years in the range of 6.5% to 10% of GOP (Table 1)13.

12 Prepared for the Ministry of Finance & the Revenue Agencies Governing Board. Accra. Ghana by a Duke University Center for International Development team lead by Profs. Graham Glenday and G.P. ShuIJa. The work was funded by the US AID Mission to Ghana through Sigma One Corporation Contract No. 641-C-OO-98-00229 over the period Octoher 2002 through Septemher 2003 13 Source: Statistical Appendix. IMF Country Repon No. 031134. May 2003; "Ghana: Selected Issues"

73

Page 76: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

-

-

...

..

..

..

74

Page 77: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

..

.,

..

III

..

III

Table I: Revenues and Expenditures as Percentage of GOP (1993-2002)

199 199 199 199 1997 199 199 200 200 200 3 4 5 6 8 9 0 1 2

Total 19.2 22.3 24.1 20.2 18.7 20.5 18.0 19.8 25.0 20.0 Revenue and Grants

I ,

Of which: Tax 14.9 16.2 14.7 15.1 14.7 15.8 14.8 ., 16.3 18.1 17.4

1

Revenues .

I

T01al 29.0 31.2 30.4 29.7 29.0 28.6 26.2 27.7 32.7 26.1 1 Expenditures i, I

Breakup of , , ,

Expenditures: i

1 I , -Recurrent 61.8 57.3 53.8 55.2 57.2 60.4 61.3 66.9 60.9 76.61

!

I I ,

- Capital 38.2 42.7 46.2 44.8 42.8 39.6 38.7 33.1 39.1 : 23.4 , i I

SurpluslOefici -9.9 -8.9 -6.4 -9.5 -10.3 -8.1 : -8.2 -7.9 -7.7 -6.1 ,

t \

, ,

Tax revenues as a share of GOP show an unsteady trend, sometimes rising and sometimes falling. For the last two years the revenues have been at 18%.

Total govemment expenditures have also been fluctuating between 26% and 32% of GOP leaving a deficit every year. It may also be noted that, capital expenditures as a share of total expenditures have generally exhibited a declining trend since 1995 while recurrent expenditures have been increasing. This expenditure pattern is clearly not desirable for the long-term investment and economic growth .

PREVIOUS PAGE BLANK 75

Page 78: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

An analysis of the various categories of tax revenues and non-tax revenues as a percentage of GOP indicates that tax revenues in all categories have shown an upward trend while non-tax revenues have been generally declining (Table II).

Table II: Tax Revenues by Tax Type (Percentage of GOP)

1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Income Taxes 2.8 3.3 3.5 3.8 4.3 4.4 4.5 5.2 5.6 5.7

Excise/SalesN AT 7.2 7.6 6.5 6.5 5.9 6.4 6.6 7.4 7.5 7.7

Trade Taxes 3.1 5.3 4.6 4.8 4.5 5.0 3.8 3.6 4.1 4.1

Non-Tax 1.8 2.6 5.8 2.5 2.7 2.6 1.5 1.5 0.9 0.5

The income taxes or direct taxes have been increasing over the past decade, mainly from a growth in PAVE taxes on employment income, some of this due to bracket creep ad some due to increases in real income. Sales taxes were fluctuating till 1998 while revenues from the Value Added Tax introduced in 1999 have been gradually going up. Trade taxes have exhibited an unsteady trend and stand at 4% of GOP.

2. 2 Revenue Potential by Tax Type

An examination of the breakdown of tax revenues as percentage of GOP by tax type for the year 2002 gives a sense of the revenue potential for each type of tax (table III). The tax bases (column C) are estimated from the national accounting data.

76

-

-

-

...

...

..

..

...

-

Page 79: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

Table III: Tax Revenues as a Share of GDP, Year 2002

Actual revenues, potential tax bases, potential revenues, effective tax base and !laps In tax bases

Actual _tax Tax raIe -- EIIocII"" I Shont 01 _b __

R.....,ua/ _GOP (b) ............, tax .~_ _:1nI8moI GOP (a) GOP -OPi 001_ ot_1

I

A B C D_B'C E=AlC 1·00 T .. _

17.4% 1. Income & 6% 92% !

PAVE lo9"k 20"10 13% 2.6% 14.5"10 I 27.3"10 --I enIo4 C8111E!f II;

, exempliollS

SeIf·Employed 0.3% 27.0% 10.0"10 2.7% 3.2% 88.3"10 Ad! nil llisbalively n __

types 01 lax more appropriate to normaiseclOr

i -Company and 3.5% 45"10 32.5% 14.6% 10.8% i 76.0"10 --other income ! ef Iofcemef It;

, exemptions and

I

inc::enti\'eS

2. Domestic 7.6% ! ! Goods & Senrice

E_.Duty 0.7% I

i VAT Iotal (net 01 4.7% 90"10 12.5% 11%

I 37%, i 59% --refunds) enIo<cemefO.

exemption and I feasibily proIlIems ,

I ,

I

Petroleum Tax 2.2% ! 3.1'-""" 4.1% ! T'-

Import duly 2.70/0 47% 10"/. 4.7% 27.3% 41.~o Administrative ~

~ exernpbons(24'\.): duly _ (18'\.)

. ~=-fees 0.5%

Export . Cocoa 0.7% I I <a) I. Base for IIlCOI11eIax IS net naliooaJ onoome. For PAVE potential base assumes abouIl.3...., ...... ng Cd 8 miion. and for sal em ed, it assumes about 4 miftion earri about Cd 3.3 million '

2. Base for VAT is private plus government consumption ~ public sector wage bill

The last but one column gives a sense of the potential base that remains untaxed for each type of tax. It is clear that there is scope of substantial improvement for increasing revenue collections from almost all the taxes.

77

!

i I : , ,

i

Page 80: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

2.3 Tax Policy Issues on the Agenda

In addition to the problem of enhancing tax revenues, there are a series of issues for the government to consider on a priority basis in order to rationalize the tax system.

