Patterns of migration and employment in Malawi

Patterns of migration and employment in Malawi: Spatial data analysis

Anderson Gondwe

International Food Policy Research Institute

29 March 2017

IntroductionGrowing interest in spatial analysis in the literature

Standard non-spatial econometric techniques produce biased results in the presence of spatial

autocorrelation and heterogeneity (Anselin, 2003);

There has also been a rapid spread of geographic information systems (GIS);

Increased availability of data sets containing the location of observations (e.g. IHS, DHS, census).

Taking advantage of these developments and specific literature gaps, thus study pursues three main aims:

to understand the spatial and temporal patterns of employment and migration in Malawi through the

application of spatial data econometric techniques;

to analyse how long-term changes in age structure affect labour force participation;

to explore the effects of land reform policy on migration and employment

Outline

This presentation covers:

Data

Methodology

Descriptive spatial analysis (employment and migration)

Spatial panel regression results

Conclusions

Data

• Census data for 1987, 1998 and 2008 (from Integrated Public UseMicrodata Series (IPUMS)-International, Minnesota Population Centre)

• No migration data in 1998

• Matched district and traditional area GIS codes (for consistency over time)

• Created a panel of 24 districts and 178 small areas

• Can now match census data with other data for similar spatial analysis

• See next Malawi's admin units...

Data cont...Figure 1: Map of Malawi showing matched administrative boundaries

Within regions are districts andwithin districts are traditionalauthorities

District boundaries have notbeen consistent over time:

• 23 districts in 1966;• 24 districts in 1977;• 24 districts in 1987;• 28 districts & 4 cities in 1998• 27 districts & 4 cities in 2008.

Methodology• Spatial models take into account the spatial dependence among observations

(based on first law of geography “Everything is related to everything else, but

near things are more related than distant things" (Tobler, 1970, p. 236))

• Panel models discussed by Belotti, Hughes, and Mortari (2013):

• Spatial autoregressive model (SAR)

• Spatial Durbin model (SDM)

• Spatial autocorrelation model (SAC)

• Spatial error model (SEM)

• Generalised spatial random errors (GSPRE)

Spatial panel modelsThe general syntax is given as

𝑦𝑦𝑖𝑖𝑖𝑖 = 𝜏𝜏𝑦𝑦𝑖𝑖𝑖𝑖−1 + 𝜌𝜌𝜌𝜌𝑦𝑦𝑖𝑖𝑖𝑖 + 𝛽𝛽𝑋𝑋𝑖𝑖𝑖𝑖 + 𝜃𝜃𝜃𝜃𝑍𝑍𝑖𝑖𝑖𝑖 + 𝑎𝑎𝑖𝑖+𝛾𝛾𝑖𝑖+ 𝑣𝑣𝑖𝑖𝑖𝑖 ; 𝑣𝑣𝑖𝑖𝑖𝑖 = 𝜆𝜆𝜆𝜆𝑣𝑣𝑖𝑖𝑖𝑖 +𝜇𝜇𝑖𝑖𝑖𝑖

Where:• y denotes outcome variable; • W is spatial weights matrix; • Wy denotes spatially lagged dependent variable (DLAG); • X denotes explanatory variables; • D is the matrix for the spatially lagged regressors; • Z denotes spatially lagged regressors; • E is the weight matrix for the spatial-error term.

EmploymentFigure 2: Spatial distribution of status in employment for 2008

Labour force participation (1)

1000 500 0 500 1000

00-04

05-09

10-14

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

65+

1987

Male labour force Female labour force Female population Male population

Labour force participation (2)

1000 500 0 500 1000

00-04

05-09

10-14

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

65+

1998

Male labour force Female labour force Female population Male population

Labour force participation (3)

1000 500 0 500 1000

00-04

05-09

10-14

15-19

20-24

25-29

30-34

35-39

40-44

45-49

50-54

55-59

60-64

65+

2008

Male labour force Female labour force Female population Male population

Employment…Figure 3: Spatial distribution of employment type in 2008

Educational attainmentFigure 4: Spatial distribution of educational attainment in 2008

Educational attainment…Figure 5: Assets, schooling and fertility

MigrationDefining migration:

-If district of residence one year ago is different=1, if same=0;

-Movement within district or city is not migration.

