18
Rachel Licker, Marina Mastrorillo, Michael Oppenheimer, Valerie Mueller, Pratikshya Bohra-Mishra, Lyndon Estes, and Ruohong Cai Human Responses to Climate Variability: The Case of South Africa CFCC15 7 July 2015

Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Embed Size (px)

Citation preview

Page 1: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Rachel Licker, Marina Mastrorillo,

Michael Oppenheimer, Valerie Mueller,

Pratikshya Bohra-Mishra, Lyndon Estes,

and Ruohong Cai

Human Responses to Climate Variability:

The Case of South Africa

CFCC15

7 July 2015

Page 2: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

The Case of South Africa

• Widespread and consequential impacts projected

• Significant internal migration rates

• Large levels of temporary migration – Apartheid legacy

• Laws confined residence, allowed temporary moves for work

• High poverty / vulnerability rates, unevenly distributed

Leibbrandt et al., 2010

Temp change Southern Africa (D-J-F) Temp change RCP4.5, 2081-2100 (D-J-F)

50%

IPCC WGI AR5, 2013

• No studies considering climate and migration

Page 3: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Research Design

New empirical, cross-scalar study

Overarching question:

How has climate variability influenced internal migration

flows in South Africa in recent history?

Approach:

(1) Top-down, province-level (origin-destination flows)

(2) Bottom-up, individual-level (individual decisions to migrate)

Page 4: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Research Design

• With multiple models, scales, and data sources we seek to:

• Increase lines of evidence

• Improve climate variability-migration estimates

• Better understand climate-migration system in SA

• Better define uncertainty

• Prepare a basis for projections

Page 5: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Province-Level Approach: Data

• SA Census 1996, 2001, 2011; Community Survey 2007

• Information on:

- Demographics

- General health and fertility

- Education and employment

- Mortality

- Housing, households and services

- Migration (previous residence* – year of move)

* Note: Different spatial disaggregation on previous residence

for each census Mastrorillo et al., submitted

Page 6: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Migrant Definition

An individual who in year y = 1996, 2001, 2007, 2011 was

living in district j belonging to province p, and moved there

from province i≠p within the last 4 years

General features of internal migrants:

• Migrants: account for ~3.5% per yr of total population

• Younger people (15-30 years old): most represented

among migrants

• Proportionately more white individuals in migrating

population

• ~80% of migrants choose urban areas as destinations Mastrorillo et al., submitted

Page 7: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Bilateral Migration Flows

Definition: Bilateral migration flows mij are the number of

migrants moving from province i to district j (not belonging to

province i) during the 4 years before the Census year

1632 obs. Mastrorillo et al., submitted

Page 8: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Macro Approach: Gravity Model

Focus on origin-destination flows of migration

log(mij

t ) =k +f j

t +adij +bXi

t +gCi

t* +eij

t

bilateral migration flows mij

t

bilateral variables (log of distance and contiguity dummy) dij

time-destination dummies f j

t

origin controls (e.g., pop, pc gdp, ethnic group, urbanization,

unemployment, agriculture variables); τ = lag time Xi

t

climate-related variables at origin (e.g., frequency of droughts,

rain variability, temperature anomalies, soil moisture) Ci

t*

Mastrorillo et al., submitted

Page 9: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Preliminary Results

Variables Sign

Demographic :

• Population

• Share of white individuals

+

+

Geographic (bilateral variables):

• Origin-destination distance

• Contiguity

-

+

Socio –

economic :

• Real per-capita GDP

• Urbanization

• Unemployment

-

+

+

Agricultural : • Percentage of agricultural GDP

• Share of people working in

agriculture

-

-

Climatic : • Frequency of droughts

• Rain variability

• Temperature anomalies

• Soil moisture

+

+

+

-

Mastrorillo et al., submitted

Page 10: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Preliminary Results

• Results on climate are robust to:

• Alternative definition of migration flows (1 year flows)

• Higher spatial disaggregation (2011 cross-section analysis)

• Alternative estimation technique (Poisson)

• Conditioning flows to ethnic group:

• Impact of climate (and other socio-econ. variables at origin)

almost not significant for white migrants

• Conversely, strong impact on black African migrants

Page 11: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Next Steps

Recap

• Influence of climate variability on province-to-district migration

flows (aggregated census data)

• Next: Influence of climate variability on probability of individual

migration (individual-level survey data)

• Corroborate province-level findings

• Not aggregating to province level: more combinations of

migrant characteristics possible

• Disaggregated, longitudinal data: better to get at migrant

motivations

• Is the climate signal more pronounced with an aggregated or

disaggregated approach, or is it the same?

Page 12: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Next Steps

• Discrete-time event history model

• Following Gray and Mueller (2012):

Model of log odds of migration event : no migration event

log(Prit/Psit) = χit + χdt + αp + εit

• Population: Adults (15+) at risk of migration

(people exit once moved, died, or not tracked)

• Spatial resolution: District council (n=52)

• Spatial extent: South Africa

• Time period: 2008-2012 (three waves)

• Unit of analysis: person-year

Prit = Probability of migration event for individual i at time t

Psit = Probability of no migration for individual i at time t

Χit = Vector of individuals i in year t Χdt = Climate in year t, district council d

αi = Fixed effects for individual i

εit = Error term

Page 13: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Next Steps

• Discrete-time event history model

• Following Gray and Mueller (2012):

Model of log odds of migration event : no migration event

log(Prit/Psit) = χit + χdt + αp + εit

• With individual fixed effects, preliminary results:

• - maximum temp extremes, - migration

• + minimum temp extremes (nighttime), + migration

• + precipitation extremes (high levels), + migration

Page 14: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Next Steps

• Province-level fixed effects and migrant characteristic controls (e.g.

race, education, age)

• Multinomial outcomes (long vs. short distance), separate genders

• Additional climate measures (drought, additional observations)

log(Prit/Psit) = χit + χhit + χid + αp + εit

Prit = Probability of migration event for individual i at time t

Psit = Probability of no migration for individual i at time t

Χit = Vector of predictor variables for individual i in year t Χhit = Vector of predictor variables for household h of individual i in

year t Χid = Vector of predictor variables for individual i in district council d

αp = Fixed effects for province p

εit = Error term

Page 15: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Next Steps

• Some similarities, some differences across models

• Explore alternative model specifications (both models) to

test comparability of results

e.g. both at province-level

Page 16: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Conclusions

• In South Africa, preliminary results suggest:

• Droughts, rain variability, increased minimum temperatures, lower

soil moisture: increased flows

• Questions around maximum temperature

• Relationship strengths differ across population groups

• Provinces with more GDP from agriculture: less migration

Page 17: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

Thank you

Page 18: Licker r 20150707_1700_upmc_jussieu_-_amphi_durand

• Migration: One possible response to climate change

• Methods relevant to study of other social responses (e.g.

conflict, labor productivity)

• Existing evidence:

• Relationships between climate variability and…

• Local & long distance moves

• Immobility

• Policy relevance: positive and negative outcomes

• Migrants

• Sending & receiving regions

• Additional research needs: e.g. more longitudinal studies,

more consideration of interactions across scales

Climate Change and Human Migration