Empir Econ (2011) 41:57–80 DOI 10.1007/s00181-010-0392-5
Financial development and poverty in developing countries: a causal analysis Salvador Perez-Moreno
Received: 16 January 2009 / Accepted: 27 April 2010 / Published online: 22 September 2010 © Springer-Verlag 2010
Abstract This article empirically examines the possible causal links between financial development and poverty in developing countries. To this end, we apply a modified form of traditional Granger causality tests to suit the short times series that are available. We conclude that the evidence supports the hypothesis that in the period of the 1970s–1980s financial development, measured by liquid assets of the financial system as a share of GDP or by money and quasi money as a percentage of GDP, leads to the reduction of moderate poverty. These results do not appear for the period of the 1980s–1990s or when financial development is measured by the ratio of the value of credits granted by financial intermediaries to the private sector to GDP, whereas they seem to be strengthened by using summary measures of financial development. Likewise, our analysis does not show any evidence of Granger causality from poverty to financial development. Keywords
Poverty · Financial development · Causality · Developing countries
JEL Classification
O16 · I32 · C23
1 Introduction A significant feature of the literature on development finance is the renewed interest in the links between financial development and the pace of economic growth. Since Schumpeter (1911), and later with Goldsmith (1969); McKinnon (1973) and
S. Perez-Moreno (B) Department of Applied Economics (Economic Policy), University of Malaga, Campus El Ejido, 29071 Málaga, Spain e-mail:
[email protected]
123
58
S. Perez-Moreno
Shaw (1973), this issue has been extensively researched, yielding considerable evidence that financial development correlates with growth. In recent decades, numerous authors have examined this relationship, contributing important theoretical and empirical arguments (Gupta 1984; Jung 1986; Demetriades and Hussein 1996; Levine 1997; Arestis and Demetriades 1997; Levine et al. 2000; Arestis et al. 2001; Calderón and Liu 2003; Christopoulos and Tsionas 2004, among others). In general terms, the empirical results suggest that financial development enhances economic growth and, simultaneously, growth propels financial development, as the expansion of the real sector may have a notable influence on the development of the financial sector. At the same time, another topic extensively studied in development economics is the relationship between economic growth and poverty reduction. Indeed, in current literature, there is widespread consensus regarding the major importance of growth in order to reduce poverty, though initial inequality may affect the impact of growth on poverty reduction. This relationship has been widely proven in different contexts and circumstances by considering the poor as a pre-specified proportion of the population— usually the lowest quintile (Dollar and Kray 2002; Foster and Szekely 2000), and by using a definition of poverty in which the poor are people with income/expenditure levels below a pre-determined threshold—for instance, purchasing power parity (PPP) adjusted US$1 per person per day, or a country-specific poverty line computed on the basis of the cost of a country-specific subsistence package (Ravallion and Chen 1997; Adams 2004). Although most researchers have focused on the analysis of the relationship in the direction of causality growth–poverty, some studies have highlighted a possible twofold causal link, taking into account that poverty reduction can also be regarded as a causal factor for economic growth in the developing world, since poverty reduction may improve economic performance and efficiency and thus contribute to boosting economic growth (Lustig et al. 2002). In contrast to the significant attention paid to these relationships, the causal linkages between financial development and poverty have received much less attention in the literature. In this context, this article attempts to carry out a causal analysis of the links between financial development and poverty in developing countries, in order to shed a little more light on the likely causal linkages. In particular, we are interested in testing not only whether there exists any causal relationship in the direction from financial development to poverty, but also whether there exists a possible causal link from poverty to financial development. In any case, the problem of testing for causality between financial development and poverty is considerable because of the scarcity of uniform annual data for most countries. The implementation of traditional time-series, Granger causality tests requires long time series. For this reason, this study relies on other econometric techniques that allow using a data panel of different countries and exploiting the cross section variation. To that end, we have reconsidered and modified the original spirit of Granger (1969) paper to apply it to the case of panel data by taking the methodological scheme used by Weinhold and Reis (2001) and Perez-Moreno (2009) as reference, though we use bootstrapped standard errors in the application of the sum-difference test in order to obtain more robust standard errors.
123
Financial development and poverty in developing countries
59
The rest of this article is organized as follows. In the next section, we briefly review the theoretical and empirical literature. Section 3 presents the panel data set and the methodology. Section 4 goes on to discuss the empirical results. Finally, the last section summarizes the conclusions reached.
