1. 1 The Growth of Poor Children in China 1991-2000: Why Food Subsidies May Matter Lars Osberg Jiaping Shao Kuan Xu Economics Department Dalhousie University UNU-WIDER Conference Advancing Health Equity 29-30 September 2006 Helsinki, Finland

2. 2 The paper in one slide  China – 1991 – 2000 Rapid growth in inequality & average income Increased Average Height-for-Age children 2-13 Elimination of food subsidies after 1995 Little increase in height of disadvantaged kids  Does income poverty now matter more for child development in China? Robust OLS & quantile regressions on panel data Poverty status & growth in height-for-age  Not significant in 1991-1993  Significant negative impact 1997-2000  General moral: Food subsidization may play an important social protection role for child well-being and development

3. 3 China Health and Nutrition Survey (CHNS)  Micro-data collected in 1989, 1991, 1993, 1997 and 2000 Surveys conducted by the Carolina Population Center at the University of North Carolina Follows a panel of about 16,000 individuals in 4,400 households in 9 provinces  Incredible level of detail  1991 – 2000 data used  Children aged 2 to 13  <2 – variable height + measurement error  >13 – variability due to puberty onset documentation available at http://www.cpc.unc.edu/china

4. 4 Survey Provinces

5. 5 Context: Very Rapid Growth in GDP Per Capita & Inequality in China  GDP per capita in China 1980 =$763 2004 = $5,419  1991-2000 annual real growth rate of per capita GDP = 9.2 % 1991 = $ 1,721 2000 = $ 3,928  World Bank GDP per capita, PPP (constant 2000 international $)  2006 = 10.9% GDP growth  Gini index of inequality in money incomes China  0.345 in 1991  0.407 in 2000. Canada  0.281 to 0.302 USA  0.338 to 0.368 Netherlands  0.266 to 0.248 Switzerland  0.309 to 0.280

6. 6 Overall Income Distribution Shifts Up – Lower Tail Stays

7. 7 Issue: Well-Being & Growth ??  Rising Inequality in Cash Income in China + Money Matters More  Social Policy Changes Marketization Elimination of food subsidies Loss of employment guarantees Loss of health care coverage  Are the poor absolutely worse off? Meng, Gregory and Wang (2005)  Does poverty affect health of poor kids?

8. 8 HAZ i = (height child i – median height same sex & age) (standard deviation of height - same age & sex)  “good indicator underlying health status” slow height growth correlates with:  perform less well in school  score poorly on tests of cognitive function  poorer psychomotor skills & fine motor skills  lower activity levels  interact less frequently in their environments  fail to acquire skills at normal rates Mansuri (2006:3)  HAZ i < -2 = “stunted”  HAZ i < -3 = “severely malnourished”  Indicator of most basic of Sen’s “Capabilities” to attain normal physical & cognitive development

9. 9 Chinese Children are taller on average - Height as % of CDC Norm Age 2 -18

10. 10 Change of average height-for-age (HAZ) by decile 1991 to 2000 (age 2-13) DecileAverage HAZ 1991 Average HAZ 2000 Absolute Change Percentage Change 1 -3.39-3.220.164.8 % 2 -2.51-2.090.4216.8 % 3 -2.12-1.580.5325.1 % 4 -1.80-1.250.5530.7 % 5 -1.50-0.950.5536.8 % 6 -1.19-0.620.5647.5 % 7 -0.87-0.340.5360.6 % 8 -0.520.010.53101.3 % 9 -0.130.450.58439.8 % 10 0.801.300.5062.1 % N 27661735

11. 11 Stagnation in % children who are “Severely Malnourished” [ HAZ i < -3] ? HAZ i < -3HAZ i < -2 1991 7.23%28.49% 1993 6.27%24.86% 1997 3.94%19.71% 2000 4.27%16.43% [1991: n=1659; 1993: n=1390;1997:n=1123; 2000:n=813]

12. 12 Income Growth + Social Policy Reform  Rapid growth in GDP in China accompanied by structural reforms Employment security Housing Food subsidies  Relatively more important for low income households “Even though income growth reduces poverty, the radical reform measures implemented in the 1990s have sufficiently offset this gain that urban poverty is higher in 2000 than in 1986.”  Meng, Gregory and Wang (2005)

13. Social Protection in 1991 and 2000 19912000 Individual %1%1 Mean 2 %Mean health subsidy rec'd last year8.18%105.72 # Bath/haircut subsidy rec'd last year23.41%78.93 # Book/newspaper subsidy rec'd last year 12.60%85.74 # House subsidy rec'd last year2.02%116.14 # Other nonfood subsidy last year15.48%264.25 # Meat/veg/oil subsidy rec'd last year30.80%144.94 # Does person have health insurance?27.77%19.94% Annual insurance premium2.8777.55 Household Average value of ration coupon last year 422.77#

14. Value of Food Coupons, 1991 [$2 PPPpoverty line = 1072 Yuan; ½ Median = 597 Yuan ] Item % 1 Average annual amount 2 Average market value per coupon (Yuan) Total coupon value (Yuan) Rice86.34%690.320.43296.84 Wheat Flour71.28%512.50.32164 Other cereal grains27.15%334.90.2480.38 Cooking oil81.05%32.362.5381.87 Eggs2.67%25.981.6342.35 Pork (or other meat)9.59%67.042.93196.43 Chicken0.29%20n/a Sugar3.60%29.341.8353.69 Other4.30%1185.740.78924.88

15. 15 Does income poverty now matter more for child development in China?  Loss of food subsidy has removed buffer to impact of income shocks  China – special factors “One child” policy means family resources are concentrated Very rapid growth in most people’s incomes  HAZ i reflects accumulation of all past health & nutrition Including any long-ago food subsidies  ∆ HAZ i – is current growth influenced by current income shocks ?