1. Examining the VAT exemptions: The VAT law in its present form exempts a wide range of goods from productive sectors including agricultural inputs (fertilizers, etc), printed matter, pharmaceuticals, and machinery. These exemptions cause the V AT system to provide negative protection for any domestic business producing these exempt goods. There is a need to study the revenue and efficiency effects of these exemptions. Some of these may need to be eliminated and others replaced by zero rating.

2. Distributional Impact of V AT: While increases in the V AT rate represent an efficient and effective means of raising tax revenues, such increases raise concern about the distributional impacts on the low-income group households and about decline in compliance that offsets some of the gains of a rate increase.

3. Examination of the structure of Personal Income Tax: A preliminary investigation of the income tax reveals that significant potential gains are possible through changes in the PIT structure that could benefit low-income employees while raising additional revenues. For instance,

a. Converting current personal allowances from deductions from taxable income to credits from tax payable to improve tax equity and increase revenues.

b. Taxing employee benefits (such as housing, furnishings, house staff, cars, fuel, club membership, low interest rate loans, etc). These benefits are lightly taxed and are also inequitable as they tend to concentrate among the high-income employees.

4. Rationalizing Corporate Income Tax Structure: The standard company tax rate is 32.5%, but there is a wide range of rate reductions for different companies or types of income, e.g. for cocoa, new businesses in fanning or rural banking, new export processing zone companies, new companies for construction of residential or commercial buildings, income from non-traditional exports, manufacturing income outside Accra and regional capitals, interest from loans to farmers or to leasing companies, manufacturing income outside Accra but in the regional capitals, hotels, companies listed on stock market etc. This variety of rates lowers the effective tax rates considerably.

5. Scrutiny of the tax structure for pensions savin!~s: There is a need to review the structure of deduction limits for private pension contributions and taxation of pension payments in order to integrate them in an equitable and efficient way with SSNIT contributions.

6. Review taxation of all investment instruments: Taxation of interest income of individuals, interest income on government securities, and investment income earned through mutual funds, and efficacy of tax breaks to induce higher private savings in pension funds needs to be closely examined.

78

..

-

..

..

..

Page 81: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

7. Other sources of revenues: Taxation of petroleum and excises on a variety of items should be looked into to assess their revenue potential.

These various issues need to be explored both from the perspective of revenue potential. economic efficiency impacts, and tax incidence on different income and consumer groups. In order to have a stable tax system, it is desirable that tax revenues keep pace with GOP and are adequate to meet the expenditure needs of the government.

3. Scope of the Study

This study evaluates the existing capacity for policy analysis and revenue estimation at the government level and in the various agencies of the government to see whether they are equipped to perform the tasks outlined above. The issue of capacity building in these areas in the Ministry of Finance through the establishment of a Tax Policy Unit (TPU) and strengthening of the existing capacity within the RAGS and the various tax agencies is also addressed. Specifically, the study aims at:

(i) Identifying key points and personnel involved in policy analysis, forecasting and decision making in the area of tax policy;

(ii) Determining the nature and extent of gaps in information, knowledge and skill base among the key functionaries;

(iii) Suggesting programs for capacity building for different agencies and management levels in terms of staffing, training, types of analysis and monitOring for decision support.

The study involved both a field visit to Ghana and a comparative study of experiences from similar economies in developing countries. In the field, a series of interviews were held with the officials of the Ministry of Finance.. Sank of Ghana, RAGS and various tax agencies. A list of people interviewed is at Annex I. This was supplemented by a four-week workshop in which middle and senior level officials of the Finance Ministry and the revenue agencies concemed partiCipated .

B. Assessing the Capacity Levels and Meeting the Institutional Gaps

4. The Structure of the Ministry of Finance:

The Minister of Finance is assisted by two deputy ministers. The bureaucratic setup is headed by the Chief Director, a position provided under the Civil Service law. The

79

Page 82: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

chief director is the overall administrative head of all the divisions, except the revenue agencies, and functions as the principal advisor to the minister.

The ministry is divided into the following departments that report to the Chief Director. Each department is' headed by a director. An organizational chart of the ministry is at Annex II.

(a) Budget Division: Responsible for estimation of the government's financial resources, public expenditure control and preparation of the annual budget.

(b) Policy Analysis Division (PAD): Meant to be a high-level 'think-tank' for research and analysis of financial, economic, social and other development strategies and priorities; expected to monitor and evaluate on-going strategies and policies in the light of financial, social and economic trends and propose changes to improve the management of the economy; and provide the macroeconomic framework for the budgetary exercise.

(c) Private Sector and Financial Institutions Division: Deals with private sector and privatization related issues, public-private enterprise management, review of financial proposals and agreements, and formulation and monitoring of financial sector pOlicies.

(d) General Administration Division: Responsible for personnel management, financial control within the ministry, institutes and departments under the ministry, commissions and committees of enquiry.

(e) International Economic Relations Department Consists of aid and debt management unit, World Bank desk, bilateral and multilateral units and is generally responsible for coordination and oversight of loans and grants.

(f) Project Implementation and Monitoring Unit: Monitors project implementation, poverty reduction expenditures, and reviews project funding documentation and contract documentation.

As would be clear from the organizational chart, the revenue agencies under RAGB have a special status. They are not answerable to the Chief Director and report directly to the Finance Minister. Aside from the newly created RAGB, which largely has an oversight and co-ordination role over tax administration, there is not an effective unit to analyze and develop tax policies in a comprehensive fashion that integrates them into the overall economic and budget policy of the Government.

5. The Existing Level of Policy Analysis within the Ministry and Other Agencies

The individual interviews with senior officers in the Government, the assessment gained from the four-week TARFseminar, and the official documents from the Finance Ministry and the various agencies reveal that not much is presently done at the level of the Ministry of Finance either in terms of policy analysis or revenue estimation. This is primarily because of a lack of strong database, deficiency in analytical capacity and the tradition of taking decision based more on individual judgments and past practices than on a systematic analysis within a well­developed budget preparation procedure at the institutional level.