2 Brief literature review Development economics theory provides different predictions about the impact of financial development on poverty as well as on income inequality, considering that both concepts are different although closely related. We can assert, therefore, that consensus does not exist in the theoretical literature on whether financial development benefits the whole population equally, or whether it disproportionately benefits the rich or the poor. A common view is that financial development might benefit the rich. In this way, Rajan and Zingales (2003) explain that the financial system, especially when institutions are weak, might mainly channel money to the rich and well connected, who are able to offer collateral and who might be more likely to repay a loan, while excluding the poor. As financial sectors become more developed, they might lend more to people who are able to provide collateral, but continue to neglect the poor, who remain unable to invest in human and physical capital, or start new businesses. Therefore, high-income households benefit more from financial development than low-income households. Other theoretical models suggest that financial development enhances growth and reduces inequality. Capital market imperfections, such as information and transaction costs, may especially affect the poor who lack collateral and credit histories. In this context, Galor and Zeira (1993); Banerjee and Newman (1993); Aghion and Bolton (1997) and Galor and Moav (2004) argue that credit constraints reduce the efficiency of capital allocation and intensify income inequality by impeding the flow of capital to poor individuals with high expected return investments. From this perspective, financial development reduces poverty both by improving the efficiency of capital allocation, which accelerates aggregate growth, and by relaxing credit constraints that restrain the poor more extensively. A third approach is the proposal developed by Greenwood and Jovanovic (1990) who suggest that it is possible that different mechanisms dominate at different levels of financial development, and predict a nonlinear relationship between income inequality and financial sector development. In their model, income inequality first increases as the financial sector develops, but later declines as more people gain access to the system. At early stages of development, only the rich can afford to access and directly profit from better financial markets. At higher levels of economic development, many people access financial markets so that financial development directly helps a larger proportion of society. On the other hand, explicit references in the literature to the impact of poverty reduction on financial development are scarce. Nonetheless, certain authors have pointed out possible reasons behind this direction of causality. Thus, Beck et al. (2007, p. 34) indicate some examples that may explain the causal links from poverty and inequality to
123
60
S. Perez-Moreno
financial development. They suggest that reductions in poverty may stimulate demand for financial services, and reductions in income inequality might lead to political pressures to create more efficient financial systems that fund projects based on market criteria, not political connections. From the empirical point of view, several studies have recently examined the relationship between financial development and income distribution with various conclusions as well, focusing particularly on the impact of financial development on poverty and inequality (Dollar and Kray 2002; Honohan 2004; Clarke et al. 2006; Beck et al. 2007; Guillaumont and Kpodar 2008; Ang 2010). Dollar and Kray (2002) show that financial development, together with other progrowth macroeconomic policies, raises average income but with little systematic effect on its distribution. However, Honohan (2004) proves that finance-intensive growth (measured by banking depth) is empirically associated with lower poverty ratios, even after allowing for mean income and inequality. Likewise, Clarke et al. (2006) find that in the long run, inequality is less when financial development is greater and, at the same time, suggest that inequality might increase as financial sector development increases at very low levels of financial sector development, although this result is not robust. In turn, Beck et al. (2007) find that the level of financial development reduces the growth rate of the Gini coefficient even when conditioning, on average, growth and lagged values of income inequality. They also reveal that financial development is strongly linked to declines in the fraction of the population living on less than $1 per day. On the other hand, Guillaumont and Kpodar (2008) seek to identify and quantify the positive and negative channels through which financial development affects poverty, and conclude that financial development is on average good for the poor, with the direct effect being stronger than the effect through economic growth. However, these authors also point out that financial instability hurts the poor and partially offsets the benefits of financial development. Finally, Ang (2010) examines how finance impacts income inequality in India, and he observes that while financial development helps reduce income inequality, financial liberalization seems to exacerbate it. 3 Data and methodology This section describes the database and discusses the methodological framework used in this article to analyze the causal relationship between financial development and poverty in developing countries.1 3.1 The data We use a panel of 35 developing countries for the years 1970, 1980, 1990 and 1998, with poverty data estimated by Sala-i-Martin (2002) for the conventional poverty 1 Note that finance studies typically focus on both developed and developing countries. In this article, in contrast to Dollar and Kray (2002); Clarke et al. (2006) and Beck et al. (2007), we only consider developing economies. Other studies that focus exclusively on developing countries are, for example, Guillaumont and Kpodar (2008) and Ang (2010), although the latter focuses on a specific country, India.
123
Financial development and poverty in developing countries
61
lines of $1 and $2 a day in 1985 PPP rates, and financial development data from World Development Indicators (World Bank 2007) and Beck et al. (2000), updated in 2007. The sample consists of all developing countries for which Sala-i-Martin (2002) provides poverty data and for which there exist financial development data for all years considered. By regions, it comprises 14 countries from Sub-Saharan Africa, 14 from Latin America and the Caribbean, 4 from South Asia, and 3 from East Asia and the Pacific (see Appendix 1 for countries by region). In measuring poverty, we therefore use two different poverty indicators that reflect the incidence (or prevalence) of poverty: poverty headcount ratio at $1 a day (extreme poverty) and poverty headcount ratio at $2 a day (moderate poverty). These indicators measure the share of the population living on less than $1 and $2 a day, respectively, based on the 1985 PPP exchange rate.2 On the other hand, financial development is measured, in principle, by two indicators commonly used in the literature, which capture different aspects of financial development: i. M3/GDP: The liquid assets of the financial system (currency plus demand and interest-bearing liabilities of banks and non banks)3 as a share of gross domestic product (GDP). This indicator is related to the ability of financial systems to provide transaction services and saving opportunities. ii. Private credit/GDP: The value of credits granted by financial intermediaries to the private sector as a share of GDP. It comprises credit to private firms and households from banks and nonbank financial intermediaries (but excludes central banks as lenders and government and state-owned enterprises as borrowers). This indicator is a good proxy variable for the extent to which private sector agents have access to financial intermediation or access to loans. According to Guillaumont and Kpodar (2008, p. 10), although the indicator Private credit/GDP has the advantage of measuring more accurately the role of financial intermediaries in channeling funds to productive agents and possibly to the poor, the use of a liquidity ratio is fundamental, as it allows us to assess whether financial intermediaries are actually helpful for the poor in supplying money balances or credits. In order to assess the robustness of the results obtained from using a broad measure of the money stock as an indicator of financial development, we additionally consider the ratio of money and quasi money (M2) to GDP (M2/GDP), given that it is also a measure commonly used in financial development literature for causality studies (see, e.g. Gupta 1984; Jung 1986 and Calderón and Liu 2003). A higher M2/GDP ratio implies a larger financial sector and therefore greater financial intermediary development. 2 Let us recall that Sala-i-Martin’s poverty data have been estimated for each country yearly by using income shares. Since microeconomic surveys are not available annually for every country, he imputes the missing data by forecasting quintile income shares or by assigning average quintile income shares, and uses a non-parametric approach to estimate a smooth income distribution for each country/year (see Sala-i-Martin 2002). 3 Liquid assets are also known as broad money, or M3. They are the sum of currency and deposits in the central bank (M0), plus transferable deposits and electronic currency (M1), plus time and savings deposits, foreign currency transferable deposits, certificates of deposit, and securities repurchase agreements (M2), plus travelers checks, foreign currency time deposits, commercial paper, and shares of mutual funds or market funds held by residents.
