DISTRIBUTIONAL EFFECTS OF GENERAL GOVERNMENT EXPENDITURE

К оглавлению
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 
17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 
34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 
51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 
68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 
85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 
102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 
119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 
136 137 138 139 140 141 142 

Before turning to the examination of government expenditures directed

towards the poor, it is helpful to review the distributional effects of

government expenditures as a whole. Studies of the distributional effects

of government activity in Thailand have focused heavily on the tax system.

Many studies have addressed the question of tax incidence, including

Medhi (1976), Pichit (1985), Chalongphob, Pranee and Tienchai (1988)

and Chalongphob and Direk (1999). The general conclusion has been that

the tax system is roughly distributionally neutral, with the exception of the

personal income tax and the corporate income tax, which are progressive,

but which represent only a small part of government revenue. Most studies

have recommended that the proportion of revenue collected from these two

taxes be increased. It has not happened.

Surprisingly little attention has been given to identifying the manner

in which government expenditures in Thailand generate differential

benefi ts among households of varying incomes. Only one study appears

to have addressed the issue systematically. Chalongphob and Direk (1999)

analyze the distributional effects of the major components of government

expenditure, including education, public health and basic infrastructure,

divided in each case into operational expenditures, investment expenditures

and loans. Although the methodology is open to criticism, the strength of

the study is that it attempts to do this within a comprehensive framework,

which permits direct comparison with its results on the tax system,

summarized above.3

Table 5.12 summarizes the results on expenditure benefi ts by taking the

estimated expenditure benefi ts by income decile and computing simple linear

regressions which relate the total benefi t from public expenditure received by

each income group (the dependent variable) to its income (the independent

variable), with the results classifi ed by type of expenditure. The fi nal column

of the table presents these results in terms of elasticities, evaluated at the

mean of the distribution. An elasticity of exactly one would mean that as

incomes increase, the percentage increase in total benefi t received is the same

as the percentage increase in income. An elasticity which is positive but

less (greater) than one, means that as income increases the benefi t received

increases by a proportion smaller (larger) than the proportional increase

in income.

The final column, reporting elasticities of expenditure benefi t with

respect to income, indicates that as incomes increase, the benefi t received

also increases, but with an elasticity of 0.4. A 10 per cent increase in income

corresponds to a 4 per cent increase in the total benefi t received from

public expenditure, overall. For transportation expenditure, this elasticity

is 1.1, indicating that benefi ts received by richer groups increase even more

rapidly with rising incomes than do incomes themselves. For education

expenditures the elasticity is 0.32 and signifi cantly different from zero. For

health care it is only 0.02 (but still signifi cantly different from zero) and

for agriculture it is –0.09 (and again signifi cantly different from zero). In

interpreting these results the methodological shortcomings of the study

need to be borne in mind.

Table 5.12 Thailand: results of regressions of expenditure benefi t against

income

Types of

expenditures Intercept t-statistic Coeffi cient t-statistic R-squared Elasticity

Total 18577.500 34.963 0.048 36.648 0.994 0.412

Education 7850.853 17.514 0.014 12.741 0.953 0.327

Health care 4833.409 63.110 0.000 2.135 0.363 0.022

Agriculture 6706.751 25.157 –0.002 –2.980 0.526 –0.086

Transportation –813.468 –3.743 0.035 66.222 0.998 1.092

Source: Calculations by authors based on data in Chalongphob and Direk (1999).

The geographical incidence of public expenditures is very relevant when

evaluating their distribution and poverty impact. Regional disparities in

incomes per capita are very signifi cant, as are the disparities in poverty

incidence (see Table 5.3 above). Bangkok, central, and east regions have

per capita incomes well above the average for the country, with Bangkok

over three times the national average. The other four regions are sharply

below the average, with mean incomes per capita in the northeast and north

being one tenth and one seventh that of Bangkok, respectively. These data

are consistent with fi ndings from other sources that the lowest income and

highest poverty levels in Thailand are to be found in the northeast, north

and south. Other sources show that these regions are relatively deprived of

economic and social infrastructure. Moreover, while the Crisis of 1997–98

had a severe adverse impact on all regions, the poorer regions seem to have

suffered the most, in terms of the size of the decline in their real incomes

and the increase in their unemployment rates.

The Comptroller General’s Offi ce accounting system allows the breakdown

of central government expenditure by location of the offi ce from which the

expenditures took place; some 85 to 90 per cent of central government

expenditure can be broken down in this fashion. The major problem with

such a breakdown is that government debt service and the salary payments

of many ministries and agencies are distributed centrally, not locally and

hence are attributed to the Bangkok region. To circumvent these problems,

these items are deleted from the data set used in the analysis that follows.

An important study undertaken by a British consulting firm (Mokoro,

1999) reviewed Thailand's public expenditure system. The Mokoro report

deals only with the expenditure system, but covers a wide range of issues

relating to expenditures, of which the distributional effect is only one. The

Mokoro report covers the role of public expenditure policy in macroeconomic

stabilization and national resource allocation and contains detailed

descriptions of the expenditure programs in education, health, agriculture

and road-building. The study also reviews expenditures related to poverty

incidence. The data assembled in the report include a new data set on the

distribution of government expenditures by region, of which there are seven,

and by province, of which there are 77, including the Bangkok metropolis.