80

-

-

...

...

...

...

...

...

-

Page 83: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

1be RAGB is in the process of putting the necessary personnel in place and training them in requisite skills for effective monitoring and revenue estimation. The three tax agencies on the other hand have made concerted efforts to improve their tax databases and enhance their capacity for revenue forecasting. This has predictably led to improvements, although the results vary across the three agencies.

The current levels of data collection, policy analysis and revenue estimation undertaken by the various agencies are summarized below.

5.1 Ministry of Fintmce

1be Policy Analysis Division in the Ministry of Finance, with a strength of six persons in position and headed by a director, is responsible for the tasks of policy analysis and revenue forecasting. Following are its main activities:

(a) Policy Analysis: The unit is responsible for initiating new policies and revenue enhancing measures. New proposals/demands could come form revenue agencies. ministers! people's representatives, and other stakeholders. In practice. however. nO policy analysis is being done in the division itself because of lack of capacity. All the proposals are forwarded to the concerned revenue agencies for analysis both on policy and revenue implications. The main role of the unit is coordination among the revenue agencies and RAGB. This is usually done through meetings in December/January every year just before finalizing the budget proposals.

(b) Forecasting and Monitoring Revenues and Expenditures: The monitoring of revenue collection takes precedence over revenue forecasting because presently the realization of revenue targets is the primary concern of the government. 1be Policy Analysis Division receives monthly collection figures from the agencies and then compares the figures with the Bank of Ghana to see how much money was actually paid into the government accounts. This exercise generally reveals huge variances that are then reconciled. The division also looks at procedural bottlenecks that might be responsible for discrepancy between collected and deposited figures. However, no follow up at sector levels is done to see whether the revenue yield from a particular sector of the economy is consistent with its growth and the prevailing economic situation and in case of a shortfall what could be the underlying reasons for that. This is supposed to be done by the revenue agencies but currently they are also not performing this task in a systematic manner. Revenue forecast is done by revenue agencies but their assumptions and results are scrutinized by PAD. While there is some capacity within the division for evaluating forecasts made by the revenue agencies, there is a complete lack of capacity for independent forecasting and authenticating the forecasts presented to the division. In the past, the analysis division used to do some preliminary revenue forecasts and issue targets to revenue agencies, but now this task is performed by RAGB. The analysis division gives its guidelines and targets to RAGB and they in tum assign targets to the revenue agencies. The RAGB targets to the revenue agencies are usually quite high compared to the government budget figures and this is mainly due to over optimistic assumptions about improvement in efficiency. 1be revenue agencies are aware of both the targets. The RAGB figures serve as the benchmark for calculating the incentives (bonuses or retention) to the agencies that are linked to performance. Part of these retentions can also go to increase the staff salaries. The

81

Page 84: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

unit has a limited role in monitoring and analysis of expenditures. The expenditure figures from the budget department are cross-checked with Bank of Ghana figures to reconcile budgetary sanctions and releases and finally a comparison made with the records of the Controller and Auditor General to see what expenditures were actually incurred against the money drawn. Often huge differences are detected.

(c) Aid and Debt Management, and Interactions with Donors: This division has also been managing the flow of foreign funds and acting as the interface with IMF and other donors. The database is maintained by another unit in the ministry but the analysis division prepares the reports and interacts with the external agencies. There is no difference between figures used for domestic purposes and those presented to donor agencies. This task used to be performed initially by the International Economic Relations Department, but over time the PAD has assumed this responsibility.

(d) Coordinate with Bank of Ghana on Public Borrowing: The division has a limited role in deciding the level of borrowings, and it is also represented during treasury bill auctions, open market operations and on the Cash Management Committee. The division advises Finance Minister on fiscal implications of monetary policies and also provides fiscal input for formulating monetary policy.

5.1.1 Manpower Requirements and Training Needs: It is clear that currently the Policy Analysis Division is not handling tax policy issue and this task is being left to the revenue agencies. The focus of the revenue agencies is naturally on revenue enhancing measures and therefore virtually no one is in charge of tax policy analysis work. This often results in adoption of policy measures without adequate examination and then those measures have to be rescinded when problems begin to surface. The division lacks both the manpower and the skills to do what is expected of it.

5.1.2 Maintenance of Database: The aggregate revenue data by tax type is currently compiled in the division using the figures forwarded by the revenue agencies and RAGB. There is a need for a full-fledged data center along with a data manager. In addition, more detailed tax databases are required that would allow tax analysis and revenue forecasting to be conducted.

5.2 TheRAGB

The main focus of the board is currently on monitoring of revenue collections. The board is also in the process of enhancing its capacity for forecasting tax revenues. Policy issues raised at the government level are mostly passed on to the board which tries to do some analysis in conjunction with the revenue agencies but neither the board nor the revenue agencies are really equipped for that kind of exercise.

5.2.1 Monitoring of Revenues (a) A common reporting format is being used for the last three years as suggested by the

IMF for preparing flash reports. The RAGB receives its data from the three tax commissioners and then compiles the information in a common format. However, the methods of data collection and the levels of monitoring the performance of field staff vary substantially across the three revenue agencies.

82

...

...

...

...

..

..

..

..

..

-

Page 85: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

III

..

..

..

III

"

"

"

(b) CEPS have the ASYCUDA system but the system has not been found to be user friendly for analytical purposes. Also, there are some problems reconciling between the departmental and ASYCUDA revenue collection figures, the later being surprisingly higher than the former. To date no satisfactory explanation for these discrepancies has been established. The system is now being transformed to GCMS which is expected to perform better in that the revenue accounting will be integrated into the system although it will take some time before the system is fully in place. Presently it will be used on a pilot basis in Terna, the airport and some other places and then extended gradually to other regions. IRS relies on physical movement of data as they are not computerized and are mostly working manually. The VATS has a semi-automated VIPS system, the headquarters and field offices use PCs but the computers are not linked. Transfer of data from the field to headquarters takes place via floppy disks. The VIPS is not yet the primary means of adntinistering and accounting for the VAT. This leaves open the possibility of discrepancies between paper and computer records. The Income Tax Service is as yet effectively uncomputerized expect for limited use of computers for maintaining records and monitoring of tax collection and adntinistration statistics and to some extent supporting tax adntinistration in the Large Taxpayer Office.