123
62
S. Perez-Moreno
The data on private credit over GDP proceed from the database on financial development and structure developed by Beck et al. (2000), updated in 2007, while the data on liquid assets of the financial system as a share of GDP, and money and quasi money as a percentage of GDP proceed from World Development Indicators, as well as those on GDP per capita (constant 1995 US$), which are utilized in our models to control for the time-invariant characteristics.
3.2 The methodology In respect to the objective of our study outlined above, our central concern is essentially a time series question: we would like to know whether financial development over time is causally related to poverty and, at the same time, whether the reverse is likely as well. In this sense, it seems obvious that a cross-section analysis, even of growth rates, will not be able to fully capture such dynamic effects. Actually, we would need to adopt some econometric technique for panel model evaluation to test for a type of Granger causality. Hence, we should basically test whether knowledge about past financial development could help us predict future changes of poverty and vice versa. Let us recall that Granger’s classic (Granger 1969) paper established a presently well-known method of testing the direction of causality in econometric models. The underlying premise of the test is based on the notion that if some variable x causes some other variable y, then the addition of lagged values of x in a regression of y on its own lagged values and other explanatory variables should significantly improve the forecasting power of the model. If the reverse case is also true, then a substantial amount of feedback is said to exist in the system. In accordance with Weinhold and Reis (2001), we have reconsidered and modified the original spirit of Granger (1969) paper to apply it to the case of panel data. We have taken into account the following four models: Vpovit = α + β1 Vpovit−1 + β2 L povit−1 + β3 Vfdit−1 + β4 L fdit−1 +β5 Vgdpit−1 + β6 L gdpit−1 + εit
(1)
Vpovit = α + β1 Vpovit−1 + β2 L povit−1 + β3 Vgdpit−1 + β4 L gdpit−1 + εit Vfdit = α + β1 Vfdit−1 + β2 L fdit−1 + β3 Vpovit−1 + β4 L povit−1
(2)
+β5 Vgdpit−1 + β6 L gdpit−1 + εit Vfdit = α + β1 Vfdit−1 + β2 L fdit−1 + β3 Vgdpit−1 + β4 L gdpit−1 + εit
(3) (4)
where Vpov and L pov are the variation rate and log-level of the respective poverty indicators (poverty headcount ratio at $1 and $2 a day), Vfd and L fd are the variation rate and log-level of the financial development measures (the ratio of the liquid assets to GDP—M3/GDP—the ratio of the value of credits granted by financial intermediaries to the private sector to GDP—Private credit/GDP—and the ratio of money and quasi money to GDP—M2/GDP), while Vgdp and L gdp are the variation rate and log-level of GDP per capita. Following standard procedure, we take log on the variables and
123
Financial development and poverty in developing countries
63
interpret their first differences as variation rates. Note that the level terms are measured at the beginning of the period. We control for the specific time-invariant characteristics (so-called “fixed effects”) that could be driving the levels of both financial development and poverty rates by modeling our variables as variation rates and by controlling for initial log-levels of both the dependent and the independent variables. We include the log-level of the GDP per capita and its first difference as control variables in so far as GDP per capita and its variation rate may be interpreted as summary variables reflecting the situation of many other socioeconomic variables. Models (1) and (2) are rival models of poverty variations, and models (3) and (4) are rival models of financial development variations. Models (1) and (3) include information on past variation rates and levels of both variables under consideration, and models (2) and (4) only include lagged endogenous variables. Thus, for instance, if model (1) can forecast poverty variations more accurately than model (2), we deduce that information about past financial development variations was important. However, if model (2) forecasts better than model (1), or there is no difference, then we conclude that including information about financial development variations does not help to predict poverty variations, and therefore there is no causal relationship in the direction from financial development to poverty. Following Granger and Huang (1997) and Weinhold and Reis (2001), we have adopted a sum-difference test that takes the following form. Consider the forecast errors η1it = η2it from models (1) and (2) (or from models (3) and (4)) where i and t denote not-in-sample cross section countries and time periods. The null hypothesis is 2 2 = η2it . H0 : η1it
(5)
It then follows that if H0 can be rejected, the model with the lowest forecast error variance should be accepted as being significantly superior to the competing model. For this purpose, Granger and Huang suggest constructing the following variables: SUMi12 = η1i + η2i DIFi12 = η1i − η2i .
