Development Economics: Macroeconomics

Size: px

Start display at page:

Download "Development Economics: Macroeconomics"

Miles Leonard
5 years ago
Views:

1 MIT OpenCourseWare Development Economics: Macroeconomics Spring 2009 For information about citing these materials or our Terms of Use, visit:

2 Wealth Accumulation and Factors Accounting for Success Anan Pawasutipaisit and Robert M. Townsend April 2, 2009

3 Motivation "It is hard to think of economic issues that are more important than the accumulation of capital..." Angus Deaton, 1992 Net worth or wealth determines how well off a household is. Using accounting framework to measure wealth. Measurement and fact finding rather than theory. The process of wealth creation is related to productivity Some succeed and some do not. Inequality and heterogeneity are related.

4 Data Townsend s Thai monthly survey. 531 households, Jan 1999-Dec Observations Chachoengsao 141 Lopburi 147 Buriram 104 Sisaket 139

5 Wealth distribution in the survey Top 1% households own about 1/3 of total wealth. But not true everywhere, there is variation across location. Wealth inequality is highest in Chachoengsao but lower in other provinces. The top 1% in Lopburi own less than 13% while the top 1% in Sisaket own less than 18%. Table 1. % of networth held by various groups defined by percentiles of the wealth distribution Year Percentile

6 Poor is getting more. The least wealthiest group at the beginning seems to be the one that rises most rapidly. Share of top 10% is going down. Table 2. % of networth held by various groups defined by percentiles of the wealth distribution in 1999 Year Percentile

7 Distribution of wealth within group Wealth distribution of the poorest groups (1&2) shift toward the right but has a wider support. Mean of the distribution increase for the poorest group.

8 On average, total growth rate is 2.5% per year. There is variation by location. Table 3. Annual Percentage Increase in Net Worth Year Chachoengsao Buriram Lopburi Sisaket Total

9 Overall picture from the survey data Wealth is growing, although there is variety in growth rate by location. Over time, every group becomes better off in general and there is less inequality between groups. But not every household is better off.

10 Decompositionofgrowthofnet worth Household level : let W i,s i and G i be net worth, saving (save or dissave) and gifts t t t (receive or give) at time t of household i. W t i = W t i 1 + S t i + G i t = W i 0 + tx ³ j=1 S j i + G i j Aggregate level : average growth rate is wealth weighted sum of growth from subgroup g = 1,...,G. ³ ³ i X g N g i=1 t W i 0 N g i=1 W i 0 1 P = p N 1 P N g 1 P N N i=1 W i 0 g=1 N g i=1 W i 0 N i=1 W i 0 1 P N W i G 1 P N g W i 1 P N g N i=1 t W 0 Andgrowthofsubgroupcan be decomposed to saving andgiftcomponent ³ 1 P N g W i i 1 P N g P t 1 P N g P t N g i=1 j=1 S j i N g i=1 j=1 Gi j 1 P N g = 1 P N g + 1 P N g N g i=1 W i 0 N g i=1 W i 0 N g i=1 W i 0 N g i=1 t W 0

11 Which component is a significant part of wealth accumulation? 82% of growth is accounted by saving component Growth of wealth of poorest group is 175%, but their wealth weight is only 6%. Overall growth is 18% over 7 years (about 2.5% APR). Growth rate decreases in initial wealth. Role of gifts is small for wealthiest but mix at the low end. Table 4. Growth of networth and its components held by various groups defined by by percentiles of the initial wealth distribution P N g Percentile Growth of Wealth N 1 g PN i=1 W 0 1 N i i=1 W i 0 Saving Gifts Total

12 The process of wealth formation is not smooth, look at monthly data Some groups experience negative growth at some points but all groups manage to grow over 7 years period. N g N g 1 X 1 X ³ W i t = S i t + G i t N g i=1 N g i=1

13 Decomposition of inequality of net worth (level and change) j Let W OL be wealth of household j with occupation O at location L, and N is total number of household Decomposing by location and occupation (definition and algebra) : 1 X j 1 X ³ I = log W OL log W N N OL j,o,l j,o,l 1 X j X N L = log N W OL j,o,l L j N log (W L) + X N L log (W L ) X N OL log (WOL ) N L X N L X N OL L N O N L O + I OL = AI + AOW C + WOWC N L

14 where ³ ³ 1 P j P N L AI = log N j,o,l W OL L N log (W L) is inequality across location P ³ N L P N OL L O AOW C = N log (W L ) N log (W L OL ) is (sum of) inequality across occupation within location WOWC = P L N ³ P L O N OL N N I OL is (sum of) inequality within occupation within L location (residual)

15 Increasing in wealth inequality across location (AI) until 2002 and decreasing after that. Smallest component among the three. Both AOWC and WOWC components have decreasing trend, where residual (WOWC) is the biggest component among the three. By region : wealth differential according to occupation (AOWC) in the NE is going up but going down for the central

16 In term of change, I = AI + AOW C + WOWC Inequality within occupations and location is going down and that is the largest single force most of the time. Inequality across locations can be large sometimes though typically moving in the opposite direction and eventually switching signs.