The full distributional effects of government expenditure policy may

be thought of at two levels. First, there is the distribution of expenditures

between provinces, and second there is the distribution among households

within provinces. If the distribution of incomes between provinces was

relatively equal, the fi rst of these distributional issues would be of minor

signifi cance. In Thailand, this is far from the case. The distributional effects

of government expenditure policy among households within a province are

unquestionably important, but extremely diffi cult to study using available

data. The distribution among provinces of widely varying incomes is also

important, but much more tractable using available information. We shall

investigate this issue below, but it must be kept in mind that the question

being asked: how the distribution of expenditures per person varies among

provinces according to their incomes per person, is at most only one

component of the broader question of how the distribution of expenditures

varies among households of varying incomes.

If expenditures were allocated in a manner that at a national level favored

poorer households relative to richer households, then, in the context where

intra-regional inequality is a signifi cant component of total inequality, we

would expect to fi nd that poorer provinces would also be favored relative to

richer provinces. If, on the other hand, it were found that the allocation of

expenditures per person favored richer over poorer provinces, this contrary

fi nding would be partial (but not necessarily conclusive) evidence that, at a

national level, the system of expenditures did not favor the poor.

To investigate these issues, average annual expenditures per person from

fi scal years 1997 to 1999 (dependent variables) have been regressed on

average incomes per person (independent variable). The latter variable is

available for 1996. Table 5.13 reports the regression coeffi cient between

these two variables and its t-statistic, the latter shown below the estimated

coeffi cient to which it refers.4

Table 5.13 Thailand: distribution of government expenditures by province,

1997 to 1999, total, current and capital

Bangkok Bangkok

included excluded

Part 1: Independent Mean provincial incomes

variable: per capita, 1996

Dependent variable

Total expenditure coeffi cient 64.18 43.43

t-value 3.52 2.27

Current expenditure coeffi cient 34.68 17.35

t-value 2.74 1.34

Capital expenditure coeffi cient 29.51 26.08

t-value 4.64 3.77

Part 2: Independent Poverty incidence

variable: (headcount) 1996

Dependent variable

Total expenditure coeffi cient –40.17 0.61

t-value –0.51 0.008

Current expenditure coeffi cient –25.17 4.44

t-value –0.47 0.09

Capital expenditure coeffi cient –15.00 –3.83

t-value –0.52 –0.14

Part 3: Independent Rural population

variable: share 1996

Dependent variable

Total expenditure coeffi cient –184.86 –58.24

t-value –2.72 –0.72

Current expenditure coeffi cient –108.39 –3.27

t-value –2.34 –0.06

Capital expenditure coeffi cient –76.47 –54.97

t-value –3.11 –1.81

Source: Calculations by authors using data from Mokoro (1999), with supplementary data

from the Comptroller General’s Offi ce, Bangkok.

The section of Table 5.13 labeled ‘Part 2’ shows a similar set of

calculations, but with poverty incidence at the provincial level used as

the independent variable in place of the income variable. In this table, if

higher levels of expenditure per person were associated with higher levels

of poverty incidence, a positive coeffi cient would be observed. A negative

coeffi cient would indicate that high levels of poverty incidence in a province

were associated with lower levels of expenditure per person. In ‘Part 3’ the

independent variable is the rural population share of the province.

The results indicate that total expenditures per person by province are

signifi cantly related to provincial income per person, and the relationship

is positive: provinces with higher incomes per person receive higher

expenditures per person. The relationship continues to apply when Bangkok

is excluded from the data set, but not quite as strongly. It is possible to

conduct these analyses separately for different public expenditure categories;

education spending, health spending and so forth. The results, in Table

5.14, indicate that both education and health spending by province were

positively related to income per person in both 1997 and 1998, but this

relationship is statistically signifi cant only for 1997. The data are weakly

indicative that adjustments to education spending from fi scal years 1997 to

1998 were negatively related to provincial incomes, but the reverse applied

to adjustments to health spending.

When poverty incidence is substituted as the explanatory variable (Part

2 of these tables), the explanatory power declines markedly. Surprisingly,

poverty incidence by province, as measured offi cially, is weakly correlated

with provincial income. Similarly, when rural population share is used as

the explanatory variable, the explanatory power is only slightly lower than

income. Expenditures per person are higher in richer provinces, which implies

that they are higher in urban-dominated provinces than rural provinces.

Tables 5.15 and 5.16 convert these results into elasticity format. First,

these tables repeat the estimated coeffi cients from Part 1 of Tables 5.13

and 5.14, respectively. These coeffi cients, labeled ‘estimated coeffi cient’

can be interpreted as the estimated change in expenditure by region

resulting from a unit increase in provincial income. The tables then convert

this marginal effect into an elasticity, which may be interpreted as the

estimated proportional change in expenditure by region resulting from

a unit proportional increase in provincial income. These elasticities have

the convenient properties described above. An elasticity between zero and

unity means that as provincial incomes (per person) increase, provincial

expenditures (per person) also increase, but in a smaller proportion.