(6) (7)
Then a test of the null hypothesis is equivalent to a test of whether δ = 0 from the regression SUMi12 = α + δ · DIFFi12 + ξi .
(8)
In our case, we regress the model (8) using bootstrapped standard errors, which are more robust than traditional standard errors. Since the methodological approach proposed only requires two observations of the variation rates, and therefore three time observations of the level variables, our data set allows examining the causal linkages between financial development and poverty in two different periods: first the 1970s and 1980s, and second the 1980s and 1990s. An advantage of this approach for data like ours, which are quite limited in the time series, is that we do not rely on precise estimates of the sample parameters but rather
123
64
S. Perez-Moreno
on running a model competition in which all models are equally handicapped by the scarcity of time series observations. In turn, the 10-year periodicity of the data implies that each observation embodies a great deal of unique information and the subsequent time series variation is less likely to be driven by cyclical and short term shocks. If the average periodicity were only a few years, then the data could be more correlated and the detection of important aspects of the dynamics would be more difficult. Nevertheless, the short time series does preclude a traditional Granger-causality approach, in which we predict out-of-sample across time. With our short panel data set, we take advantage of the fact that we have variation across space as well as through time. In particular, we estimate the models on N-1 cross section observations, and use the resulting coefficient estimates to generate a forecast of the dependent variable for the remaining (not-in-sample) unit. In this way, we systematically generate N different forecast errors for each model. Then, we test whether these forecast errors are statistically different from each other, as described above.
4 Empirical results In Tables 1, 2, 3 and 4 we present the estimated results of the proposed models for the periods 1970s–1980s (Tables 1 and 3) and 1980s–1990s (Tables 2 and 4). Although the scarcity of data entails important limitations in estimating the models and the adjustments are not satisfactory enough, these estimations are useful in order to chiefly assess the sign and significance of the linkages between the variation rates of poverty in time t (Vpov it ) and the variation rates of financial development in time t −1 (Vfdit−1 ) (Tables 1 and 2), as well as between the variation rates of financial development in time t (Vfdit ) and the variation rates of poverty in time t − 1 (Vpov it−1 ) (Table 3 and 4). In Table 1, the estimated models reflect that Vfdit−1 is a negative exploratory variable of Vpov it , which might be interpreted as evidence that those countries with a growth of financial development in the 1970s reduced poverty in the 1980s and conversely. This linkage is highly significant when financial development is measured by liquid assets of the financial system as a share of GDP (M3/GDP) and by money and quasi money as a percentage of GDP (M2/GDP), and poverty by the headcount ratio at $2 a day. However, Table 2 shows that in the period of the 1980s–1990s this connection between financial development and poverty is positive, although not significant, with all the indicators. Regarding the models in which the variation in financial development is a function of the variation in poverty, Tables 3 and 4 manifestly reveal that the relationship is not statistically significant and the sign is not clear, though it is mostly negative when the poverty headcount ratio at $2 a day is used. Tables 5, 6, 7 and 8 reflect the results of the modified form of the traditional Granger causality test discussed in the previous section. In particular, Tables 5 and 6 present summary statistics of the mean squared forecasting error from the two competing models of poverty variations (models 1 and 2) and the two competing models of financial development variations (models 3 and 4) for the period of 1970s–1980s and
123
(1.29)
−0.2293 0.5801
(−0.27) (1.71)*
−0.8102 −0.5580
−0.3759
(−0.61)
−0.9176
−0.5689
(0.93)
0.3992
(2.20)**
3.5701
(2)
−0.8295
(−0.22)
−1.1703
(0.30)
1.6874
(1)
(1.34)
2.6457