17 Income differential across activities There are 5 primary sources of income for households in the survey. Most households are labor households, and cultivation households. Table 7. Number of households in each occupation Primary Occupation Chachoengsao Buriram Cultivation Livestock 3 0 Fish/Shrimp 28 0 Lopburi Sisaket Business Labor Total

18 Almost every occupation performs better than cultivation. There is convergence in income Table 8. Ratio of mean income from non-cultivation to cultivation occupation Year Mean Livestock Fish/Shrimp Business Labor

19 Consumption Distribution in the Survey Unlike wealth distribution, consumption is distributed less unequally. Consumption share of the top 1% households are around 4-6%, which are a lot lower than 33% as in wealth share of the same group. Bottom 50% households consume about 1/3 of total consumption although their wealth shareislessthan10%. Table 9. % of consumption held by various groups defined by percentiles of the wealth distribution Year Percentile

20 If we partition households into 4 groups with equal size. Consumption share is increasing in wealth and it fluctuates over time. By Theil-L, consumption inequality fluctuate with overall increasing trend. Table 10. % of consumption held by various groups defined by percentiles of the wealth distribution in 1999 Year Percentile

21 Income Distribution in the Survey Income distribution is also less unequal than wealth distribution. Observe negative income share at some point (the top 1% in 2004). Half of the households who are relatively poor (by wealth) produce about 30% of total income. Table 11. % of income held by various groups defined by percentiles of the wealth distribution Year Percentile

22 Partition households into 4 groups with equal size. Income share is also increasing in wealth, the top 25% sometimes own more than half of income produced but it fluctuates over time. By Theil-L, income inequality fluctuates with overall decreasing trend Table 12. % of income held by various groups defined by percentiles of the wealth distribution in 1999 Year Percentile

$By occupation, the fraction of poor households in business is declining$

23 Poverty Reduction and Occupation Poverty line by 1) Thai official income number for each province 2) $2.16 a day in 1993 PPP. By income, clear cycle in the NE. There are many more poor in the Northeast, but the biggest gains are in Buriram. By occupation, the fraction of poor households in business is declining substantially.

By location, the biggest gains are still in Buriram.

24 By consumption, there are a lower number of poor than using income measure, especially in the Central. By location, the biggest gains are still in Buriram. By occupation, labor and business escape poverty relatively fast.

25 HeterogeneityinGrowthofNet Worth Computing growth per month and take arithmetic mean overall period for each household. Positive growth on average. But 40% of households have negative growth rate. Distribution is much wider in the NE. Table 15. Distribution of Growth of Net Worth (% per month) min max mean std. dev. Q1 Q2 Q3 Chachoengsao Lopburi Buriram Sisaket All

Example of some households who experience large increases and decreases in their

26 Taking the advantage of the long monthly panel, we can track relative position of net worth within changwat for each household. Example of some households who experience large increases and decreases in their relative position. Histogram of change in relative position over 7 years, center is at zero but (std. dev. =13.75, min=-57, max=80).

27 Decomposing change in net worth to savings and gift NW i t = S i t + G i ³ t for all i ³ cov NW i,s i cov NW i,g i 1 = ³ + ³ for all i var NW i var NW i Variation in wealth change for most households is better explained by variation in savings rather than gifts, as distribution has center around 100.

28 Growth of Net Worth, Productivity and Savings Rate Saving can be think of a combination of savings rate, productivity and assets level St i = Yt i Ct i S i π i = t t A i π i A i t ³ t t = s i t ROA i t A i t This interpretation is suitable for households that use assets to generate income. But it is harder to interpret for households that have labor earning as primary source of income. When we look at ROA, look only at non-labor households.

29 Household saving rate Excluding negative net income, and small positive net income. 3 types of observations: household-month as data is originally collected, average to household-year, and household (average over 7 years). Stronger positive association of growth of net worth and savings rate over the long run as correlations tend to be higher when we aggregate over calendar year, and over 7 years. By location there is significant and positive correlation between growth of net worth and savings rate in Chachoengsao at almost all levels for both non-labor and labor households. There are more significant correlations for non-labor than labor households.