(c) Since each revenue agency gets its data in different ways and different formats. it is difficult to have an effective monitoring system in place. In the board meetings of the RAGB, comntissioners keep revising their collection figures till the last minute and as a result it is often difficult to cover all the three departments in a meeting and get a clear picture of their performance on a regular basis. The data collection and transmission have to be on a timely basis and the reporting format has to improve for better monitoring and meaningful discussions.

(d) The RAGB needs to monitor the costs of collection as well in order to increase the efficiency of the staff. The IRS has been doing this exercise for quite sometime but CEPS and V A TS are still in the process of learning.

5.2.2 Revenue Forecasting: At present. the level of revenue forecasting is quite poor. More manpower and training are needed on the pattern of the four-week training program offered last year in Accra for the middle level officials and the program on Tax Analysis and Revenue Forecasting (T ARF) at Duke University for the senior management. Only a long­term strategy with consistent efforts will be effective in the long run to build and sustain capacity.

5.3 The Revenue Agencies:

The present capacity for performance monitoring and revenue forecasting differs across the revenue agencies although the three agencies have a more or less similar units for research. monitoring and planning purposes.

5.3. 1 VA T Department

5.3. 1. 1 Research MonitOring and Planning Unit (RMPU):

(a) The main tasks of tltis division are monitoring and forecasting revenues. In terms of monitoring, revenue collections from seven field offices are reviewed on a monthly basis to ensure if they are on target and the reasons for the shortfall. if any. so that

83

Page 86: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

corrective measures may be undertaken. Although the department has a computer system, the field offices are not linked to headquarters and because certain offices do not have computers, some data is received in paper format. This renders the task of data collection and monitoring somewhat difficult. An additional problem arises because the RMPU does not have a regularly updated master file at its disposal (as is feasible within the computer capacity in the unit) to monitor VAT collection, arrears and refund performance on an ongoing basis and reconcile it with reported collections by office and sector.

(b) Sectoral monitoring - how each sector is performing in terms of revenue yield and whether revenues from a specific sector are keeping pace with the growth in that sector - is part of the department's agenda but currently it is not being done for lack of manpower and capacity. At the most there is an annual review, but no close monitoring takes place. This would become feasible if the department is able to acquire some more computer support.

(c) Currently the Monthly Receipts model is being used or revenue forecasting with suitable adjustments for growth and inflation and the forecasts have been generally found to be accurate and useful.

(d) The Ministry of Finance hands over tax policy proposals and suggestions received from various stakeholders to the RAGB for analysis, which in tum consults with the revenue agencies. Sometimes donor agencies have their suggestions while some policy initiatives, particularly revenue enhancing measures and discretionary changes, are suggested by the agency itself. The agency does respond to these various proposals, but it is not possible to do any in-depth study for two reasons. First, officers are busy pursuing revenue targets, trying to improve the existing procedures and removing bottlenecks in processing tax refund. Second, there is lack of adequate capacity for doing a good job . For instance, a study of distributional impact of VAT was done last year with the help of CEPA, but the study was not thorough. The VAT agency should have independent capacity for estimating tax incidence and doing an impact analysis of policy proposals and revenue enhancing measures that are initiated either by the ministry or RAGB or the agency itself on consumers in various income groups.

5.3.1.2 Manpower and Space Constraint: (a) Constraint of space has led to constraint of staff. In addition to the Deputy

Commissioner heading the research and monitoring division, there is only one other person who can engage in policy analysis. There is no one who can handle revenue forecasting on a full time basis within the division. The RMPU has been borrowing personnel from the operations division and this is clearly not sustainable over the long term.

(b) To be able to do the tasks ofrevenue forecasting, monitoring and impact! incidence analysis, more staff is needed and more training would be warranted.

5.3.2 Customs Excise and Preventive Services (CEPS) 5.3.2.1 Research and Monitoring Unit:

(a) Three units are involved in this work - Data Development Unit, Monitoring Unit, Policy Development and Analysis Unit.

(b) Data Development Unit is quite developed and they coordinate with ASYCUDA and the regional offices for collecting and inputting data. About 75% data comes from ASYCUDA and the rest is collected manually as hard copies from frontier units. It is expected that after switching to GeMS, there will be full computerization of all the

84

..

..

..

..

..

..

-

Page 87: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

..

-

..

units at headquarters and most of the units in the field. There may still be some interior units where data collection will remain manual but those will be very few.

(c) Monitoring Unit is primarily devoted to the monitoring performance of various operational units, which amounts primarily to monitoring of revenue collections.

(d) There is no sector wise performance monitoring. Only the imports and tariffs from some key commodities are monitored but that too is done on an ad hoc basis. There is no reconciliation between the revenue liabilities established by the ASYCUDA and GCMS and the collection accounts.

(e) The Policy Development and Analysis unit is in place and has also got requisite manpower and computers but it is not fully developed yet. This unit lacks the capacity to do any policy analysis and it simply examines those policy and administrative issues that might be affecting revenues. If some policy measure affects an economic sector in an unexpected way, this unit would go into the causes and effects. For instance, they would analyze the impact on revenues if certain tariff exemptions are given, but would not go into the effects on the importing industry. Again, ifCEPS gets a proposal from some stakeholder, they would examine its implications but this is mostly oriented to revenue implications rather than impact on consumers or the industries. In its present form, the functions of this unit overlap with that of the monitoring unit. The reason is simply a lack of capacity among the personnel for doing any kind of meaningful economic or impact analysis. Effective use of the extensive trade and customs databases that have been collected through the ASYCUDA and GCMS are not been used to analyze the impacts of proposed tariff changes. The unit needs extensive training.