(1)
−0.8102 −0.5619
(−0.27) (1.12) −0.5689
(0.93)
0.3992
(2.20)**
3.5701
(2)
Poverty headcount ratio at $2 a day
−0.2293 0.4709
(1.29)
5.9563
(2)
Poverty headcount ratio at $1 a day
Private credit/GDP
−0.7091
(−0.48)
(0.30)
0.4507
(1)
−0.8102 −0.4517
(−0.27) (1.43)
−0.5689
(0.93)
0.3992
(2.20)**
3.5701
(2)
Poverty headcount ratio at $1 a day
−0.2293 0.5646
(1.29)
−0.2978
5.9563
−6.1305
(2)
(−0.81)
(1)
Poverty headcount ratio at $1 a day
M2/GPD
35
35
0.2761
35
0.6227
0.6893 35
0.3468
0.4237
(−1.38)
−0.2386
(0.60)
0.4136
−1.1572
35
0.2553
0.3867
(−0.39)
−0.2833
(−0.90)
−4.1188
(−1.02)
−1.1578
(−0.56)
−0.1193
35
0.2761
0.3612
35
0.3312
0.4492
(−0.99) (−0.70)
−0.6283 −0.1500
(−0.99) (0.84)
−5.1068 0.5891
(−1.37)
−0.1750
(−0.76)
35
0.3468
0.4237
(−1.38)
−0.2386
(0.60)
0.4136
−6.5614
35
0.4760
0.5685
(0.10)
0.0802
(−1.17)
−5.1380
(−2.05)**
−4.9539
(−1.72)*
−0.5840
35
0.2761
0.3612
35
0.5546
0.6332
(−0.99) (−0.01)
−0.6283 −0.0016
(−0.99) (1.28)
−5.1068 0.7492
(−4.29)***
−0.7417
(−2.61)**
35
0.3468
0.4237
(−1.38)
−0.2386
(0.60)
0.4136
Estimates were obtained using ordinary least squares, with the dependent variable being the difference in log of the respective poverty measure (Vpov it ). Heteroskedasticityconsistent t statistics are shown in parentheses *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level
0.4396
N
0.3612
(−0.99) (−0.31)
(−0.29)
0.5385
−0.6283 −0.0439
−0.2137
Adjusted R 2
(−0.99) (1.57)
(−1.01)
(−5.28)***
(−1.88)*
−5.1068 0.8645
−0.8290
−4.6742
−4.3458
−0.7686
(−4.32)***
−6.2249
(−1.62)
(−6.09)*** (−2.00)* (−4.60)*** (−3.19)*** (−3.25)*** (−2.00)* (−3.22)*** (−3.19)*** (−2.64)** (−2.00)* (−2.70)** (−3.19)***
(0.92)
1.1528
5.9563
−2.2542
(0.33)
R2
L gdpit−1
Vgdpit−1
L fdit−1
Vfdit−1
L povit−1
Vpovit−1
Constant
(1)
(2)
Poverty headcount ratio at $2 a day
(1)
Dependent variable Vpovit Poverty headcount ratio at $1 a day
M3/GPD
Table 1 Models of poverty variation including and excluding financial development variation (1970s–1980s)
Financial development and poverty in developing countries 65
123
123
(−0.05)
−0.0833
(−0.12) (1.34)
0.1937
(4.14)***
N
−0.0488
35
35
0.5458
0.6260
(0.95)
0.1350
35
0.5077
0.5656
(0.55)
0.0570
(−0.59)
−0.1914
(1.77)*
0.1760
(5.64)***
35
−0.0895
0.1028
(−0.58)
−0.6536
(−1.36)
−5.8074
(0.77)
0.8479
(1.05)
1.3762
(−0.02)
−0.0135
(−0.27)
−0.0533
(0.52)
4.9692
(1)
35
−0.0488
0.0746
(−0.46)
−0.4143
(−1.29)
−5.2281
(−0.05)
−0.0337
(−0.67)
−01259
(0.28)
1.9344
(2)
Poverty headcount ratio at $1 a day
Private credit/GDP
35
0.4998
0.5881
(1.03)
0.1178
(−0.20)
−0.0754
(−1.01)
−0.1271
(0.10)
0.0174
(1.72)*
0.1918
(4.27)***
0.6477
(−1.66)
−1.8278
(1)
35
0.5077
0.5656
(0.55)
0.0570
(−0.59)
−0.1914
(1.77)*
0.1760
(5.64)***
0.6238
(−1.19)
−1.158
(2)
Poverty headcount ratio at $2 a day
35
0.0434
0.2122
(−0.36)
−0.3996
(−0.97)
−4.1932
(0.12)
0.1914
(1.31)
4.6129
(−0.10)
−0.0739
(0.19)
0.0447
(0.26)
2.6194
(1)
35
−0.0488
0.0746
(−0.46)
−0.4143
(−1.29)
−5.2281
(−0.05)
−0.0337
(−0.67)
−01259
(0.28)
1.9344
(2)
Poverty headcount ratio at $1 a day
M2/GPD
35
0.5209
0.6055
(0.94)
0.1237
(0.04)
0.0151
(−1.04)
−0.2663
(0.86)
0.2087
(1.64)
0.1897
(5.04)***
0.6327
(−1.49)
−2.0147
(1)
35
0.5077
0.5656
(0.55)
0.0570
(−0.59)
−0.1914
(1.77)*
0.1760
(5.64)***
0.6238
(−1.19)
−1.158
(2)
Poverty headcount ratio at $1 a day
Estimates were obtained using ordinary least squares, with the dependent variable being the difference in log of the respective poverty measure (Vpov it ). Heteroskedasticityconsistent t statistics are shown in parentheses. *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level
−0.0174
35
Adjusted R 2
0.0746
−0.4143
(−0.46)
−0.4881
(−0.43)
0.1622
(−1.29)
(−1.20) (0.12)
(−1.03) 0.0503
−5.2281
−5.7451
−0.1000
(−0.05)
(1.07) −0.2894
(1.09)
0.3177
−0.0337
(−0.37)
3.9563
−01259
(−0.67)
−0.1042 0.6238
−1.158 (−1.19)
−2.1264 (−1.45)
1.9344
(0.28)
2.1840
(0.21) 0.6463
(2)
(1)
(1)
(2)
Poverty headcount ratio at $2 a day
Poverty headcount ratio at $1 a day
R2
L gdpit−1
Vgdpit−1
L fdit−1
Vfdit−1
L povit−1
Vpovit−1
Constant
Dependent variable Vpovit
M3/GPD
Table 2 Models of poverty variation including and excluding financial development variation (1980s–1990s)
66 S. Perez-Moreno
(−0.17)
(−1.43)
−0.2670
−0.1458
35
Adjusted R 2
N
35
−0.0950
0.0338 −0.1289 35
35
−0.0950
0.0338
(−0.44)
35
0.0326
0.2033
(1.20)
0.1612
