The Price Effects of Cash Versus In-Kind Transfers

Jesse Cunha (Naval Postgraduate School)

Giacomo De Giorgi (Stanford University)

Seema Jayachandran (Northwestern University)

September 2012

In-kind versus cash transfers

• Government transfers are often made in-kind

• One rationale is paternalism—boost consumption of certain goods

• Other potential reasons are self-targeting, political economy

• Weighed against constraining consumer choice and administrative

costs

• This paper: Price effects are another factor in this policy choice

Price effects of transfers

• Cash and in-kind transfers have an income (demand) effect

• Demand and prices for normal goods ↑

• In-kind transfers also inject supply into the local economy

– Setting where goods are provided (public housing) rather than

vouchers (Food Stamps)

• Influx of supply reduces prices

⇒ This paper: Empirically assess the size of price effects for cash

and in-kind transfers

Overview of paper

• Examine price effects of food transfer program in Mexico

– Randomized experiment across villages: in-kind transfers, cash

transfers, control group

• Find that prices decline for in-kind transfers relative to cash

transfers

• Larger effects in more remote villages, consistent with both...

– More closed economy

– Less competition

• Magnitudes small in non-remote villages; large in remote villages

• Differential effects for households that produce food

Effect of cash and in-kind transfers on prices

P

Q

MC

MR0MRin‐kind

Supply provided by govt

Income effect

MRcash

ΔPin‐kind  <   ΔPcash

ΔPcash

ΔPcash > 0

ΔPin‐kind

Hypotheses

• Cash transfers have positive income effect on prices

⇒ ∆pcash > 0

• In-kind transfers have positive income effect + negative supply

effect on prices

⇒ ∆pinkind < ∆pcash

• Sign of ∆pinkind is theoretically ambiguous without restrictions

on preferences

– But for, e.g., homothetic preferences, ∆pinkind < 0

Imperfect competition

• Can generate same predictions with imperfect competition

– Cash transfers cause price increase

– In-kind transfers cause prices to decrease, relatively

• Effects probably more likely to persist in the long run under

imperfect competition

– Long-run supply curve flatter than short-run curve

– If inherent barriers to entry lead to imperfect competition, may

persist in long run

Normative implications of price effects

• Lower price of transferred good furthers paternalistic goal of

encouraging consumption of transferred goods

• In-kind transfers redistribute from producers to consumers

(relative to cash transfers)

• If govt wants to tax producers to make transfers to consumers,

in-kind transfers could be a second-best tax instrument (Coate,

Johnson, and Zeckhauser 1994)

• With imperfect competition, no longer just a pecuniary externality

– Goods are undersupplied by the market

– Govt influx of supply could increase efficiency

Outline of rest of talk

• Background on PAL program + data

• Results

– Overall price effects for cash versus in-kind program

– Heterogeneity based on remoteness of village

– Quantifying the total effect

– Producer versus consumer households

• Conclusion

Transfer program we study

• Mexico’s food assistance program, Programa de Apoyo

Alimentario (PAL)

• PAL nationwide (in 2009): 200,000 households in 5,000 villages

• Targets poor households in villages too poor to be receiving

Oportunidades

PAL experiment

• Experiment in 2003-05: 208 villages

• Village-level randomization among eligible villages (small, rural,

poor) in 6 southern states

• Household-level targeting: 89% of households eligible

• 3 treatment arms – eligible HHs receive the following each month:

– Food box with 10 goods

– 150 pesos cash

– No transfer (control group)

Items in food box

Item TypeAmount per

box (kg)

Value per box (pre-program,

in pesos)

Calories, as % of total

box

Village change in supply (∆Supply)

(1) (2) (3) (4) (5)Corn flour basic 3 15.7 20% 1.00Rice basic 2 12.7 12% 0.61Beans basic 2 21.0 13% 0.29Fortified powdered milk basic 1.92 76.2 17% 8.62Packaged pasta soup basic 1.2 16.2 8% 0.93Vegetable oil basic 1 (lt) 10.4 16% 0.25Biscuits basic 1 18.7 8% 0.81Lentils supplementary 1 10.3 2% 3.73Canned tuna/sardines supplementary 0.6 14.8 2% 1.55Breakfast cereal supplementary 0.2 9.3 1% 0.90

Notes:(1) Value is calculated using the average of pre-treatment village-level median unit values. 10 pesos ≈ 1 USD. (2) ∆Supply measures the PAL supply influx into villages, relative to what would have been consumed absent the program. It is constructed as the average across all in-kind villages of the total amount of the good transferred to the village divided by the average consumption of the good in control villages in the post-period.(3) We do not know whether a household received canned tuna fish (0.35kg) or canned sardines (0.8kg); the analysis assumes the mean weight and calories throughout.(4) Biscuits are excluded from our analysis as post-program prices are missing.