• Regional migration

• District-level migration

-Generally, people stay in their home regions but move across districts

• Mostly rural-rural migration

• See next...

Regional migration

Table 1: Inter-regional migration transitions by genderYear

North Centre South Total North Centre South Total1987 North 96.73 1.91 1.36 100 97.71 1.38 0.91 100

Centre 0.68 97.59 1.73 100 0.44 98.38 1.19 100South 0.48 1.83 97.69 100 0.26 0.94 98.80 100Total 11.56 40.13 48.31 100 11.49 38.28 50.23 100

2008North 90.51 5.65 3.84 100 91.57 5.16 3.27 100Centre 1.78 94.30 3.91 100 1.57 94.83 3.60 100South 1.17 6.21 92.61 100 0.81 4.37 94.82 100Total 12.37 43.36 44.27 100 12.52 41.91 45.57 100

Male Female

District level migrationFigure 6: Population proportions that moved by gender and district

Spatial dependenciesLaw of geography (near things are more related than distant things)Table 2: Spatial dependencies

Variables Moran's I SD p-value Geary's c SD p-value

Migration 0.397 0.009 0.000*** 0.612 0.011 0.000***

Employed 0.063 0.009 0.000*** 0.930 0.013 0.000***

Years of schooling 0.167 0.009 0.000*** 0.793 0.013 0.000***

Sex 0.153 0.009 0.000*** 0.857 0.013 0.000***

Age 0.056 0.009 0.000*** 0.954 0.015 0.002***

Married 0.127 0.009 0.000*** 0.863 0.012 0.000***

Dependency ratio 0.075 0.009 0.000*** 0.927 0.013 0.000***

Global spatial autocorrelationFigure 7: Spatial correlogram

Moran's I spatial correlogram(mean) migration

I

Distance bands0-100000 0-200000 0-300000 0-400000 0-500000 0-600000 0-700000

0.00

0.20

0.40

0.60

0.80

Local spatial autocorrelationPositive Moran and negative Geary z-values indicate similarity

Figure 8: Local spatial autocorrelations for employment and migration

What about the land reform policy?

Malawi Community Based Rural Land Development Project

• Aimed at addressing social conflicts related to unequal access to land

• Targeted landless (mostly Southern region districts)

• Launched in 2004 but run through 2009.

• Households voluntarily purchased plots from fallow estates & resettled

• The World Bank (2012) has rated the land reform policy as satisfactory

and exceeding expectation.

Set up of difference-in-difference

Regression set up of DID• Interest: whether land reform policy had effects on migration and

employment• Observations: 320 and 36 small places are in the control and treatment

groups, respectively• Problem: the two groups may be different from each other• Key assumption: the two groups would follow the same time trend in the

absence of the treatment.• Fixed effects controls for the influence of unobserved and time invariant

characteristics• Robustness: Combine PSM with DD by matching units in common support

Set up of difference-in-difference…• Model summary:

y𝑖𝑖𝑖𝑖 = time ∗ 𝑡𝑡𝑡𝑡𝑡𝑡𝑎𝑎𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖 + 𝛽𝛽𝑡𝑡𝑡𝑡𝑡𝑡𝑎𝑎𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖 + 𝛿𝛿𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑡𝑖𝑖 + 𝜀𝜀𝑖𝑖𝑖𝑖

DID = time ∗ treatment; this is our interest parameter

• Before 2004 (0); post-2004 (1). • Treatment group (1) : small geographical areas in Thyolo and Mulanje• Control (0): for the rest of the country.