−0.1844
35
0.0001
0.1177
(0.64)
0.0759
(0.29)
0.1245
(−1.49)
−0.2677
(−0.62)
35
0.0190
0.1921
(0.97)
0.1889
(−0.31)
−0.1827
(0.91)
0.2229
(−0.22)
−0.0792
(−1.30)
−0.2469
(−0.48)
−0.1522
(−1.18)
−2.4641
(3)
35
0.0001
0.1177
(0.64)
0.0759
(0.29)
0.1245
(−1.49)
−0.2677
(−0.62)
−0.1844
(−1.08)
−1.0046
(4)
Poverty headcount ratio at $2 a day
35
−0.0348
0.1478
(0.92)
0.0608
(−0.43)
−0.1502
(1.50)
0.0431
(0.21)
0.0191
(0.23)
0.0326
(0.43)
0.0992
(−0.85)
−0.4930
(3)
35
−0.1098
0.0207
(−0.35)
−0.0248
(−0.14)
−0.0419
(0.53)
0.0775
(0.59)
0.1263
(0.32)
0.1921
(4)
Poverty headcount ratio at $1 a day
M2/GPD
35
−0.0782
0.1120
(0.32)
0.0426
(−0.73)
−0.2935
(0.64)
0.1463
(−0.45)
−0.0960
(0.51)
0.0678
(0.56)
0.1169
(−0.45)
−0.7613
(3)
35
−0.1098
0.0207
(−0.35)
−0.0248
(−0.14)
−0.0419
(0.53)
0.0775
(0.59)
0.1263
(0.32)
0.1921
(4)
Poverty headcount ratio at $1 a day
Estimates were obtained using ordinary least squares, with the dependent variable being the difference in log of the respective financial development measure (Vfdit ) Heteroskedasticity-consistent t statistics are shown in parentheses *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level
0.0564
0.0703
(−0.44)
(0.08)
−0.0212
−0.0329 (−0.32)
−0.0212
0.0046
(−0.35)
(0.27)
(−0.26)
(0.27)
(0.06)
(0.88) −0.1515
0.0729
0.0729
0.0206
(0.24)
(0.67)
0.0489
−0.1148
0.0290
0.0148
(−0.12)
(0.03)
−0.0215
(−0.41)
−0.1335
(−0.42)
(−0.17)
(−0.13)
0.0044
−0.1452 (−0.77)
(0.02)
−0.0215
−0.0164
(−0.53)
−0.1114
−1.0046 (−1.08)
−1.6389 (−1.50)
−0.0123
(−0.77)
0.1897 (0.43)
−0.1090
−0.1452
−0.1166
(−0.54)
0.1936 (0.18)
(4)
(3)
Poverty headcount ratio at $1 a day
Private credit/GDP
0.0023
0.1897
(0.43)
0.0010
(0.00)
(4)
(3)
(3)
(4)
Poverty headcount ratio at $2 a day
Poverty headcount ratio at $1 a day
R2
L gdpit−1
Vgdpit−1
L povit−1
Vpovit−1
L fdit−1
Vfdit−1
Constant
Dependent variable Vfdit
M3/GPD
Table 3 Models of financial development variation including and excluding poverty variation (1970s–1980s)
Financial development and poverty in developing countries 67
123
123
(−0.70)
(−0.48)
(−0.69)
−0.1662
(−0.78)
35
N
35
0.1192 35
0.1013
0.2599
(0.71)
0.1503
(1.08)
−0.6725
35
0.1192
0.2228
(1.57)
0.0923
(1.98)*
0.9112
(−0.70)
−0.1175
(−0.68)
−0.2180
(−1.29)
−2.5037
35
0.2264
0.3629
(1.87)*
0.3117
(2.47)**
1.5053
(0.41)
0.0335
(0.14)
0.0044
(−1.63)
−0.2763
(−0.23)
−0.0627
(−1.83)
−2.0535
35
0.2578
0.3451
(2.54)**
0.2509
(3.59)***
1.4734
(−1.81)*
−0.2761
(−0.14)
−0.0310
(−2.36)**
35
0.2629
0.3930
(1.60)
0.4347
(1.63)
1.6320
(0.65)
0.2027
(0.45)
0.2001
(−1.68)
−0.2738
(0.25)
0.7009
(−1.33)
−3.9239
(3)
35
0.2578
0.3451
(2.54)**
0.2509
(3.59)***
1.4734
(−1.81)*
−0.2761
(−0.14)
−0.0310
(−2.36)**
−2.0535
(4)
Poverty headcount ratio at $2 a day
35
0.1450
0.2959
(0.68)
0.0996
(0.80)
0.4492
(0.35)
0.0319
(−0.50)
−0.0158
(0.00)
0.0005
(−0.03)
−0.0084
(−0.57)
−0.7068
(3)
35
0.0972
0.2034
(0.79)
0.0468
(1.58)
0.6332
(0.21)
0.0321
(0.55)
0.1687
(−0.46)
−0.2456
(4)
Poverty headcount ratio at $1 a day
M2/GPD
35
0.0677
0.2322
(0.88)
0.1475
(1.35)
0.6183
(0.56)
0.1172
(−0.05)
−0.0079
(0.00)
0.0002
(0.45)
0.1396
(−0.68)
−1.3426
(3)
35
0.0972
0.2034
(0.79)
0.0468
(1.58)
0.6332
(0.21)
0.0321
(0.55)
0.1687
(−0.46)
−0.2456
(4)
Poverty headcount ratio at $1 a day
Estimates were obtained using ordinary least squares, with the dependent variable being the difference in log of the respective financial development measure (Vfdit ) Heteroskedasticity-consistent t statistics are shown in parentheses. *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level
0.1605
Adjusted R 2
(1.57)
0.2228
(0.89)
0.3087
0.0923
0.1413
(1.98)*
(1.11)
0.7867
(0.27)
0.9112
(0.36)
0.7204
0.0719
0.0350
(−0.34)
−0.1175
−0.0795
(−0.34)
(−0.68)
(−0.85)
−0.2977
−0.1146
−0.2180
−0.2891
−1.3514 (−0.54)
−0.0090
−0.6725
(−1.29)
−0.9963
(−0.75)
(4)
(3)
(4)
(3)
(3)
Poverty headcount ratio at $1 a day
Poverty headcount ratio at $2 a day
Poverty headcount ratio at $1 a day (4)
Private credit/GDP
M3/GPD
R2
L gdpit−1
Vgdpit−1
L povit−1
Vpovit−1
L fdit−1
Vfdit−1
Constant
Dependent variable Vfdit
Table 4 Models of financial development variation including and excluding poverty variation (1980s–1990s)
68 S. Perez-Moreno
15.5094 (3.9940)
0.0859 (0.2971)
0.0721 (0.2724)
Model 2
Model 3
Model 4
0.0721 (0.2724)
0.0877 (0.3003)
0.2587 (0.5154)
0.1505 (0.3934)
