Public Investments and Inclusive Growth in Uganda Stephen D. - PowerPoint PPT Presentation

Public Investments and Inclusive Growth in Uganda Stephen D. Younger, Sarah Ssewanyana, Ibrahim Kasirye and Elaine Hill Cornell University and Economic Policy Research Centre, Uganda

Overview of presentation • Why the focus on inclusive growth in Uganda • Infrastructure • Data and methods • Decompositions spatial differences in outcomes • Simulation of poverty impacts of infrastructure • Conclusions

Background and Motivation • Drivers of economic growth and poverty reduction are a matter of public concern in Uganda – Positive macroeconomic growth in past 20 years – Poverty reduction from 56% in 1992/93 to 25% by 2009/10 – worsening inequality; Gini increased from 0.33 to 0.42 – Pace of poverty reduction has slowed down in the recent past.

Large geographical variation in poverty over time Uganda: Trends in income poverty 1992/3-2009/10 Headcount poverty (%) 1992/93 1999/00 2002/3 2005/6 2009/10 All Uganda 54.9 33.4 38.8 31.1 24.5 Rural 58.5 37.4 42.7 34.2 27.2 Urban 27 9.6 14.3 13.7 9.1 By geographic location Central 45.6 19.3 22.5 16.4 10.7 Eastern 58.8 34.2 45.9 35.9 24.3 Northern 72.2 63.4 62.9 60.7 46.2 Western 53.1 25.9 32.9 20.5 21.8

Background and Motivation • Growth is necessary but not sufficient for sustainable poverty reduction in Uganda – Some authors suggest that inequality is driven by unequal access to public goods (Fox et al ., 2009) • E.g. education, basic infrastructure and transport services – Both education and returns to education are lower in rural areas and in urban area, outside the centre – How can public investment programs be used to reduce such discrepancies?

General Approach of the Paper • We estimate regression decompositions (Oaxaca-Blinder, 1973) using a nationally representative LSMS type survey. – Extent to which spatial differences in outcomes are due differences in endowments or returns to endowments • Skilled labour, infrastructure etc – Dependent variable: household consumption expenditure per adult equivalent – Regressions based on 4 geographical regions of Uganda.

Approach (contd) = β + ε ln( ) W X i i i i

Approach (contd) [ ] − = β − β E ln( W ) ln( W ) X X C N C C N N [ ] ( ) ( ) − = β − β + β − β E ln( W ) ln( W ) X X X X C N C C C N C N N N ) ( ) ( = β − β + − β X X X C C N C N N

Data • Uganda National Household Survey (UNHS) 2009/10 – Conducted during May 2009 and April 2010 by Uganda Bureau of Statistics – Covered all areas of Uganda – At least 6,800 households surveyed – Similar to Living Standards Measurement Surveys (LSMS) by the World Bank – Based on two stage stratified random sampling procedure – Enumeration areas (EAs) are principal sampling units (PSU) and 10 households randomly selected from each EA .

Regression Decompositions • Dependent variable – Household consumption expenditure per adult equivalent • Explanatory variables – Physical infrastructure i.e. Roads, Agricultural markets, health facilities, electricity, phone service, factory – Household head’s characteristics i.e. education attainment, age, spatial location • Infrastructure regressors measured at community level – E.g. electricity means household is resident in community/LC1 with electricity and not household itself has access

Issues with infrastructure variables • Presence of infrastructure in community/LC1 is co-linear • Risky to interpret the individual infrastructure coefficients (“returns”) • Hence we focus on sum of coefficients e.g. “all infrastructure” – weighted average of all coefficients on each infrastructure service for the endowments criteria – Weights are actual frequencies observed in the region • Interpretation: % change in welfare resulting from the presence of all of the infrastructure in the region – E.g. Impact of the infrastructure package observed in central Uganda is to increase welfare by 15%