Box of in-kind goods

PAL transfers being trucked into villages

Influx of cash and food was large

• Transfers are large: 19% of baseline monthly food expenditures

for recipients, 12% of total expenditures

• Given 89% eligibility, influx of 17% of baseline monthly food

expenditures for village

• Cash transfer was 8% of baseline total expenditures for village

Income effect of transfers

• Is the income effect from cash and in-kind transfers the same?

• Could be smaller effect for in-kind transfers – recipients value the

bundle less than its market value

• Could be larger for in-kind transfers, e.g., transfer signals quality

• Roughly, income effect is similar for both transfers in our setting

Equivalence of income effect for PAL program

• Cost for the in-kind box at prices in village was 206 pesos

• Government procurement cost was 150 pesos so set cash transfer

at this amount

• In-kind goods can’t be costlessly resold, so value is <206 pesos

What is value of PAL in-kind transfer?

• 116 pesos of the bundle was inframarginal, based on examining

control group’s consumption (Cunha 2012)

• In-kind HHs consume 34 pesos more of these goods than they

would have with cash transfer

• Another 56 pesos of transfer is extramarginal but not consumed;

transaction costs from resale

• Assume deadweight loss erodes two thirds of value in both cases;

90 pesos nominal value but valued at 30 pesos

• In-kind box valued at ∼146 pesos

• Even if consumers only value the inframarginal portion, different

income effect cannot explain magnitude of our results

Supply side of the market

• Food is sourced from manufacturers outside these villages

– We focus on only the local GE effects, ignoring possible Mexico-

wide effects

• Supply side within the village are grocery stores/shopkeepers

• Agricultural producers in the village supply substitute goods

Village stores

Village stores

Data

• Matched panel surveys of households and stores

– Pre-intervention (2003) / Follow-up (2005)

– Program underway for ≈ 1 year at follow-up

– For HH survey, interviewed random sample of 33 HHs per

village

• 14 of 208 villages not included because of missing data or program

began before baseline

• Final sample: 194 villages, 360 stores

• Randomization seems to have worked (Table 2)

Data on food prices

• 9 PAL food items

– 6 basic goods: corn flour, rice, beans, pasta, oil, fortified milk

– 3 supplementary goods: canned fish, packaged breakfast cereal,

and lentils

– Data for 10th PAL good (biscuits) not collected

• 51 non-PAL food items

• No price data for non-food items

Price data

• Our outcome variable is the good-store-village price (12,940

observations)

• Price surveys of local stores in each village

• Up to 3 stores per village but typically 1 or 2

• Looked for, or asked for, lowest priced product

• Incomplete baseline store data

Baseline price data: Unit values

• Household survey has food consumption, expenditure,

consumption out of own production by item, 7-day recall

• We calculate unit values (expenditures per unit purchased)

• Use median price for village-good

• Interpolate from other villages in municipality if missing

• Also use store prices, imputing missing values

Basic regression

pgsv = α+ β1InKindv + β2Cashv + φpgsv,t−1 + σXgv + εgsv

• g is good, s is store, v is village

• Control for lagged prices

• Control for indicator if lagged prices is imputed

• Cluster on village

• Two predictions are β1 < β2 and β2 > 0

Effect of transfer program on price of PAL goods

All PAL goods

Basic PAL goods 

All PAL goods

Basic PAL goods 

All PAL goods

Basic PAL goods 

All PAL goods

Basic PAL goods 

Outcome =  price price price price price price ln(price) ln(price)(1) (2) (3) (4) (5) (6) (7) (8)

In‐kind ‐0.037* ‐0.033 ‐0.036* ‐0.033 ‐0.032* ‐0.025 ‐0.037 ‐0.015(0.020) (0.020) (0.020) (0.020) (0.017) (0.017) (0.025) (0.022)