30 Table 16. Correlation of growth of net worth and savings rate for nonlabor househo All Chachoengsao Buriram Lopburi Sisaket HH-month (0.7475) (0.0876) (0.1191) (0.7507) (0.8263) HH-year (0.0025) (0.0118) (0.0071) (0.0002) (0.2019) HH (0.0005) (0.0017) (0.2472) (0.0000) (0.1568) Table 17. Correlation of growth of net worth and savings rate for labor households All Chachoengsao Buriram Lopburi Sisaket HH-month (0.8229) (0.2738) (0.8421) (0.5515) (0.9717) HH-year (0.0462) (0.0009) (0.2669) (0.0166) (0.4314) HH (0.0107) (0.0000) (0.2099) (0.5350) (0.3089)

31 Return on assets (ROA) A standard accounting concept for financial analysts, but here is for household. There is significant and positive correlation between growth of net worth and ROA at all levels. Correlation tends to be higher when we average overall 7 years. Table 18. Correlation of growth of net worth and ROA for nonlabor households All Chachoengsao Buriram Lopburi Sisaket HH-month (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) HH-year (0.0000) (0.0000) (0.0000) (0.0000) (0.0000) HH (0.0000) (0.0000) (0.0000) (0.0000) (0.0000)

32 If we plot density of growth of net worth by ROA. The density shifts toward the right as ROA is in higher group. Households with low ROA tend to have lower growth of net worth, but there are few exception. Higher ROA households tend to have more dispersion in term of growth.

33 Factors Accounting for Success Education Most household s head (80% ) have primary level education, but the younger the head, the higher the education. Table 19. Head s education at baseline No edu. Primary Secondary Vocational Univ. Chachoengsao Buriram Lopuri Sisaket Total Table 20. OLS estimate of head s education on head sage Chachoengsao Buriram Lopuri Sisaket Total (0.003) (0.027) (0.002) (0.001) (0.000)

Using all household members, education attainment varies across location, occupation and over time. Education level is higher in the central than the NE.

34 Using all household members, education attainment varies across location, occupation and over time. Education level is higher in the central than the NE. Cultivation households have lowest level of education while livestock, fish/shrimp and business households have higher education. The average level of education of labor households is also low but the maximum one (16 years) are all at labor households. Education attainment increases over time, younger members stay in school longer than previous generation.

35 Measures for household education Consider all members in that month and use the mean or median as representative of that month, or max (education of existing members who have the highest level of education). These measures can vary over time depending on educational attainment of members and household composition. High education is associated with high ROA. By occupation, business households have positive and significant regression coefficient for all measures. Table 22. Regression coefficient of ROA on education, by occupation (household-y e Cultivation Livestock Fish/Shrimp Business All max_edu (0.312) (0.899) (0.055) (0.015) (0.000) mean_edu (0.333) (0.427) (0.403) (0.001) (0.000) median_edu (0.393) (0.159) (0.624) (0.002) (0.000)

36 Multivariate regression Regressors : education, debt normalized by assets, initial wealth, occupation (dummy), household size, head age and control for time and location. There are four specifications where dependent variable is ROA, and regressors are with and without labor supply, with and without household fixed effect (HHFE) High ROAisassociatedwithhighdebt/assetratio andhousehold size Households with lower initial wealth, younger head tend to have higher ROA. Education is not robust in all specifications, either negative and significant or positive but not significant. Own work is positive and significant both with and without HHFE. Size of estimate of own labor supply is also bigger than paid labor supply.

37 The explanatory power increases several times when we control for household fixed effect, indicating that factors accounting for success is also specific to each household. Table 23. Multivariate regression Dependent variable : ROA Debt-Asset Ratio Education (0.000) (0.054) (0.000) (0.060) (0.000) (0.134) (0.011) (0.183) Household Size Head Age (0.005) (0.000) -1.14e (0.002) (0.003) (0.000) -1.18e (0.002) Initial Wealth (0.000) (0.000) Own Hour Work (0.000) (0.040) Paid Hour Work (0.002) (0.611) Household Fixed Effect No Yes No Yes Number Obs. 2,070 2,077 2,070 2,077 adj-r

38 Persistent of ROA ROA regression indicates that a factor specific to household can considerably account for variation in success. This factor may not change over time or change slowly. Whether a high ROA household today is more or less likely to have high ROA again in the future? We compute average ROA for the first half and second half of the overall sample frame and rank them for each time period (from lowest to highest).