Table: Average Endowments and Returns to Those Endowments, by Region, UNHS 2009/10 Returns Endowments Central Eastern Northern Western Central Eastern Northern Western Education Level of HHH < primary 7 11 12 13 0.38 0.45 0.46 0.41 primary 31 31 28 30 0.34 0.30 0.28 0.28 O-level 50 45 54 57 0.08 0.06 0.04 0.04 A-level 93 81 73 70 0.07 0.04 0.04 0.04 Post-secondary 129 137 125 107 0.02 0.01 0.00 0.01 All education: 26 21 18 19 Age of HHH 0 0 0 1 44 44 42 44 urban residence 19 23 22 18 0.29 0.08 0.14 0.08 Presence in LC1 of: Public health centre -6 -1 -9 13 0.06 0.08 0.06 0.11 Public hospital -9 8 0 24 0.00 0.02 0.00 0.00 All health infrastructure: -15 7 -9 37 All season feeder road -4 -7 12 5 0.81 0.65 0.39 0.69 Tarmac trunk road 3 14 8 13 0.23 0.13 0.11 0.07 Factory within 10km 15 -8 17 -1 0.16 0.03 0.01 0.05 Telephone service 6 13 7 3 0.71 0.41 0.19 0.64 Agricultural input market 6 -4 -4 2 0.32 0.38 0.27 0.43 Agricultural output market -7 3 11 -2 0.57 0.70 0.44 0.71 Community has electricity 25 9 5 16 0.54 0.17 0.14 0.12 All physical infrastructure: 15 4 11 9 All phys. infra. except electricity: 2 3 10 7

Endowments • Central region has better infrastructure – Interpretation: share clusters in the survey reporting having a service in its LC1/community – Exceptions are health centres and agricultural markets • Northern region is worse off compared to rest • Education attainment levels better in central – Difference mainly due to higher secondary and post-secondary levels – May reflect migration of the highly educated from other regions

Returns to endowments • Returns to education are higher in Central Region but difference not very large e.g. 26% vs.20% • Returns to health irregular – Negative in central and Northern Uganda probably due to collinearity – Possible that health infrastructure benefits children more than others – No contemporaneous variable to measure impact of previous health investments for today’s adults. • Returns to physical infrastructure higher in central and lowest in Eastern – Regional differences influenced by electricity • Overall: No equity-efficiency trade off for non electricity infrastructure – Only for electricity are returns higher in Central Region

Oaxaca-Blinder Decompositions • We decompose the means differences between central and each of the other regions – Central richest in incomes and infrastructure • Sources of welfare differences across regions – Decompose Eq3 further to report the returns and endowment effects for household characteristics • Separate the intercept coefficients from the “return effects” to determine unexplained effect.

Oaxaca-Blinder Decomposition of Average Welfare Differences, Central Region vs. Others (%) Eastern Northern Western Return effects 14 33 5 on household characteristics -9 -10 -17 on infrastructure 6 23 -12 sub-total, returns: Endowment effects 10 11 11 household characteristics 10 14 11 infrastructure 20 25 22 sub-total, endowments: 43 78 36 unexplained 69 126 47 Total difference in welfare Notes: "Unexplained" is the difference between the constant coefficient in Central region minus the listed region The return effects show the change in the listed region's average welfare if it had Central region's regression coefficients (returns to assets) The endowment effects show the change in the listed region's average welfare if that region had Central region's endowments Reported values are averages of the estimated effects using Central region and the listed region as the reference region Total differences in welfare are for the regression samples only

Oaxaca-Blinder decompositions • If other regions had Central’s endowments, average welfare levels would be in range of 20%-25% higher – 10%-11% due to better education endowments – 10%-14% due to better infrastructure in Central • Return to effects are more varied – Northern region would increase welfare by 23% – Eastern region only 6% – Western region worse off • Overall no equity-efficiency trade off for infrastructure investments – Returns are better in poorer regions including Northern

Poverty impacts of social and infrastructure investments

Simulated Poverty Impacts of Social and Infrastructure Services, by Region Variable All Uganda Central Eastern Northern Western HH head's educational attainment < primary 0.03 0.01 0.05 0.03 0.02 primary 0.06 0.05 0.07 0.05 0.03 O-level 0.02 0.02 0.02 0.01 0.01 A-level 0.03 0.03 0.03 0.01 0.01 post-secondary 0.01 0.02 0.01 0.00 0.00 All education 0.15 0.13 0.17 0.10 0.07 HH head's age Urban 0.03 0.03 0.01 0.02 0.01 Infrastructure in LC1 Health centre 0.01 0.00 0.00 0.00 0.01 All season feeder road 0.01 -0.01 -0.03 0.03 0.01 Tarmac trunk road 0.01 0.01 0.02 0.00 0.01 Factory w/ > 10 people 0.01 0.01 0.00 0.00 0.00 Telephone service 0.03 0.02 0.05 0.01 0.01 Agricultural input market 0.01 0.01 -0.01 0.00 0.00 Agricultural output market 0.00 -0.01 0.02 0.03 -0.01 Electricity 0.03 0.07 0.01 0.00 0.01 Notes: simulations based on the results in Table 5. All infrastructure 0.09 0.09 0.06 0.06 0.03 Reported value is the increase in the poverty headcount that would obtain if that variable were zero instead of its observed value .