(f) In terms of revenue forecasting, little is being done at present. The manpower is in place but there is no capacity outside of the limited training received to date. This is another area that needs extensive training.

5.3.3 Internal Revenue Services (IRS) 5.3.3.1 Research Planning and Monitoring Unit (RPMU):

(a) Revenue monitoring is done by the RPMU. This unit has three divisions: Tax examination, Tax audit, and Investigation.

(b) The tax examination division, headed by an assistant commissioner. deals "ith all procedural matters and verification of data and information coming from the field. The officials from this division take field trips and visit every field office at least once a year to check the authenticity of the facts and figures emanating from the field offices and submit a report to the IRS commissioner. Revenue collections are faxed once every month by the field offices. This information is used for flash reports. reconciliation with Bank of Ghana and for field verification.

(c) The tax audit division monitors and examines both the IRS work and the records of the companies.

(d) The investigation division investigates the departmental officers and also taxpayers in cases of evasion or tax fraud.

(e) The RPMU also does corporate planning for expansion of office space and staff of the IRS and modemization of its different parts. The biggest bottleneck in modernization is lack of computerization and this adversely affects the work of monitoring and forecasting.

(f) The RPMU has a statistical unit that maintains records of tax collections by type and office as well as other administrative performance data.

(g) Presently any policy analysis work is done within the IRS by the RPMU and not by the Ministry of Finance. While the primary focus is on assessing the revenue impact

85

Page 88: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

of a new measure, its effect on taxpayers is also examined. This work is currently done by a group of four people with knowledge of tax laws and legal issues and with good computer expertise. Proposals for policy change are received from several sources - MOF, employee organizations, businesses and IRS officers themselves.

(h) Sector analysis is done on an annual basis both by research unit of the RPMU and regional offices to assess the status of each major sector that should be yielding revenues. For instance, recently fishing and timber and related activities were examined. While the logging itself is not interesting from the revenue point of view, the downstream activities like sawing and furniture making were found to be good sources of income tax revenues.

(i) Currently there is no formal system of consultation with the businesses and other taxpayers. However, some groups (e.g. accountants) and individuals approach IRS with specific suggestions, or problems and questions of clarification and IRS examines those points and responds to them. Recently two proposals were presented to IRS - increasing pension contribution from 17.5% to 35% of salary and making the entire amount tax deductible and giving tax exemption to NGOs. Both proposals were examined and their revenue impacts assessed by RPMU.

(j) Extensive consultation among departmental heads within the IRS takes place every year before the budget exercise and new ideas/proposals are discussed. Also, their revenue implications and taxpayer impact are examined before presenting to the MOF in the annual retreat.

(k) In terms of revenue forecasting there is capacity limitation, lack of computers and problem of data. For example, prior to the recent data surveys, there has been no reliable data on the distribution of different types of taxpayer by tax bracket in order to estimate changes in the structure of the personal income tax. As a result, no serious work is currently being done in this area. The main problem is lack of computerization of data without which both monitoring and forecasting remain ad hoc exercises.

(I) Tax expenditure analysis is done on a regular basis and revenue impact of any concession or incentive is assessed. A sensitivity analysis is done to see the impact of alternative proposals and presented to the government. Tax expenditure analysis of old existing laws, however, is not done or presented to the government with a view to re-examining those measures.

(m) The IRS also has a significant legal capacity. It has a Legal Department that is involved in work on legal interpretation of the law within the court system and assisting in reviewing legal drafting conducted by Attorney Generals Departruent. Senior management in IRS also reviews amendments to Income tax legislation. Legal Department is also involved in negotiation of Double Taxation Treaties. In addition, a compliance unit will use court procedures in the collection of tax arrears. (n) Coordination with the Policy Analysis Division (PAD): Currently both the Policy

Analysis Division (PAD) and Economic Policy and Planning Cell (EPPC) within the MOF hold monthly meetings with tax agencies. However, the main discussions in these meetings focus on revenue targets and the ways of achieving the targets. Issues like data sharing, economic trends and their impact on revenues, inflation and other macro-economic variables and policy reforms are not taken up.

5.4 The Budget Process

86

..

-

-

..

..

..

..

..

..

-

Page 89: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

Tax policy and revenue forecasting is closely linked with the budget process. This section deals with the budgeting practice currently followed in Ghana and its interaction with policy analysis and revenue estimation work done by other departments in the govenunent.

(a) The Budget Framework: Ghana had initially a line-item or incremental budgetary process, but the government switched to a performance and activity based budgeting since 1980s. For budget preparation, govenunent now follows the Medium Term Expenditure Framework (MTEF) based on the logical framework of a "Strategic Planning Model". All the 37 ministries, departments and agencies (MDAs) are subdivided into five sectors: administrative services, economic services, infrastructure, social services and public safety services. Each MDA stands on four pillars - rational for existence, core business area (deliverables), key stakeholders (beneficiaries or recipients of the services) and core values (provide services for what. keeping quality standards). These four pillars should be reflected in the mission statement of the MDA. Each MDA has to define its broad objectives under MTEF. Also, the number of objectives cannot exceed nine due to budget limitations. Each objective should have a set of outputs or deliverables that should be prioritized. The MDA should then outline activities to be undertaken for achieving a particular objective or deliverables. Finally, five most essential inputs required to carry out a particular activity have to be identified and quantity, frequency and unit cost of each input has to be laid down to calculate the total cost of a particular activity. This exercise then extends into making of the budget (activity based budgeting). In many cases, this whole process remains incomplete but ideally this is the sequence to be followed.

(b) Annual Budget Making Exercise: The budget department gives a resource envelope to each MDA comprising both domestic (tax revenue and non-tax revenue generated by MDA itself) and foreign funding available (the top-down pan of MTEF). The re-enue estimates should ideally be based on forecasting model but currently it is being done based on the guesstimate of revenue agencies, discussions with them and extrapolation from the previous year's figures. A macro-model for revenue forecasting exists but has never been used because of some technical problems. Budget guidelines are prepared for the MDAs along with their budget ceilings and then they have to redo their strategic panning model based on the resource envelope (bottom up pan ofMTEF). The MDA's annual budget should reflect the objectives and milestones of "Mission 2020" and "Ghana Poverty Reduction Strategy".

