The impact of the Kenya CT-OVC Program on household spending. Kenya CT-OVC Evaluation Team Presented by Tia Palermo Naivasha, Kenya January 2011

Transcription

1 The impact of the Kenya CT-OVC Program on household spending Kenya CT-OVC Evaluation Team Presented by Tia Palermo Naivasha, Kenya January 2011

2 Kenya Cash Transfer Program for Orphans and Vulnerable Children (CT- OVC) Government s flagship social protection program. Reached 100,000 households and 230,000 OVC across the country as of mid Targets households who are ultra-poor and contain an OVC Eligible households receive a flat monthly transfer of 21 USD (Ksh 1500). Care and protection of the resident OVC is hh s responsibility for receiving the cash payment. Currently no punitive sanctions for noncompliance.

3 Aim of the paper To investigate whether the CT-OVC has changed the preferences of households in terms of their consumption behavior

4 Outline 1. CT-OVC Evaluation Design and Sample 2. Impact of the CT-OVC on Expenditure Patterns 3. Ex-Ante Predictions of Program Impacts 4. Testing for Elasticity Changes 5. Conclusion

5 1. Evaluation Sample and Design 2,759 households in the evaluation sample 2/3 assigned to treatment 17% drawn from wider population and does not enter into estimation of program impacts Among true Treatment (T) and Controls (C) [N=2294], 1907 were again interviewed in 2009 (17% attrition rate) Final sample for analysis: N=1849 in the panel (excludes households with missing information on key variables)

6 Category Food Domestic Health and Hygiene Transportation and Communication Household items Items included All foods Paraffin, kerosene, charcoal, firewood, washing powder, laundry soap, detergents, matches, candles, batteries, bulbs, textiles, kitchen equipment, lanterns, lamps, torches toilet soap, other toiletries, hair cut and personal services, medicines and medical supplies (excluding AIDS drugs), consultation and treatment fees, laboratory and diagnostic fees, visits to traditional healers, hospitalization fees, other health expenditures bus fares, matutu, taxis, other transport expenses, transport to and from health facility, transport to and from school Housing Education Clothing Other Alcohol and Tobacco water and utilities [Note: excludes rent due to the fact that only a small proportion of households (8.8% at baseline) reported paying any rent.] tuition fees, registration fees, exam fees, other fees, private tuition, uniforms including school shoes, school supplies including textbooks men s clothing, children s clothing, women s clothing, footwear books, notebooks, newspapers, stationery, TKK, recreation, personal items (umbrella, watch, lighter, etc.) alcohol and tobacco

7 Food categories Category Cereal Tubers Meats and Fish Dairy Fruits and Vegetables Other Foods Foods Eaten Out Items included maize, rice, other grains, bread cassava, potatoes, arrow roots, yams, cooking banana fresh fish, dried fish, smoked fish, beef, chicken, other meat milk, eggs sukuma wiki, other vegetables, bananas, other fruits, beans, pulses, nuts cooking fat, cooking oils, sugars, spices, tea leaves/tea bags, coffee and other nonalcoholic drinks meals eaten out and other ready-made foods

8 Table 1. Mean Spending, broad household items Panel A. Broad household item groups (1) (2) (3) (4) (5) Treatment Control Wave 1 Wave 2 Wave 1 Wave 2 AE Monthly Expenditure p-value Total Food Domestic Health, Hygiene Trnspt, Communication Housing Education Clothing Other Alcohol, Tobacco Shares Food Domestic Health, Hygiene Trnspt, Communication Housing Education Clothing Other Alcohol, Tobacco

9 Table 1. Mean Spending, food groups Panel B. Food groups (1) (2) (3) (4) (5) Treatment Control Wave 1 Wave 2 Wave 1 Wave 2 AE Monthly Expenditure p-value Cereals Tubers Meat, Fish Dairy Fruit, Vegetables Other Food Food Eaten Out Shares Cereals Tubers Meat, Fish Dairy Fruit, Vegetables Other Food Food Eaten Out

10 2. Impact of the CT-OVC on Expenditure Patterns

11 Table 2. DD estimates, broad groups Panel A. Broad household item groups (1) (3) (5) (7) (9) (11) (13) (15) (17) Food Domestic Health Trnsprt&Comm Housing Education Clothing Other Alc & Tobacco Treat (β3) (-0.20) (-1.00) (-3.19) (-3.06) (-0.36) (0.47) (1.10) (0.16) (1.32) Time (β4) (5.44) (-5.84) (0.32) (-5.22) (-3.88) (-1.01) (-0.31) (3.12) (-0.29) Treat*Time (β5) (3.62) (1.19) (2.58) (2.57) (1.59) (0.79) (3.57) (-0.94) (-1.69) Constant 1, (33.55) (31.28) (11.22) (11.62) (8.40) (2.03) (7.45) (2.45) (4.27) Observations 3,698 3,698 3,697 3,698 3,698 3,698 3,698 3,698 3,698 R-squared