• We add spatial effects since migration and employment are spatial

Effects of land reform policy on migrationMale Female

OLS Non spatial SAR OLS Non spatial SARTime -0.180*** -0.188*** -0.089*** -0.153*** -0.076** 0.001

(0.010) (0.028) (0.029) (0.009) (0.031) (0.025)Treated -0.039*** -0.036***

(0.014) (0.012)DID 0.035* 0.029 0.019 0.036** 0.038* 0.027**

(0.020) (0.020) (0.014) (0.017) (0.020) (0.013)Schooling -0.004 -0.004 -0.009 0.000 -0.022 -0.023**

(0.004) (0.014) (0.009) (0.003) (0.014) (0.009)Age 0.056 0.106 0.097* -0.045 -0.123 -0.130**

(0.053) (0.083) (0.056) (0.036) (0.078) (0.052)Age squared -0.001 -0.002 -0.002** 0.001 0.002 0.002**

(0.001) (0.001) (0.001) (0.001) (0.001) (0.001)Married 0.229** 0.427** 0.389*** 0.083 0.174 0.108

(0.095) (0.177) (0.120) (0.061) (0.152) (0.102)Employed -0.036 -0.08 -0.081** -0.112*** -0.116*** -0.114***

(0.039) (0.057) (0.038) (0.022) (0.035) (0.023)Assets 0.039*** 0.034** 0.034*** 0.020*** -0.002 -0.005

(0.005) (0.014) (0.010) (0.005) (0.015) (0.010)W*migration 0.726*** 0.783***

(0.160) (0.131)Observations 356 356 356 356 356 356Spatial units 178 178 178 178 178 178

Effects of land reform policy on agric. employmentMale Female

OLS Non spatial SAR OLS Non spatial SARTime -0.289*** -0.390*** -0.306*** -0.316*** -0.443*** -0.252***

(0.035) (0.045) (0.106) (0.028) (0.056) (0.087)Treated 0.080*** 0.032

(0.027) (0.021)DID 0.036 0.025 0.025 0.042 0.048* 0.045**

(0.037) (0.029) (0.020) (0.029) (0.028) (0.019)Schooling -0.064*** 0.077** 0.074*** -0.042*** 0.060 0.057**

(0.010) (0.033) (0.023) (0.006) (0.040) (0.027)Age 2.989*** 0.225 0.237 1.840*** 0.986* 1.020***

(0.549) (0.777) (0.538) (0.237) (0.507) (0.348)Age squared -0.048*** -0.004 -0.004 -0.030*** -0.016* -0.016***

(0.009) (0.012) (0.009) (0.004) (0.008) (0.006)Married 0.109 -0.071 -0.103 0.217 -0.100 -0.103

(0.282) (0.425) (0.297) (0.146) (0.378) (0.260)Dependency 0.677*** -0.262 -0.257* 0.199* 0.105 0.111

(0.188) (0.202) (0.140) (0.118) (0.155) (0.106)W*employed 0.268 0.569**

(0.324) (0.233)Observations 356 356 356 356 356 356Spatial units 178 178 178 178 178 178

Effects of land reform policy on gov. employmentMale Female

OLS Non spatial SAR OLS Non spatial SAR

Time 0.042*** 0.063*** -0.010 0.017*** -0.006 -0.032**(0.007) (0.015) (0.011) (0.006) (0.014) (0.013)

Treated -0.011** -0.011**(0.005) (0.004)

DID 0.029*** 0.027*** 0.022*** 0.015** 0.013* 0.012**(0.008) (0.009) (0.006) (0.006) (0.007) (0.005)

Schooling 0.015*** -0.005 0.004 0.008*** 0.027*** 0.029***(0.002) (0.011) (0.007) (0.001) (0.010) (0.007)

Age -0.335*** -0.415 -0.344** -0.116** 0.455*** 0.453***(0.113) (0.252) (0.162) (0.049) (0.130) (0.089)