(c) Budget Hearings: FITSt a policy hearing between Budget depanment and the MDA takes place. This is followed by a budget hearing where cash plan for spending is discussed between the MDA and the depanment of Finance. Finally. the parliamentary hearings take place after the budget is cleared by the cabinet.

(d) Budget Cycle: Ghana has adopted he calendar year as its fiscal year and the budget cycle starts in May-June when the budget guidelines and resource envelopes are intimated to the MDAs. The process continues till November when the Finance Minister submits draft estimate to parliament along with his statement on economic and fiscal policies. Provisional budget is approved in December before the house adjourns for Christmas when 25% of all personnel emoluments and 20% of administrative expenses (administration. overheads. cost of rendering services) and government investment are approved. The budget proposals are fully discussed in the months of January-February and the budget passed in March. Actual discretional) revenue measures are often only finalized and announced at this time after the expenditure-side of the budget has already been approved and is being implemented. Budget depanment issues sanctions for expenditures on a quanerly basis - general

87

Page 90: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

warrants are issued for emoluments and administrative expenses while special warrants are issued for costs of services and investment expenditures.

(e) Expenditure Monitoring: Expenditure returns are submitted for each activity on a quarterly basis by the respective agency. The Public Expenditure Monitoring Unit (PEMU) receives these reports from MDAs and analyzes them. Performance monitoring is not fully in place yet. Physical monitoring is done only partially by the Project Inspection and Monitoring Unit (PIMU). With the given manpower and capacity in the department, it is not possible to inspect all the capital works. MDAs submit certificates of expenditure on a quarterly basis. PIMU sends its officers to the field for actual inspection and verification of work done. Thus a system of sample inspections is in place.

(f) Impact or Outcome Analysis: Again, this is not being done at present due to lack of personnel and capacity. Some line agencies are expected to do this task on their own. For instance, Ghana Poverty Reduction Strategy Unit does conduct impact analysis in some selected sectors. Thus the budget implementation part has not yet fully developed.

(g) Revenue Forecast: The Director Revenues used to do some preliminary revenue forecasts in the past but now this work has been totally delegated to the RAGB. This type of forecast is made available to the budget department once every year at the beginning of the budget cycle.

(h) Tax Expenditure Analysis: While tax refunds (duty drawback or VAT refunds) are reported as line items in the budget, tax expenditure items are not reported as such. Currently, no one is doing this analysis in the Ministry of Finance even for internal consumption.

(i) Rationalization Exercise: No exercise has been done in the past to rationalize the structure of the ministry. There is a committee looking into this and also a civil service improvement program is underway.

G) Departmental Structure: The budget department is headed by a director who is assisted by four directors: Director for Public Safety and Infrastructure sector, director for Social Service sector, director for Economic Services sector and director for Administrative Services sector. Thus the administrative structure of the department follows the budget classification by sectors. Each director has one scheduled officer for each MDA under him.

5.5 Summary

To summarize, the following issues emerge from the preceding analysis of the existing state of affairs in the Ministry of Finance and the tax agencies:

(i) The focal point for decision making on issues of fiscal policy and fiscal management is the office of the Minister of Finance and the Policy Analysis Division (PAD) is meant to assist the minister in this task. However, given its limited manpower and capacity for tax policy analysis and revenue estimation, it only plays a marginal role in the process. In practice, all the policy and revenue related issues are passed on to the RAGB and the revenue agencies.

(ii) The RAGB has so far focused on revenue enhancing measures and monitoring of revenue performance of the three revenue agencies. It is in the

88

-

...

...

..

Page 91: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

.,

..

.,

.,

.,

..

.,

process of developing capacity for analyzing revenue implications of policy measures and forecasting of revenues.

(iii) The databases maintained by the tax agencies to date are not well organized and have generally not been reconciled with actual collections. The capacity to forecast revenue potential and compliance levels are, at present, limited. The agencies have been making serious efforts to improve their database and their revenue forecasting capacity, but a great deal of worK remains to be done.

(iv) Currently, the task of policy analysis is not being done in a thorough or systematic way by any agency. Most of the effort is focused on revenue enhancing measures and their revenue implications.

(v) At present, the budget department does not get the type of support it should receive on revenue estimation for budget making and budget revision from other departments in the govemment.

(vi) The sequencing of making firm revenue estimates and introducing firm revenue measures needs to be accelerated such that budget expenditures can be allocated from a more predictable revenue envelope in order to minimize the scope for unintended deficits .

6. Bridging the Institutional Gap

In order to meet this institutional gap at various levels in terms of database, policy analysis skills and forecasting techniques, establishment of an integrated system is envisaged. Under this proposed setup, a central unit in the government along with an assembly of sub-units in RAGB and tax agencies with appropriate coordination mechanism needs to be created simultaneously.

6.1 Tax Policy Unit in Ministry of Finance

The central unit at the govemment level, referred to as the Tax Policy Unit (TPU), will be the primary unit responsible for tax policy as expressed in tax legislation and budget statements. It will be primarily responsible for tax policy formulation process within budget process. Its main focus will be upwards to advise the Minister of Finance and horizontally to co-ordinate inputs with other players in budget process, the senior management in fiscal and budget departments .

The TPU would also help improve data collection and enhance govemment's capability for tax policy analYSis and revenue forecasting. In this role, the unit closely coordinates with the Economic Planning and Statistical Directorate in the govemment as well as the revenue agencies .