12 Table 2. DD estimates, food groups (1) (3) (5) (7) (9) (11) (13) Cereal Tubers Meat & Fish Dairy Fruit & Veg Other Food Eaten Out lnadjexp (β2) Treat (β3) (-0.32) (4.50) (-0.83) (-0.67) (0.28) (-0.31) (-0.71) Time (β4) (11.25) (-9.61) (-0.36) (0.55) (3.04) (-0.34) (1.73) Treat*Time (β5) (3.24) (-4.02) (3.32) (4.86) (0.64) (2.02) (0.49) Constant (25.60) (16.47) (15.81) (9.69) (29.06) (23.67) (5.54) Observations 3,698 3,698 3,698 3,698 3,698 3,698 3,698 R-squared

13 Key questions Are these effects high or low? Do the estimated impacts come solely from the increase discretionary money available to T households, or has the program somehow changed fundamental behaviors with respect to spending patterns? For example, have households changed their rates of time preference and become more forward looking because they can rely on a steady, reliable source of income, and have thus fundamentally changed their spending patterns accordingly?

14 3. Ex-Ante Predictions of Program Impacts

15

16 Table 3. Engel Curve Estimates and Expenditure Elasticities, HH items Panel A. Broad household item groups (1) (2) (3) (4) (5) (6) (7) (8) (9) Food Domestic Health Transprt&Comm Housing Education Clothing Other Alc & Tobacco lnadjexp (β2) (-11.73) (-6.75) (3.07) (13.11) (-1.19) (7.65) (6.56) (6.74) (0.27) Treatment (1.16) (-0.80) (-3.16) (-1.28) (1.19) (1.47) (0.88) (-0.46) (0.21) Constant (22.16) (12.95) (-0.53) (-10.11) (2.38) (-7.85) (-5.45) (-5.99) (0.53) N 1,849 1,849 1,849 1,849 1,849 1,849 1,849 1,849 1,849 R-squared Elasticity

17 Table 3. Engel Curve Estimates and Expenditure Elasticities, Food groups Panel B. Food groups (1) (2) (3) (4) (5) (6) (7) Cereal Tubers Meat & Fish Dairy Fruit & Veg Oth foods Eaten Out lnadjexp (β2) (-17.81) (-2.81) (12.24) (1.88) (-3.26) (-10.45) (4.45) Treatment (0.74) (2.61) (-0.68) (-0.15) (1.08) (-0.33) (0.18) Constant (22.04) (5.85) (-8.50) (-1.13) (9.43) (12.97) (-3.73) N 1,849 1,849 1,849 1,849 1,849 1,849 1,849 R-squared Elasticity

18 Ex-ante predictions of program impact on expenditure shares % change in spending in category i=20*ei Ex-ante predicted impact=(%change in spending in category i)*(mean spending at baseline in category i)

19 Table 4. Ex-ante predictions of program impact on expenditure shares, hh categories Panel A. Broad household item groups (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Food Domestic Health Transprt&Comm Housing Education Clothing Other Alc & Tobacco Total Pooled elasticity % change in ADJEXP % change in spending on group mean spending at baseline Ex-ante predicted impact Actual DD impact estimate (36.554) (9.697) (13.407) (11.761) (5.100) (11.275) (6.779) (4.362) (2.196) Treatment only elasticitiy % change in ADJEXP % change in spending on group mean spending at baseline Ex-ante predicted impact Actual DD impact estimate (36.554) (9.697) (13.407) (11.761) (5.100) (11.275) (6.779) (4.362) (2.196) Note: Standard errors of DD coefficient estimates in parentheses.

20 Table 4. Ex-ante predictions of program impact on expenditure shares, food groups (1) (2) (3) (4) (5) (6) (7) (8) Cereal Tubers Meat & Fish Dairy Fruit & Veg Oth foods Eaten Out Total Pooled elasticity % change in ADJEXP % change in spending on group mean spending at baseline Ex-ante predicted impact Actual DD impact estimate (14.455) (3.46) (12.306) (6.101) (10.21) (9.290) (3.167) Treatment only elasticity % change in ADJEXP % change in spending on group mean spending at baseline Ex-ante predicted impact Actual DD impact estimate (14.455) (3.46) (12.306) (6.101) (10.21) (9.290) (3.167)

21 4. Testing for Elasticity Changes

22 Table 5. Estimates of Program Impacts on Expenditures Shares, Triple Differences Panel A. Broad household item groups (1) (2) (3) (4) (5) (6) (7) (8) (9) Food Domestic Health Transprt&Comm Housing Education Clothing Other Alc & Tobacco lnadjexp (β2) (-8.59) (-5.08) (3.58) (11.78) (1.45) (4.89) (1.72) (2.39) (-2.25) Treat (β3) (-1.57) (-1.03) (1.67) (3.67) (2.58) (0.74) (-2.62) (0.34) (-2.86) Time (β4) (-5.57) (-2.66) (2.22) (5.47) (1.65) (3.59) (-1.17) (-0.23) (-3.34) Treat*Time (β5) (2.15) (2.78) (-2.71) (-3.21) (-2.73) (-1.56) (1.58) (0.50) (2.65) lnpce*treat (β6) (1.52) (1.00) (-1.96) (-3.84) (-2.43) (-0.61) (2.68) (-0.30) (2.89) lnpce*time (β7) (5.19) (1.86) (-2.13) (-5.99) (-1.94) (-3.61) (1.03) (0.57) (3.22) lnpce*treat*time (β8) (-2.08) (-2.77) (2.76) (3.28) (2.69) (1.49) (-1.34) (-0.61) (-2.76) Constant (12.11) (10.30) (-1.87) (-10.36) (-0.56) (-5.19) (-1.36) (-2.30) (2.82) Observations 3,697 3,697 3,697 3,697 3,697 3,697 3,697 3,697 3,697 R-squared Treatment elasticity, baseline Treatment elasticity, follow-up Control elasticity, baseline Control elasticity, follow-up Note: t-statistics in parentheses.