Poverty headcount ratio at $2 a day N = 35
0.2300 (0.4862)
0.2474 (0.5047)
15.5093 (3.9940)
15.9716 (4.0527)
Poverty headcount ratio at $1 a day N = 35
Private credit/GDP
Out-of-sample causality test. Standard deviations of the forecast error are shown in parentheses
12.5990 (3.6012)
Model 1
Poverty headcount ratio at $1 a day N = 35
M3/GPD
Table 5 Mean-squared forecast error (1970s–1980s)
0.2300 (0.4862)
0.2579 (0.5145)
0.2587 (0.5154)
0.2623 (0.5189)
Poverty headcount ratio at $2 a day N = 35
0.1051 (0.3290)
0.1066 (0.3313)
15.5094 (3.9940)
12.4113 (3.5716)
Poverty headcount ratio at $1 a day N = 35
M2/GPD
0.1052 (0.3290)
0.1197 (0.3510)
0.2587 (0.5154)
0.1858 (0.4365)
Poverty headcount ratio at $2 a day N = 35
Financial development and poverty in developing countries 69
123
123
13.3870 (3.6968)
0.1938 (0.4443)
0.1254 (0.3592)
Model 2
Model 3
Model 4
0.1254 (0.3592)
0.1843 (0.4355)
0.1384 (0.3773)
0.1422 (0.3825)
Poverty headcount ratio at $2 a day N = 35
0.2363 (0.4931)
0.2929 (0.5482)
10.6626 (3.3077)
14.1285 (3.7961)
Poverty headcount ratio at $1 a day N = 35
Private credit/GDP
Out-of-sample causality test. Standard deviations of the forecast error are shown in parentheses
14.8072 (3.8880)
Model 1
Poverty headcount ratio at $1 a day N = 35
M3/GPD
Table 6 Mean-squared forecast error (1980s–1990s)
0.2363 (0.4931)
0.3137 (0.5677)
0.1384 (0.3773)
0.1479 (0.3900)
Poverty headcount ratio at $2 a day N = 35
0.1259 (0.3597)
0.1852 (0.4339)
13.3870 (3.6968)
14.4966 (3.8464)
Poverty headcount ratio at $1 a day N = 35
M2/GPD
0.1259 (0.3597)
0.1528 (0.3962)
0.1384 (0.3773)
0.1451 (0.3864)
Poverty headcount ratio at $2 a day N = 35
70 S. Perez-Moreno
1.7553 (0.79)
DIFFi j 4.0005 (3.62)
−0.9438 (−2.93)*** −0.0399 (−0.45)
−0.0119 (−0.13)
DIFFi j
Constant
−0.0273 (−0.19)
−0.5406 (−1.14)†
−0.0938 (−0.08)
Constant
2.2011 (1.10)
0.0529 (0.32)
0.2864 (0.26)
−0.2370 (−0.17)
Poverty headcount ratio at $1 a day N = 35
Private credit/GDP
1.9443 (1.45)
0.0301 (0.20)
0.2688 (0.26)
−0.0542 (−0.30)
Poverty headcount ratio at $2 a day N = 35
t Statistics shown in parentheses based on bootstrapped standard errors obtained from 100 bootstrap replications *** Significant at the 0.01 level; ** Significant at the 0.05 level; † Significant at the 0.25 level; ‡ Significant at the 0.30 level
Models 3 and 4
Models 1 and 2
Poverty headcount ratio at $2 a day N = 35
Poverty headcount ratio at $1 a day N = 35
M3/GPD
0.0057 (0.05) 1.8220 (1.07)
0.1977 (0.11)
−0.9061 (−1.99)**
−0.0517 (−0.32)
Poverty headcount ratio at $2 a day N = 35
0.0035 (0.03)
−0.6864 (−1.12)‡
−0.2716 (−0.23)
Poverty headcount ratio at $1 a day N = 35
M2/GPD
Table 7 Sum-difference test results for poverty variation (models 1 and 2) and financial development variation (models 3 and 4) (1970s–1980s)
Financial development and poverty in developing countries 71
123
123
0.0834 (0.78)
3.3835 (4.09)
DIFFi j
0.8359 (1.38)
DIFFi j
Constant
0.6695 (0.48)
Constant
3.0113 (2.88)
3.7309 (2.20)
1.5572 (2.40) 0.0546 (0.33)
0.1429 (0.23)
0.2710 (0.23)
−0.0303 (−0.25)
0.0145 (0.11)
0.8010 (0.68)
−0.0074 (−0.04)
0.7332 (0.79)
0.0168 (0.12)
Poverty headcount ratio at $2 a day N = 35
t statistics shown in parentheses based on bootstrapped standard errors obtained from 100 bootstrap replications. *** Significant at the 0.01 level; ** Significant at the 0.05 level; † Significant at the 0.25 level; ‡ Significant at the 0.30 level
Models 3 and 4
Models 1 and 2
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $2 a day N = 35
Private credit/GDP
M3/GPD
−0.0212 (−0.17) 2.8812 (1.73)
3.2098 (4.50)
0.4404 (0.55)
0.0150 (0.11)
Poverty headcount ratio at $2 a day N = 35
0.0497 (0.44)
0.6219 (0.72)
0.6740 (0.49)
Poverty headcount ratio at $1 a day N = 35
M2/GPD
Table 8 Sum-difference test results for poverty variation (models 1 and 2) and financial development variation (models 3 and 4) (1980s–1990s)
72 S. Perez-Moreno
Financial development and poverty in developing countries
73