The present Policy Analysis Division (PAD) may be suitably re-engineered to serve the purpose by making it responsible for the specific tasks of policy analYSis and revenue estimation and augmenting it in terms of manpower and skills for performing the assigned tasks. The TPU would initially primarily be a research and monitoring unit with a focus on fiscal policy issues including the impact analYSis of tax laws and fiscal measures in terms of their economic efficiency, equity considerations and revenue implications. As the capacity of the TPU developed it would assume

89

Page 92: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

leadership in coordinating the policy consultations and analysis of tax policy proposals from all stakeholders and the distillation of efficient policy structures for consideration by the government. It would also gradually provide leadership in coordinating the legal drafting and interpretation of tax legislation. The TPU would draw upon the specialized capacities within the revenue agencies and Attorney­Generals office on a routine basis during the budget formulation process in the analysis, design and drafting of tax proposals. For example, the TPU may second members of the revenue agencies during the budget preparation process and also during any major tax or sector reform involving a particular tax structure such as pension or health financing reforms.

The unit will also actively interact with the Central Sank to ensure the harmony between the fiscal and monetary policies pursued by the state.

6.2 Sub-Units in RAGB and Revenue Agencies

At the same time, it is imperative that RAGS and the different tax agencies take concrete steps to establish their own units that would strengthen their respective management information systems and databases and will also sharpen their skills for sector analysis and revenue forecasting. These units should also have some capacity to perform economic and revenue analysis of their own proposals as well as suggestions received from other stakeholders with respect to their taxes with a special focus on the administration and compliance issues relating to tax policy changes. This will create an active group of sub-units around a central unit and their relationship will be complementary in nature. These units in the departments will be the primary source of database for the TPU and for one another. A close and strong working relationship among the sub-units within RAGS and the tax agencies, and between the central unit and the sub-units is essential for the success of this scheme.

The functions of the TPU at the Ministry level and the units within the RAGB and the tax agencies along with the inter-relationships are described below.

c. Functions of TPU and Sub-Units, Organizational Issues and Support Services

7. Functions of the Tax Policy Unit (TPU)

The main functions of the Tax Policy Unit are summarized below. The whole menu may appear to be rather large but it is better to have the comprehensive picture in view at this time. During the actual implementation stage, these activities may be suitably phased in.

90

..

-

-

..

..

Page 93: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

..

..

..

7.1 Policy Formulation Process - Tax reform design and discretioTUlry measures: TPU should be responsible for initiating all the tax policy measures ranging from major reform designs to more limited discretionary changes in specific areas of taxation. This implies the impact analysis of all the proposals and involves the following steps:

(a) Identification of policy proposals through I. Internal reviews

ii. Unsolicited proposals lll. Consultations with stakeholders iv. Government commitments from international or regional agreements

(b) Tax analysis of proposals in terms of their economic impacts. This implies: i. Estimation of the economic impacts of taxes in terms of revenue growth

(short and long term stability), efficiency, incidence, equity, administrative and compliance costs

ii. Assessing the impact of changes in one tax regime on revenues from other taxes

lll. Comparative analysis of different revenue raising measures (c) Scrutiny of non-tax revenue sources and curtailment of "nuisance taxes" (d) Decision making on tax and non-tax proposals (e) Issues of allocation of revenue functions to different levels of government (f) Co-ordination of legal drafting of accepted proposals (g) Ensuring revenue agencies conduct tax administration training (internal circulars and

training) and taxpayer education to implement new measures

7.2 Tax Analysis of specific sectors, special measures and non-tax economic policics

The TPU should be routinely analyzing the performance of specific economic sectors, special measures and economic policies that are not tax related but have a significant impact on tax revenues. The TPU has to have an economy-wide perspective so that tax policy is well integrated into the overall economic strategies of the government. This will include:

(a) Tax expenditure analysis arising out of special provisions of i. V AT exemptions ii. Investment incentives iii.Tariff exemptions for selected items/industries/free zones iv. Tax payment deferrals

These provisions tend to make the tax base narrow and the tax system less buoyant. Also, a concession to one party makes policy precedence and pressures for others. These concessions and exemptions amount to spending from the budget and should.. therefore, be spelled out clearly in the budget for the sake of transparency and public debate. This exercise should be performed both for existing provisions and new proposals.

(b) Sector studies of special sectors such as: i. Taxation of pensions, insurance and banking sector ii. Taxation of various types of savings, investment and export promotion

and its implication for overall savings and investment promotion in the economy

(c) Exploration of the possibility of charging user fees for public sector semces wherever it is possible to exclude the non-payers

91

Page 94: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

(d) Analyzing foreign tax and regional trade agreements and their impact on revenues and businesses (e) Intergovernmental fiscal relations (f) Evaluation of Impact of Non-Tax Economic Policies:

There are many economic policies that are not related directly to taxation but have profound impact on tax revenues. Such policies have to be analyzed carefully:

i. Estimate tax revenue impacts of alternative premises about the economy such as economic growth rate or growth rate of specific economic sectors ii. Measure the effect of macroeconomic policy changes on tax revenues such as (a) removal of quantitative restrictions on imports, (b) trade liberalization policies and their effect on import substitution and export promotion, (c) deregulation of certain economic activities, (d) currency fluctuation, (e) outcome of interest rate changes in terms of business profits

7.3 Monitoring of Revenue Potentiol and Forecasting of Revenues

This task is one of the core responsibilities of TPLI and involves:

(a) Estimation of country's tax capacity and identification of tax handles with growth potential

(b) Development of reliable data bases and forecasting models and their maintenance

(c) Base case forecasts by types of tax (d) Computation of elasticity and buoyancy of various types of taxes and the

entire tax system (e) Evaluation of the effect of inflation on tax revenues by tax types (f) Forecasting short-term seasonal patterns for annual cash flow management (e) Come up with a set of publications for the use of officials in line ministries and for

general circulation on a regular basis

7.4 Coordination with Budget Division in budget formulation and implementation:

The Tax Policy Unit has to play an important role during the annual budget cycle by providing inputs of revenue estimates and targets. More specifically through the following steps:

(a) Initial medium term revenue forecasts within the macro framework for preliminary budget estimates at least six months before a budget year

(b) Revised medium term forecast based on fiscal policy for final budget estimates about two-three months before a budget year

(c) After close of prior year, revise current year revenue forecasts for purposes of expenditure control

(d) Near close of current year, revise current year estimates for the purposes of expenditure control and basis for forecasting next year

7.5 Tax collection and administration performance monitoring

92

...

