﻿

# DISTRIBUTIONAL EFFECTS OF GOVERNMENT POVERTY-RELATED EXPENDITURE

К оглавлению

Having considered the allocation of general categories of government

expenditure we now focus on our main concern: the impact of those programs

targeted at the poor. Given a lack of information on intra-provincial

allocations our focus is on the allocation between provinces and the extent

to which poorer provinces are favored or not. In particular, we ask whether

the level of expenditure per person in a given province is related to:

(a) the level of poverty incidence (P) in that province;

(b) the level of household income per household member (Y) in that

province;

(c) the size of the rural population relative to the size of the total population

(nR) in that province; and

(d) the overall size of the province, as measured by its total population

(N).

The statistical analysis which follows is based on the provincial allocation

of the cumulative total of expenditures for the years 2000 to 2002, as

summarized in Table 5.11 above. Four models are estimated:

Ei

j = i0

+ i1

ln Pj + 2 ln Nj (1)

Ei

j = i0

+ i1

ln Yj + i2

ln Nj (2)

Ei

j = i0

+ i1

ln Pj + i2

ln Nj + i3

nRj

(3)

Ei

j = i0

+ i1

ln Yj + i2

ln Nj + i3

nRj

(4)

where Ei

j denotes expenditure per person of type i in province j, Yj denotes

average income per person in province j in the year 2000, Pj denotes the

offi cial estimate of poverty incidence (percentage of the population with

incomes per head below the offi cial poverty line) in province j in the year

2000, Nj denotes total population of province j in the year 2000, and nRj

denotes the share of the population of province j residing in rural areas in

the year 2000.

Models 3 and 4 were discarded because the rural share of the total

provincial population (nR

j ) was strongly correlated with poverty incidence

at the provincial level. As noted above, poverty in Thailand is a strongly rural

phenomenon. The inclusion of both poverty incidence and rural population

share thus introduces strong multicollinearity into the regressions. The

subsequent discussion will therefore focus on models 1 and 2, which exclude

the variable nRj

.

The regression results are summarized in Tables 5.17 to 5.23. First, Table

5.17 indicates that the total ‘poverty-related’ budget is positively related

to poverty incidence and negatively related to income per person – poorer

provinces receive larger budgets per person, but that these statistical

relationships are insignifi cant. The signifi cant explanatory variable is the

size of the province. Large provinces receive smaller allocations per person.

In terms of the allocation of ‘poverty-related’ expenditures, it pays to be

small. It makes little difference whether the province is rich or poor.

Tables 5.18 to 5.23 now perform a similar exercise for each of the six

components of ‘poverty related’ expenditures summarized in Table 5.11.

• Poor and low-income people

• Infrastructure

• Agriculture and natural resources

• Health and social welfare

• Education and training

• Others

The coeffi cient on poverty is positive in four of the six components and

negative in two (Agriculture and natural resources, and Health and social

welfare) but the relationships are far from being statistically signifi cant.

Poverty incidence has little to do with the allocation of these expenditures

across provinces. Income per capita similarly has little relationship to the

allocation of expenditures in all categories except ‘Poor and low-income

people’, where the expected negative coeffi cient is observed. For this category

at least, corresponding to just under 6 per cent of the total ‘poverty-related’

expenditures reported in Table 5.11, expenditures are seemingly poverty

targeted.5

The Poor and Low-income People Program had three component

categories in 2000. These were:

• Assistance to farmers and the poor

• Health care loan for low-income people

• Health care for low-income people

The first category is assistance to farmers – supposedly poor farmers – who

rent land. In some documents this category is called ‘Agricultural land rent

control for farmer and assistance to farmer and the poor’. Expenditure in

this sub-category is positively related to provincial income per capita in

each year 2000 to 2002, but the relationship is not statistically signifi cant

in any one of these years. The second category is loans for health care. It

was negatively related to income per person in 2000 and 2001 (coeffi cients

marginally insignifi cant at the 90 per cent level), but was discontinued in

2002. The third category is grants for health care and it was negatively

and signifi cantly related to income per person in 2000 and 2001 (at the

95 per cent level), but it was also discontinued in 2002. That is, there

appears to have been a lessening of the poverty-targeting feature of these

expenditures in 2002.