Households successful over the first half of the sample are

39 There is considerable persistence, especially for households in the three provinces except Buriram. Households successful over the first half of the sample are likely successful over the second, or luck per se is not an explanation for success.

40 In Buriram, persistence is much lower (correlation is 0.07 and not statistically significant). One possibility to explain less persistence in Buriram is that household composition is less stable there. A household in which most of the individuals change should not have persistent ROA over the two subperiods. Regression : variation in household size can undercut persistence of ROA. ³ ROA 2,i = b 0 + b 1 ROA 1,i + b 2 ROA 1,i sd(hhsize i ) + b 3 sd(hhsize i )+u i Table 24. Estimate from regression ROA 1 ROA 1 sd(hhsize) sd(hhsize) adj-r 2 Number Obs (0.000 ) (0.024) (0.542 )

Predictive power of ROA Classify households into 3 groups by ROA: high, middle, and low Whether there is any difference in financial and physical assets

41 Predictive power of ROA Classify households into 3 groups by ROA: high, middle, and low Whether there is any difference in financial and physical assets accumulation, and consumption behavior? Financial assets are growing for almost all groups and regions (exception is for the low group of Buriram and Sisaket).

42 Physical assets of high ROA group fluctuates with an increasing trend. Physical assets of middle and low ROA groups fluctuate with decreasing trend. One interpretation could be that high ROA households put their wealth back to their income generating activities to get a high return.

Consumption : using mean consumption of the first half and the second half to compute its growth rate. In this way, it does not appear that households with higher ROA have higher consumption growth.

43 Consumption : using mean consumption of the first half and the second half to compute its growth rate. In this way, it does not appear that households with higher ROA have higher consumption growth. And correlation of consumption growth and ROA are not statistically significant. If household is in autarky : consumption Euler equation should be n u 0 (c t )=βe t u 0 (c t+1 )(1 δ + MPK) o If ROA is related to MPK, then consumption growth should be related to ROA.

44 Financial Stretegies and Potential Imperfection Financing cash flow deficit where D t = C t + I t Y t F 1 t,...,f n t D t = F t F t n be overall deficit, it must be financed by devices ³ cov D, F 1 cov (D, F n ) 1= var (D) var (D) High ROA households are not using capital assets to smooth consumption and conversely use consumption to finance investment deficits. High ROA households are more actively involved in financial markets in the sense of using formal savings accounts and borrowing, and less informal, i.e., less gifts. But high ROA households do use cash more than the low ROA households.

45 Potential Imperfection in Credit Markets If we impose the Cobb-Douglas functional form β K β L yjt = A jt K jt L jt where y jt,k jt,l jt are net income from productioin activities (cultivation, livestock, fish/shrimp, business), level of physical assets, labor supply of household j at time t n o A jt =exp A 0 + α j + u jt where A 0 is common productivity or mean efficiency across households, α j is time-invariant household specific fixed effect for household j, and u jt is error term for household j at time t. Estimating OLS of the log-form of above equation using household-year (average by mean) Table 27. Estimate from regression β K β L adj-r 2 Number Obs ,762 (0.021) (0.000)

Computing marginal product of capital (MPK) for each household and compare to the interest rate. The median of monthly interest rate 0.75% (or 9% per year).

46 Computing marginal product of capital (MPK) for each household and compare to the interest rate. The median of monthly interest rate 0.75% (or 9% per year). Scatter plot of average MPK against (median) interest rate. There is variation in MPK, high capital households are associated with low MPK : potential imperfection in credit markets.

More productive households have higher debt-asset ratio than the other group, but there are

47 Whether households with relatively high MPK borrow more than the other group? Group by whether MPK is above or below interest rate and look at debt-asset ratio. More productive households have higher debt-asset ratio than the other group, but there are some outliers. Overall it seems that the financial system is working to some extent, but the system is not working perfectly.

48 Using estimate of A j as TFP for household j and see how that is related to ROA and growth of net worth. TFP is positively but weakly related to ROA and growth of net worth. Table 28. Correlations of TFP, ROA, Growth of Wealth TFP ROA ROA (0.0080) Growth of Wealth (0.0074) (0.0000)

49 Story from selected case study household ROA regression indicates that a factor specific to household can considerably account for variation in success. We look at some successful households and see what makes them successful.

Lending Services of Local Financial Institutions in Semi-Urban and Rural Thailand

Lending Services of Local Financial Institutions in Semi-Urban and Rural Thailand Robert Townsend Principal Investigator Joe Kaboski Research Associate June 1999 This report summarizes the lending services