...

...

...

..

..

..

Page 95: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

."

..

."

."

revenue enhancing measures and their practical implications. During the months leading up to a budget, the members of this committee could be seconded to the TPU to work intensively on the analysis and preparation of budget revenue measures for presentation to the Minister.

(b) A revenue forecasting coordination committee with representatives from TPU, RAGB, tax agencies, the budget department and the directorate of Statistics to facilitate data sharing and division of labor in terms of the revenue forecasting exercise.

10.2 Needfor common "langUilge"

There will be a need for multi-level training for this system to work. The professionals . manning the TPU, the RPMU in RAGB, the RPMUs in revenue agencies and the officials of the fiscal division in the Directorate of Statistics should undergo a common basic training program in tax issues and forecasting techniques. Some personnel in the TPU, the RAGB and the revenue agencies should be imparted advance training on modeling and revenue forecasting. In addition, the senior management in the ministry of finance, the RAGB and the revenue agencies should receive an orientation seminar to familiarize them with the capacity building programs in the various units. This will enable them not only to appreciate the complete framework but also to have the right kind of expectations. All this can be achieved through the following series of programs.

(a) An intensive two-week in·country training program for all the assigned personnel (new recruits and old staff) in TPU, and RPMUs in RAGB and other agencies on major issues in tax analysis and techniques of revenue forecasting. The workshop would focus on the analytics of current tax policy issues facing the country. It may tum out to be a refresher course for some people who have been through the earlier four weeks program offered in Accra.

(b) A three-day hands on revenue forecasting workshop for each unit (TPU, RPMU in RAGB, IRS, VAT and CEPS). These wouldfocus on the installation and maintelUII/ce of data collection and analysis systems. 17.ese should be organized separately for professionals of each unit and should be held in the premises of the unit concerned. The entire exercise should take abow two weeks as these workshops may be held simultaneously if enough resource personnel are made available.

(c) Afour to six hour seminar, spread over two days ifnecessary, for the top policy makers to familiarize them with what is available to them through the central TPU and the sub-units in terms of policy analysis andforecasting instrun.enls.

(d) A short up-date or refresher in-country program after 6 months have elapsed_ It would be primarily for assessing the progress. problem solving and correctional steps if any.

(e) An extemaJ training program phased over three years for the senior and top level personnel in the Ministry of Finance. TPU. RAGB, the three revenue agencies and directorate of statistics to give them a wider perspectil'e based on international experience.

11. Tax Policy Support Services

In order for the TPU to function smoothly and effectively, a staged process of developing several support services wi/I be necessary. Some of these services are outlined below.

95

Page 96: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

(a) Tax monitoring systems: computer systems for tax collection data and modeling in the various units; computer network within TPU and possible connections to other RMPUs in RAGB and other agencies; computer network management and support

(b) Legal provision for sharing of revenue and taxpayer data and information with TPU personnel; legislative changes may be necessary

(c) A personnel policy with suitable remuneration and career development goals for tax policy officers

(d) Human resource development and provision for imparting specialized skills training to the staff in the areas of

Economics and tax policy Computer use/data analysis/modeling Accounting Tax Law, tax administration systems and procedures (where appropriate TPU

officers should attend training programs offered by revenue agencies on tax law and administration matters)

Special sectors analysis such as finance, mining, pensions, etc Management of tax policy and strategic development of tax system

12. Conclusion

The establishment of the TPU and strengthening of RPMUs in RAGB and revenue agencies and the development of support services will need to be undertaken on a time bound basis. By its very nature, it is a multidimensional task and some things have to be done simultaneously while some others have to be implemented in a close sequence. The process requires resources and detailed planning. However, as the system comes into existence and becomes effective, it should help improve the revenue side of the budget and make it more stable in due course.

Some of the initial steps in building a TPU in terms of the details provided above include:

t. Establishment of TPU within PAD with appropriate positions and terms of reference

2. Agreement on human resource strategy to sustain TPU 3. Installation of all existing tax databases and models along appropriate

computer capacity and support 4. Review of the revenue forecasting and revision cycle within budget

cycle. 5. Establishment of revenue monitoring committee and procedures 6. Establishment of tax policy committee and related responsibilities and

procedures. 7. Conduct of training programs to support these functions.

96

-

-

-

...

..

..

..

-

Page 97: Revenue Forecasting Models and Tax Policy Analysis for …pdf.usaid.gov/pdf_docs/pnacx581.pdf · WI August 2003 Revenue Forecasting Models and Tax Policy Analy~is for Ghana Submitted

Annex)

List of Government Of Ghana Contacted .. 1. Dr Anthony Osei, Advisor and Deputy Minister for Finance 2. Mr. Harry Owusu, Chief Executive, Revenue Agencies Governing Board of Ghana 3. Mrs. Janet Abrafi Opoku-Ayearnpong, Commissioner of Income Tax 4. Mr. Odartey Blankson, Commissioner of Value-Added Tax Service 5. Mr John Prempeh, Controller and Accountant General 6. Mr Kwabena Boadu Oku-Afari, PAD, Ministry of Finance 7. Mr. Samuel Danquah Arkhurst, PAD, Ministry of Finance 8. Mr. Paul Nkasah, Customs and Excise Preventive Service 9. Mr. Willy L.S. Moroy, Head of IRS LTO 10. Mr. Ken Bentsi-Enchil, Deputy Commissioner of VATS ., 11. Mr. George Blankson, Deputy Commissioner of V A TS 12. Mr. Nana Tjumasi, IRS 13. Mr. Nyamordey, IRS 14. Mr. Jackson Berko, IRS 15. Head of IRS Adabraka District Office

., 16. Head of IRS Osu District Office 17. Head of Computer Unit, IRS 18. Head of VAT LTO, VAT Industrial District Office

..

..

.,

fI 97