Cash 0.002 0.014 0.003 0.012 0.001 0.011 0.006 0.039(0.023) (0.027) (0.023) (0.026) (0.020) (0.022) (0.028) (0.026)

Lagged normalized unit value 0.027 0.127***(0.021) (0.042)

Lagged normalized store price 0.325*** 0.335***(0.052) (0.064)

Lagged ln(unit value) 0.857*** 0.861***(0.025) (0.037)

Observations 2,335 1,617 2,335 1,617 2,335 1,617 2,335 1,617

Effect size:  In‐kind ‐ Cash ‐0.039** ‐0.047** ‐0.038** ‐0.045** ‐0.034** ‐0.036** ‐0.044** ‐0.054**

H 0 :  In‐kind = Cash (p‐value) 0.02 0.04 0.03 0.04 0.03 0.04 0.04 0.02

Notes:  *** p<0.01, ** p<0.05, * p<0.1(1) The outcome variable in columns 1 to 6 is the post‐treatment price; it varies at the village‐store‐good level. It is normalized by good; the price is divided by the average price of the good across all observations in the control group. The outcome in columns 7‐8 is the logarithm of the post‐treatment store price, with no normalization.(2) Lagged unit value is the village median unit‐value, imputed geographically if missing (see text), and it varies at the village‐good level; it is normalized using the good‐specific control group mean.(3) Lagged store price is the village median store price, imputed with the village median unit‐value if missing (see text), and it varies at the village‐good level; it is normalized using the good‐specific control group mean.(4) Regressions in columns 1 to 4, 7 and 8 include one indicator for imputed pre‐program prices; those in columns 5 and 6 include two imputation indicators (see text).(5) Standard errors are clustered at the village level. 194 villages.

Robustness checks

• Results similar across specifications

– Control for store prices

– Use log prices

• No evidence that results are driven by changes in quality

Heterogeneity based on remoteness of village

• Two reasons to expect larger price effects in more remote areas

– Less open economy (steeper supply curve)

– Imperfect competition

• Measured as travel time to market with fresh meat, vegetables,

fruit

– Use village median of self-reports in household survey

Results on remoteness of the village

Above‐median 

remotenes

Below‐median 

remotenesAll villages

Above‐median 

remotenes

Below‐median 

remotenesAll villages

Outcome =  price price price price price price(1) (2) (3) (4) (5) (6)

In‐kind ‐0.030 ‐0.044* ‐0.050 ‐0.014 ‐0.045* ‐0.033(0.033) (0.024) (0.030) (0.027) (0.027) (0.031)

Cash 0.050 ‐0.029 0.013 0.062** ‐0.015 0.032(0.034) (0.031) (0.031) (0.031) (0.038) (0.036)

ln(Remoteness) x In‐kind ‐0.028 ‐0.007(0.033) (0.036)

ln(Remoteness) x Cash 0.023 0.033(0.033) (0.037)

Observations 865 1,470 2,130 603 1,014 1,471

Effect size:  In‐kind ‐ Cash ‐0.081*** ‐0.015  ‐0.076*** ‐0.030 H 0 :  In‐kind = Cash (p‐value) 0.00 0.56 0.00 0.35

Effect size:  ln(Remoteness) x In‐kind ‐ ln(Remoteness) x Cash ‐0.050** ‐0.040*

H 0 :  ln(Remoteness) x In‐kind = ln(Remoteness) x Cash (p‐value)

0.02 0.08

All PAL goods Basic PAL goods only

Notes:  *** p<0.01, ** p<0.05, * p<0.1(1) The outcome variable is the post‐treatment price; it varies at the village‐store‐good level. It is normalized by good; the price is divided by the average price of the good across all observations in the control group.  Standard errors are clustered at the village level.(2) Regressions control for the main effects of the interaction terms reported, as well as for the pre‐period normalized unit value and an indicator for imputed pre‐program prices (see text). (3) Remoteness is defined as the time required to travel to a larger market that sells fruit, vegetables, and meat. It is constructed as the village median of household self‐reports. It is missing (though below median) if no household in the village reports leaving the village to purchase these foods.