Table 5.17 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: total budget

allocation

Intercept lnP lnN R2

16372.910 95.008 –929.764 0.242

(6.050*) (0.871) (–4.623*)

Intercept lnY lnN R2

18215.220 –106.248 –989.278 0.256

(4.904*) (–0.316) (–4.997*)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations using data from Bureau of the Budget, Bangkok.

Table 5.18 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: poor and low-income

people program

Intercept lnP lnN R2

277.368 11.868 –0.860 0.025

(1.249) (1.325) (–0.052)

Intercept lnY lnN R2

797.152 –61.206 –1.789 0.066

(2.681*) (–2.274*) (–0.113)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

Table 5.19 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: infrastructure

program

Intercept lnP lnN R2

4621.135 48.768 –266.062 0.101

(3.397*) (0.889) (–2.632*)

Intercept lnY lnN R2

5911.286 –167.805 –256.272 0.098

(3.200*) (–1.004) (–2.603*)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

Table 5.20 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: agriculture and

natural resources program

Intercept lnP lnN R2

748.956 –7.558 –32.593 0.034

(2.503*) (–0.626) (–1.466)

Intercept lnY lnN R2

511.407 27.167 –32.520 0.036

(1.243) –0.730 (–1.483)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

Table 5.21 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: health and social

welfare program

Intercept lnP lnN R2

1476.924 –3.503 –64.324 0.022

(2.121*) (–0.125) (–1.243)

Intercept lnY lnN R2

1071.784 62.793 –71.742 0.034

(1.141) (0.739) (–1.433)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

Table 5.22 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: education and

training program

Intercept lnP lnN R2

667.284 13.281 –38.145 0.038

(1.795) (0.886) (–1.381)

Intercept lnY lnN R2

1010.139 –33.265 –41.814 0.040

(2.020*) (–0.735) (–1.569)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

Table 5.23 Thailand: regression results on relationship between povertyrelated

expenditure, poverty and income: other programs

Intercept lnP lnN R2

8581.245 32.150 –527.780 0.485

(9.719*) (0.903) (–8.043*)

Intercept lnY lnN R2

8913.450 66.067 –585.141 0.465

(6.455*) (0.529) (–7.950*)

Note: t values are in brackets and * indicates statistically signifi cant at 95 per cent level.

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

To test this proposition, data are assembled in Table 5.24 based on the

results of regression analyses for all project components of the expenditures

summarized in Tables 5.10 and 5.11. This was done separately for each year

2000, 2001 and 2002. The regressions performed were the same as Model 2

above, except that the population size variable (Nj) was deleted. For each of

these years, the projects were then classifi ed into three categories:

• Pro-poor, meaning that expenditure was negatively related to income

per person, signifi cant at 90 per cent level or better.

• Neutral, meaning that expenditure was not related to income per

person, at 90 per cent level or better.

• Pro-rich, meaning that expenditure was positively related to income

per person, signifi cant at 90 per cent level or better.6

The expenditures corresponding to each project, so classifi ed, were then

added and the results are summarized in Table 5.24.

Table 5.24 Thailand: poverty reduction expenditures classifi ed by relationship

to income per person, 2000 to 2002

Type of Project 2000 2001 2002

Pro–poor 25 664.33 38 245.60 20 964.64

(signifi cant 90%)

Neutral 21 707.07 17 349.39 15 432.76

(insignifi cant)

Pro–rich 26 365.57 24 692.97 41 127.00

(signifi cant 90%)

Total 73 736.97 80 287.96 77 524.40

Source: Authors’ calculations, using data from Bureau of the Budget, Bangkok.

In 2000 these three categories were of similar size. The net effect of the

‘poverty-related’ expenditures was approximately neutral between poor

and non-poor provinces. However, the expenditures corresponding to the

pro-poor and neutral categories have contracted over the three years, while

those corresponding to the pro-rich category have expanded, especially in

2002. By 2002 the ‘poverty-related’ expenditures were, on balance, positively

related to provincial incomes per person – richer provinces received greater

benefi ts per person.