From Table 5.15, the estimated elasticity of total expenditures with

respect to income is approximately 0.4, a similar result to that derived

from estimates of the household distribution of expenditures, discussed

above. When Bangkok is excluded from the data set the relationship between

provincial expenditures and provincial incomes declines, but does not vanish.

Finally, Table 5.16 summarizes, in a similar way, the estimated relationship

Table 5.14 Thailand: distribution of government expenditures by province,

1997 to 1999, sectoral components

Bangkok Bangkok

included excluded

Part 1: Independent Mean provincial incomes

variable: per capita, 1996

Dependent variable

Education coeffi cient 5.15 1.92

t-value 1.77 0.63

Health coeffi cient 2.34 2.13

t-value 1.44 1.19

Social services coeffi cient 2.45 1.26

t-value 0.46 0.22

Agriculture coeffi cient 14.90 15.91

t-value 5.04 4.90

Part 2: Independent Poverty incidence

variable: (headcount) 1996

Dependent variable

Education coeffi cient –11.5633 –6.52

t-value –0.97 –0.057

Health coeffi cient –3.91 –3.15

t-value –0.59 –0.47

Social services coeffi cient 23.60 26.21

t-value 1.12 1.22

Agriculture coeffi cient –7.94 –5.73

t-value –0.58 –0.41

Part 3: Independent Rural population

variable: share 1996

Dependent variable

Education coeffi cient –25.47 –10.43

t-value –2.46 –0.83

Health coeffi cient –10.47 –11.50

t-value –1.80 –1.59

Social services coeffi cient 5.40 19.50

t-value 0.28 0.82

Agriculture coeffi cient 2.42 16.55

t-value 0.19 1.09

Source: Calculations by authors using data from Mokoro (1999), with supplementary data

from the Comptroller General’s Offi ce, Bangkok.

Table 5.15 Thailand: government expenditures by province, 1997 to 1999,

estimated coeffi cients and elasticities with respect to provincial

incomes

Estimated coeffi cient

Bangkok included Bangkok excluded

Total expenditure 64.18 43.43

Current 34.68 17.35

Capital 29.51 26.08

Implied elasticity

Bangkok included Bangkok excluded

Total expenditure 0.421 0.287

Current 0.386 0.195

Capital 0.471 0.420

Source: Calculated by authors from results in Table 5.13 and data on which they are based.

Table 5.16 Thailand: government expenditures by province, 1997 to 1999,

sectoral components; estimated coeffi cients and elasticities

with respect to provincial incomes

Estimated coeffi cient

Bangkok included Bangkok excluded

Education 5.15 1.92

Health 2.34 2.13

Social 2.45 1.26

Agriculture 14.90 15.91

Implied elasticity

Bangkok included Bangkok excluded

Education 0.236 0.087

Health 0.166 0.146

Social 0.078 0.038

Agriculture 1.598 1.677

Source: Calculated by authors from results in Table 5.14 and data on which they are based.

between the provincial expenditures by sector and provincial incomes. The

estimated elasticities are all positive and between zero and unity for all

sectors except agriculture. For education, the estimated elasticities are

substantially higher when Bangkok is included in the data set, indicating that

educational expenditures favor Bangkok heavily, but that among provinces

outside Bangkok, richer provinces are not signifi cantly favored.

In the case of health expenditures, there is a positive and signifi cant

relationship between per capita expenditures and incomes whether Bangkok

is included in the data or not. Social services expenditures are positively

related to incomes, but this relationship is not statistically signifi cant. What

these results do indicate, however, is that the claim that social services

expenditures favor poorer provinces is unsupported by these data.

Agricultural expenditures are the most strongly related to provincial

incomes. Agricultural expenditures apparently favor richer provinces and

this relationship is the strongest (estimated elasticity over one) of all of

the forms of expenditure covered by the data. This surprising result is

unchanged by the removal of Bangkok from the data set. While it is true

that Thailand's rural populations tend to be the poorest, Table 5.14 shows

that agricultural expenditures per person are not signifi cantly related to the

rural population share.

These results are, in general, strikingly supportive of the results reported

by Chalongphob and Direk (1999) and summarized in Table 5.12. The strong

exception relates to agriculture. Chalongphob and Direk’s estimates imply

a negative relationship with household incomes. This could be reconciled

with the strongly positive relationship between provincial expenditure

and provincial income, shown in Table 5.13, only if it were supposed that

agricultural expenditures within provinces were allocated in a manner which

very strongly favored lower income groups. This seems improbable. Future

research may illuminate this matter further.

In summary, in so far as total expenditures, education spending and

health spending are concerned, the data suggest that the provinces with

higher incomes per person receive higher levels of expenditure per person,

with elasticities between zero and one. This is even more true of spending

on agriculture, where the elasticity is between one and two.