16. 16 Variables1991-2000 Dependent variableHAZ_2000 HAZ_initial year0.472*** (0.132) Father weight (kg)_initial year0.012 (0.008) Father height (cm) _initial year0.018 (0.020) Father age at child’s birth0.013 (0.027) Father # of yrs formal education_initial year-0.053*** (0.018) Mother weight (kg) _initial year0.023** (0.011) Mother height (cm) _initial year0.029* (0.015) Mother age at child’s birth-0.026 (0.089) Mother age at child’s birth squared0.000 (0.001) Mother # of yrs formal education initial year0.063*** (0.017) Change of number of household members-0.101 (0.087) Log of total equivalent income in the period-0.130 (0.169) Dummy=1 if income <$2 / day_910.156 (0.129) Dummy=1 if income <$2 / day _930.410*** (0.128) Dummy=1 if income <$2 / day_97-0.603*** (0.148) Dummy=1 if income <$2 / day _2000-0.794*** (0.134) Panel OLS 1991-2000; 2-4 become 11-13 Poverty line=$2 / day ; very small sample

17. 17 Estimates from panel – Robust OLS – Early and late 1990s compared  Poverty line = $2 / day / per capita  Controls for tap water, region  Larger sample 2 to 11 in 1991 were 4 to 13 in 1993 2 to 10 in 1997 were 5 to 13 in 2000  Poverty status insignificant in 1991 & 1993 Statistically significant & negative in 1997 & 2000  Same results if ½ median poverty line Either national median or rural/urban separately

18. Variables1991-19931997-2000 Dependent variableHAZ_1993HAZ_2000 HAZ initial year0.711*** (0.042)0.537*** (0.041) Father weight (kg)0.001 (0.003)0.016*** (0.005) Father height(cm)0.010** (0.005)0.003 (0.007) Father age at child’s birth0.009 (0.009)0.005 (0.013) Father education0.015 (0.010)-0.005 (0.013) Mother weight(kg)0.007* (0.004)0.008* (0.004) Mother height (cm)0.003 (0.006)0.019*** (0.007) Mother age at childbirth0.013 (0.037)0.112** (0.048) Mother age squared0.000 (0.001)-0.002*** (0.001) Mother education-0.005 (0.009)0.020* (0.010) Change number in hhold0.023 (0.046)-0.033** (0.016) Log total equiv income0.027 (0.046)-0.087 (0.061) income <$2 / day_91-0.029 (0.066) income <$2 / day _930.069 (0.083) income <$2 / day_97-0.193** (0.098) income <$2 / day _2000-0.153* (0.082)

19. 19 OLS – presumes common impact of RHS variables on conditional mean  OLS – choose β to minimize sum squared residuals Implies: outliers acquire greatest weight Presumes: symmetric loss function for errors  Social important issue – stunting & deprivation Child growth & development – arguably a different process for the stunted & for the thriving  Quantile regression Estimates the differing determinants of outcomes at each point in the outcome distribution Linear, asymmetric loss function

20. 20 Panel 1991-93 – Robust OLS & Quantile Regression Estimated Impact of Poverty on HAZ i

21. 21 Panel 1997-2000 – Robust OLS & Quantile Regression Estimated Impact of Poverty on HAZ i

22. 22 Quantile Regression Estimates  1991 & 1993 quantile point estimates vary relatively little & close to OLS estimates - i.e. β = 0 nothing gained by using quantile regression  1997 fairly close to OLS β = -0.2 (≠0 @ 95%)  2000 Increasingly large negative impact of poverty for larger HAZ  Poverty especially reduces the chances of substantial growth

23. 23 Caveats  China = 1.3 Billion people HUGE diversity of circumstances  CHNS data – very fine BUT Great Audacity would be required to generalize from 4,400 households in 9 provinces Sample selection implies much smaller sample size for regressions reported here  N = 1278, 587, 129  Simplest poverty measure used + 3 alternative poverty lines  Suggestive results NOT conclusive evidence – to be corroborated with larger data sets ??

24. 24 Possible Implications  Bottom decile of children aged 2-13 in China ≈ 13 million Failure to keep up with increase in stature of other children matters 1: For future output, health & well-being 2: Because children are citizens NOW  General Moral? Subsidies to basic food availability may play a crucial safety net role for child well-being  Historically important in OECD nations  Worth considering in Social Policy Reform in LDCs