Some Broad Targeting Experiments

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 

Given the very modest sums of money attached to direct anti-poverty

programs and given the apparently weak poverty impact, an important

question is how far broad targeting based on general categories of public

expenditure can meet the goal of poverty reduction. To address this question

we apply a simulation exercise looking at the poverty and growth impact of

different expenditure packages. The calculations are based on parameters

from a regression model developed by one of the authors (Balisacan and

Pernia, 2003).

Balisacan and Pernia (2003) compiled longitudinal provincial data for

the 1980s and 1990s to examine empirically the link between the average

expenditure of the various population quintiles, on the one hand, and overall

income growth and other factors on the other. The dependent variable

is average per capita expenditure in each quintile. The impact of overall

provincial income growth on poverty reduction is distinguished from the

direct impact of certain economic and institutional factors.

The explanatory variables are categorized into two groups, namely initial

condition variables and time-varying variables. Included in the fi rst group are

province-specifi c human capital endowment, farm and land characteristics,

social capital, geographic attributes, and political economy characteristics.

The proxy for initial human capital endowment is the (three-year lagged)

average years of schooling of household heads. Two alternative variables

representing farm characteristics are average farm size and irrigation.

The latter, expressed as the ratio of irrigated land to total farm area, is

a proxy for the quality of agricultural land. Geographic attributes are an

indication of spatial isolation or high transport cost (given by a dummy

variable indicating whether a province is landlocked or not) and the average

frequency of typhoons hitting the province. These variables are intended to

capture geographic poverty traps. Meanwhile, the initial political economy

variables aim to refl ect the quality of local governance and access to fi scal

resources. One variable is ‘local political dynasty’, defi ned as the proportion

of local offi cials – related to each other by blood or affi nity – out of the total

number of elective positions. This variable is meant to capture the extent

of collusion or competition in local politics. The other political economy

variable pertains to the political party affi liation of the provincial chief

executive. This is represented by a dummy variable indicating whether the

provincial governor belongs to the national President’s political party.

The time-varying variables include relative price incentives, road access

and electricity, agrarian reform, and overall average per capita income. The

price incentives variable is given by the agricultural terms-of-trade, defi ned

as the ratio of the price of agricultural to non-agricultural products. The

time-varying infrastructure variables pertain to road access and electricity.

The roads variable, representing access to markets, off-farm employment,

and social services, is defi ned as quality-adjusted road length per square

kilometer of land area. Electricity is used as a proxy for access to technology,

or simply the ability to use modern equipment. It is defi ned as the proportion

of households with access to electricity. The agrarian reform variable,

defi ned as the proportion of the cumulative completed agrarian reform

area to total potential land reform area, serves as a proxy for households’

ability to smooth consumption in response to shocks, given imperfections

in credit markets.

Certain variables may have strong complementarities, so the impact of

one variable on the living standards of the poor may be conditioned by the

values of the other variables. To allow for this possibility, interaction terms

on certain variables are introduced; in particular, schooling and roads, and

schooling and electricity.

The econometric estimation takes into account the possibility of a reverse

causation in the poverty–growth relationship, so overall mean income may

systematically respond to changes in the average living standards of the

poor. The regression results for each of the quintiles are reproduced in

Table 6.5.

In general, the results for the second quintile closely resemble those for

the fi rst quintile. This is signifi cant considering that estimates of poverty in

the Philippines vary widely – from 20 per cent to 40 per cent – depending

on, among other things, the poverty norm employed. The offi cial estimate

roughly corresponds to the bottom 40 per cent of the population.

Other observations are also worth noting:

• The growth elasticity of poverty tends to increase monotonically with

income quintile. This confi rms what has been noted above, that the

benefi ts of growth are unevenly spread throughout the various income


• The roads variable is signifi cant, but has a negative sign for the fi rst

three quintiles, suggesting that roads per se directly reduce the welfare

of the poor, unless complementary factors like schooling are present.

In contrast, this variable is signifi cant and has a positive sign for the

top quintile, indicating that roads raise directly the average welfare

of the richest group in society, as expected.

• Apart from its impact through other channels, overall schooling does

not seem to have a direct, signifi cant effect on average welfare for

all quintiles. However, as noted above, when interacted with roads,

schooling tends to raise the average income and welfare in the fi rst

three quintiles. This suggests that complementarity matters for other

quintiles as well.6

• Other things being equal, agrarian reform raises the average welfare

of all quintiles, except the top one. Note that those in the top 20

per cent do not normally depend on agriculture for employment and


• Irrigation tends to have a pro-poor bias. Farm size does not have

signifi cant effects on the average welfare of all but the richest group,

implying that it is the quality of the land, not farm size per se, that

favorably affects the welfare of the lower-income groups.