Testing between competition and closed economyexplanations

• Don’t have census of stores per village

• Poor quality data when tried to collect it retrospectively in 2011

• Suggestive evidence using number of stores in the data collection

• Price effects persist for a year, even though long-run MC curve

seems like it should be flat⇒ Also suggests imperfect competition

Results on number of stores

Outcome =  price price price price(1) (2) (3) (4)

In‐kind ‐0.030 ‐0.039 ‐0.018 ‐0.020(0.058) (0.062) (0.064) (0.069)

Cash 0.065 0.056 0.109 0.104(0.067) (0.071) (0.071) (0.076)

# stores x In‐kind ‐0.004 ‐0.006 ‐0.006 ‐0.007(0.026) (0.025) (0.029) (0.029)

# stores x Cash ‐0.032 ‐0.022 ‐0.047 ‐0.037(0.028) (0.030) (0.030) (0.031)

ln(Remoteness) x In‐kind ‐0.025 ‐0.006(0.034) (0.037)

ln(Remoteness) x Cash 0.022 0.026(0.035) (0.038)

Observations 2,130 2,130 1,471 1,471

Effect size:  In‐kind ‐ Cash ‐0.095* ‐0.096* ‐0.127** ‐0.124**H 0 :  In‐kind = Cash (p‐value) 0.06 0.06 0.02 0.02

Effect size:  # stores x In‐kind ‐ # stores x Cash 0.028  0.016  0.040* 0.030 H 0 : # stores  x In‐kind = # stores x Cash (p‐value) 0.15 0.47 0.05 0.18

Effect size:  ln(Remoteness) x In‐kind ‐ ln(Remoteness) x Cash ‐0.047** ‐0.033 H 0 :  ln(Remoteness) x In‐kind = ln(Remoteness) x Cash (p‐value) 0.03 0.16

All PAL goods Basic PAL goods only

Notes:  *** p<0.01, ** p<0.05, * p<0.1(1) The outcome variable is the post‐treatment price; it varies at the village‐store‐good level. It is normalized by good; the price is divided by the average price of the good across all observations in the control group.  Standard errors are clustered at the village level.(2) Regressions control for the main effects of the interaction terms reported, as well as for the pre‐period normalized unit value and an indicator for imputed pre‐program prices (see text). (3) Remoteness is defined as the time required to travel to a larger market that sells fruit, vegetables, and meat. It is constructed as the village median of household self‐reports. The number of stores is the number of stores included in the follow‐up price survey; a maximum of three stores were surveyed per village.

Effects on non-PAL goods

• Other food items are substitutes for PAL goods

• Identify subset of goods that are close substitutes for PAL goods

• Examine price effects for all other food items

• No data for non-food prices

Substitutes

All villagesAbove‐median 

remoteness

Below‐median 

remotenessAll villages All villages

Above‐median 

remoteness

Below‐median 

remotenessAll villages

Outcome =  price price price price price price price price(1) (2) (3) (4) (5) (6) (7) (8)

In‐kind ‐0.013 0.010 ‐0.024 ‐0.014 0.010 0.000 0.014 ‐0.005(0.025) (0.032) (0.036) (0.029) (0.019) (0.029) (0.024) (0.023)

Cash 0.027 0.035 0.024 0.024 0.009 0.039 ‐0.012 0.013(0.031) (0.034) (0.045) (0.033) (0.022) (0.042) (0.023) (0.034)

ln(Remoteness) x In‐kind ‐0.006 ‐0.022(0.034) (0.028)

ln(Remoteness) x Cash 0.002 0.014(0.036) (0.032)

Observations 1,442 498 944 1,307 10,648 3,765 6,883 9,698

Effect size:  In‐kind ‐ Cash ‐0.039  ‐0.025  ‐0.048  0.001  ‐0.039  0.026 H 0 :  In‐kind = Cash (p‐value) 0.15 0.41 0.22 0.95 0.34 0.24Effect size:  ln(Remoteness) x In‐kind ‐ ln(Remoteness) x Cash ‐0.008  ‐0.036 H 0 :  ln(Remoteness) x In‐kind = ln(Remoteness) x Cash (p‐value)

0.77 0.27

Set of PAL substitutes All non‐PAL goods

Notes:  *** p<0.01, ** p<0.05, * p<0.1(1) The outcome variable is the post‐treatment price; it varies at the village‐store‐good level. It is normalized by good; the price is divided by the average price of the good across all observations in the control group.  Standard errors are clustered at the village level.(2) Regressions control for the main effects of the interaction terms reported, as well as for the pre‐period normalized unit value and an indicator for imputed pre‐program prices (see text).   (3) Remoteness is defined as the time required to travel to a larger market that sells fruit, vegetables, and meat. It is constructed as the village median of household self‐reports.(4) Columns 1‐4 include 7 non‐PAL goods that we identified as PAL substitutes: corn tortillas, corn kernels, liquid milk, cheese, yogurt, potatoes, and plantains. Columns 5‐8 include all 51 non‐PAL goods.