23 Table 5. Estimates of Program Impacts on Expenditures Shares, Triple Differences Panel B. Food groups (1) (2) (3) (4) (5) (6) (7) Cereal Tubers Meat & Fish Dairy Fruit & Veg Oth foods Eaten Out lnadjexp (β2) (-8.26) (-3.02) (6.09) (1.46) (-3.13) (-8.33) (3.25) Treat (β3) (0.43) (-1.63) (-1.26) (0.26) (-1.59) (-1.36) (1.30) Time (β4) (1.45) (-2.12) (-0.01) (-1.94) (-1.00) (-5.82) (0.99) Treat*Time (β5) (-0.73) (0.98) (1.57) (-1.78) (1.89) (2.11) (0.06) lnpce*treat (β6) (-0.36) (1.99) (1.15) (-0.28) (1.70) (1.30) (-1.29) lnpce*time (β7) (-0.28) (1.20) (-0.12) (1.89) (1.23) (5.53) (-0.81) lnpce*treat*time (β8) (0.64) (-1.30) (-1.37) (1.98) (-2.07) (-2.09) (-0.05) Constant (11.79) (5.40) (-4.30) (-1.02) (7.44) (11.51) (-2.94) Observations 3,697 3,697 3,697 3,697 3,697 3,697 3,697 R-squared Treatment elasticity, baseline Treatment elasticity, follow-up Control elasticity, baseline Control elasticity, follow-up Note: t-statistics in parentheses.

24 Table 6. Estimates of Program Impacts on Expenditures Shares, Triple Differences, by HH size Panel A. Broad household item groups lnpce*treat*time (β8) (1) (2) (3) (4) (5) (6) (7) (8) (9) Sample Food Domestic Health Transprt&Comm Housing Education Clothing Other Alc & Tobacco Pooled (-2.08) (-2.77) (2.76) (3.28) (2.69) (1.49) (-1.34) (-0.61) (-2.76) Small HH (-1.23) (-1.72) (2.93) (1.07) (2.48) (0.03) (-2.80) (0.49) (0.93) Large HH (-1.75) (-1.70) (1.06) (3.26) (1.58) (1.71) (-0.96) (-0.50) (-3.48) Panel B. Food groups lnpce*treat*time (β8) (1) (2) (3) (4) (5) (6) (7) Sample Cereal Tubers Meat & Fish Dairy Fruit & Veg Oth foods Eaten Out Pooled (0.64) (-1.30) (-1.37) (1.98) (-2.07) (-2.09) (-0.05) Small HH (1.34) (0.19) (-1.50) (0.44) (-1.37) (-1.83) (0.33) Large HH (-0.49) (-0.46) (-0.82) (1.94) (-0.94) (-1.48) (-0.50)

25 Conclusions As a result of the program, participating households have significantly higher expenditures than control households in four of the nine household item categories. We also see positive program impacts on spending in four of the seven food groups. The main contribution of this paper is our in-depth analysis of whether the program has shifted underlying preferences. As evidenced by the DDD results, the program appears to have impacts beyond simple income effects (movements along the Engel Curve) and has caused preferences to shift (a shifting of the Engel Curve) for some goods, including health and alcohol and tobacco

26 Next steps Investigate why the program induced such a large increase in health expenditures, especially over a period when health expenditures in control households were decreasing. May indicate more forward thinking. We can also compare savings (at follow-up only) between T and C households to verify the hypothesis that the program has induced a change in time preference data: module on time preference and discount rate.

27 References Deaton, A., & Muellbauer, J. (1980). Economics and Consumer Behavior. Cambridge: Cambridge University Press.

28 Thank you Contact information:

29 Extra slides

30 Proportion of HH with positive spending, broad HH groups (1) (2) (3) (4) (5) Treatment Control Wave 1 Wave 2 Wave 1 Wave 2 Proportion w/ Positive Spending Food Domestic Health, Hygiene Trnspt, Communication Housing Education Clothing Other Alcohol, Tobacco

31 Proportion of HH with positive spending, food groups (1) (2) (3) (4) (5) Treatment Control Wave 1 Wave 2 Wave 1 Wave 2 Proportion w/ Positive Spending Cereals Tubers Meat, Fish Dairy Fruit, Vegetables Other Food Food Eaten Out

32 Engel Curves

33 Engel Curves