Age squared 0.005*** 0.007* 0.006** 0.002** -0.007*** -0.007***(0.002) (0.004) (0.003) (0.001) (0.002) (0.001)

Married -0.004 -0.008 -0.058 -0.019 0.161* 0.134**(0.058) (0.138) (0.089) (0.030) (0.097) (0.067)

Dependency 0.039 0.061 0.04 0.006 0.044 0.041(0.039) (0.065) (0.042) (0.024) (0.040) (0.027)

W*employed 0.913*** 0.648***(0.060) (0.195)

Observations 356 356 356 356 356 356Spatial units 178 178 178 178 178 178

Effects of land reform policy on self and private employment (OLS)

Male FemaleSelf Private Self Private

Time 0.228*** 0.185*** 0.213*** 0.108***(0.020) (0.018) (0.019) (0.013)

Treated -0.018 -0.007 0.002 -0.014(0.015) (0.014) (0.015) (0.010)

DID -0.043** -0.040** -0.021 -0.041***(0.021) (0.019) (0.020) (0.014)

Schooling -0.001 0.020*** 0.008* 0.011***(0.006) (0.005) (0.004) (0.003)

Age -1.436*** -0.298 -0.691*** -0.106(0.310) (0.284) (0.161) (0.110)

Age squared 0.023*** 0.005 0.011*** 0.002(0.005) (0.005) (0.003) (0.002)

Married -0.274* -0.047 -0.241** 0.090(0.159) (0.146) (0.099) (0.068)

Dependency -0.134 -0.324*** -0.030 -0.129**(0.106) (0.097) (0.080) (0.055)

Constant 22.764*** 4.543 10.835*** 1.62(4.826) (4.416) (2.428) (1.658)

R-squared 0.580 0.572 0.657 0.428Observations 356 356 356 356

Discussion

Gender dimension• The gender dimension of our results has implications for policy and poverty

reduction because the participation of females in employment contributes toeconomic growth.

• For example, in a recent study, the Food and Agriculture Organisation (2011)found that by closing the resource gap between men and women in developingcountries, agricultural yields for females could potentially improve by between20% and 30%.

• As a result, total agricultural output of developing countries would grow bybetween 2.5% and 4.0%. In turn, the levels of malnutrition would decline bybetween 12% and 17%, which would globally imply that 100 to 150 million peoplewould be lifted out of hunger.

Conclusions• Clear evidence of spatial spillovers

• Most indicators are heterogeneous within the country but at low geographical level

• Positive effects of land reform on migration and employment

• There are gender effects: agricultural and government employment improved, more particularly for the females

• Paper allows the feasibility of integrating census data with other data for similar spatial analysis

• Rural-urban migration is not as massive as perceived. On the contrary, the observed patterns of migration in Malawi are largely rural-rural.

Implications

• A similar programme can work for the whole country

• However, there will be need for access to functioning services such as

health, water, transport and markets.

• Further study can use individual level data rather than aggregated data

-aggregated data ignores the rich cross-sectional evidence of the individual

observations and fails to say anything about the industries, firms and individuals

themselves (Arbia, 2016).

Balancing tests for base periodDescription Weighted Variable(s) Mean Control Mean Treated Diff. t P-valueMale

Migration 0.150 0.106 -0.044 3.12 0.0021***Schooling 2.565 2.472 -0.093 0.74 0.4600Age 20.736 20.622 -0.115 0.72 0.4720Married 0.333 0.326 -0.007 1.48 0.1400Employed 0.646 0.636 -0.01 1.18 0.2410Assets -0.410 -0.500 -0.09 0.56 0.5770

FemaleMigration 0.148 0.093 -0.054 3.69 0.0003***Schooling 1.275 1.233 -0.042 0.4 0.6888Age 21.681 21.758 0.077 0.35 0.7268Married 0.364 0.361 -0.003 0.82 0.4132Employed 0.623 0.627 0.004 0.22 0.8271Assets -0.463 -0.595 -0.132 0.9 0.3707