Magnitude of the effects

• Multiply estimated change in prices by expenditure amount to

quantify price effect in pesos

• Expenditures per HH on PAL goods is 200 pesos and on non-PAL

goods, 1050 pesos per month (in control villages)

• Applies to non-recipients too

• Price effects small for non-remote villages

Magnitude of the effects

• Price effects have negligible effect on household purchasing power

for non-remote villages

• Price effects large in remote villages

⇒ Difference between in-kind and cash transfers equivalent to 60

extra pesos for a consumer (>30% of direct transfer)

Effects on food-producing households

• HHs are mainly consumers of the PAL goods, but some HHs are

agric. producers

• Welfare effects differ for producers

– e.g., Price increase from cash transfer is an extra benefit

• Production decisions might respond to the program

– e.g., Produce/sell more when prices rise in cash villages

• Income effect also could affect production, e.g., investment

affected if credit-constrained

Effects for food-producing HHs

Outcome = Farm profits

Farm costs

ln(Expenditure per capita)

ln(Expenditure per capita)

Asset index

Asset index

(1) (2) (3) (4) (5) (6)

In‐kind 143.87 134.01 0.115** 0.084(89.839) (119.511) (0.046) (0.075)

Cash 186.16* 345.32** 0.064 ‐0.040(106.082) (140.378) (0.052) (0.106)

Producer x In‐Kind  0.001 ‐0.018 0.077 0.055(0.060) (0.046) (0.115) (0.088)

Producer x Cash 0.087 0.015 0.266* 0.229**(0.068) (0.051) (0.142) (0.109)

Producer ‐0.161*** ‐0.003 ‐0.308*** ‐0.007(0.050) (0.036) (0.092) (0.071)

Control for pre‐period outcome? yes yes yes yes yes yesVillage FE yes yesObservations 4,924 5,038 5,534 5,534 5,571 5,571

Effect size:  In‐kind ‐ Cash ‐42.29  ‐211.31* 0.050  0.124 H 0 :  In‐kind = Cash (p‐value) 0.67 0.08 0.25 0.20

Effect size:  Producer x In‐Kind ‐ Producer x Cash ‐0.086  ‐0.033  ‐0.189  ‐0.174*H 0 :  Producer x In‐Kind = Producer x Cash (p‐value) 0.13 0.47 0.13 0.07

Notes:  *** p<0.01, ** p<0.05, * p<0.1(1) Observations are at the household level.  Standard errors are clustered at the village level.                                        (2) Profits and costs are measured in pesos and they are for the preceeding year; samples are trimmed of outliers greater than three standard deviations above the median.(3) Producer is an indicator for households that, at baseline, either auto‐consume their production or report planting or reaping produce or grain or raising animals.(4) Expenditure is the value of non‐durable items (food and non‐food) consumed in the preceding month, measured in pesos.(5) The asset index is the sum of binary indicators for whether the household owns the following goods: radio or TV, refrigerator, gas stove, washing machine, VCR, and car or motorcycle.

Summary of findings

• In rural Mexico, in-kind transfers cause prices of transferred goods

to fall relative to cash villages

• Results driven by more remote villages, perhaps because of less

supply-side competition

• In remote villages, in-kind transfers deliver 30% more to

consumers than cash transfers

• Welfare consequences of price changes are the opposite for

producers

– Price increase from cash transfers increases farm profits

– Lower prices from in-kind transfers hurt farm profits

Concluding thoughts

• Long-run effects might differ as supply adjusts

• Many other considerations when choosing in-kind vs. cash

transfers

– Paternalistic goals versus constraining household choices

– Govt may not be as efficient a supplier as the private sector

• But price effects are too large to ignore in remote villages

– High eligibility for social programs (typically ultra-poor)

– Fewer stores

– Less integrated with the outside economy

• In-kind transfers are one tool to reduce oligopolistic inefficiencies