1980s–1990s. In addition, Tables 7 and 8 include the corresponding results of the sum-difference test by using bootstrapped standard errors. In Table 5, we observe that in the 1970s–1980s period when financial development is measured by M3/GDP or M2/GDP, the models of poverty variations that include information on former variations of financial development (model 1) perform better at out-of-sample forecasting than the model that excludes financial development variations (model 2). The sum-difference test results presented in Table 7 indicate that the difference is statistically significant, particularly when we consider moderate poverty (threshold of $2 a day) rather than extreme poverty (threshold of $1 a day).4 Thus, there is compelling evidence to indicate that in this first period examined financial development leads to decreases of the percentage of the population living on less than $2 a day, as the financial development indicator basically reflects the ability of financial systems to provide transaction services and saving opportunities. Nevertheless, Table 6 shows that in the second period (1980s–1990s), the models of poverty variations do not perform better when information on financial development is included in the specification. Hence, we find that the causality relationship from financial development to poverty does not appear in the later period, in contrast with the earlier one. To properly interpret these differences in the results between both the periods, it would be necessary to research in depth the historical context and circumstances of developing countries during the three decades considered, examining key issues such as the evolution of the world economy, the political situation in developing nations, the financial sector reforms and the monetary and financial policies implemented at the international and national level, the debt crisis and its social costs, the IMF and the World Bank adjustment programs, or the pro-growth and pro-poor policies applied in these countries over time. If financial development is measured by the value of credits granted by financial intermediaries to the private sector as a share of GDP (Private credit/GDP), all the models of poverty variations, irrespective of the poverty line and period studied, show worse forecasting performance when information on financial development is included. Therefore, we cannot conclude that there exists a causal link from financial development to poverty when we focus on the access of private sector agents to financial intermediation and loans to measure financial development. In relation to the possible causal links from poverty to financial development, we find that the models of financial development variations perform worse when information on poverty is included in the specification (models 3 and 4), in both periods and for all the indicators of financial development and poverty employed (Tables 5 and 6). Hence, on the basis of our empirical evidence, we cannot affirm that in developing countries, changes in poverty rates are followed by modifications in the level of financial development, in a Granger-causal fashion. As the effects of financial development on poverty are to some extent sensitive to the indicators of financial development and there is no uniform argument as to which proxies are most appropriate for measuring financial development, it can be useful 4 Although the models of poverty variations with the headcount ratio at $1 a day also show a better forecasting performance when information on financial development is included, the difference of performance is hardly statistically significant (Table 7).
123
74
S. Perez-Moreno
Table 9 Mean-squared forecast error for summary measures of financial development (1970s–1980s) Summary measure 1
Summary measure 2
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $2 a day N = 35
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $2 a day N = 35
Model 1
15.2868 (3.9660)
0.2211 (0.4764)
15.5015 (3.9929)
0.2268 (0.4823)
Model 2
15.5094 (3.9940)
0.2587 (0.5154)
15.5094 (3.9940)
0.2587 (0.5154)
Model 3
0.1123 (0.3401)
0.1182 (0.3485)
0.1284 (0.3636)
0.1312 (0.3672)
Model 4
0.1004 (0.3214)
0.1004 (0.3214)
0.1154 (0.3445)
0.1154 (0.3445)
Out-of-sample causality test. Standard deviations of the forecast error are shown in parentheses
Table 10 Mean-squared forecast error for summary measures of financial development (1980s–1990s) Summary measure 1
Summary measure 2
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $2 a day N = 35
Poverty headcount ratio at $1 a day N = 35
Poverty headcount ratio at $2 a day N = 35
Model 1
14.9011 (3.8999)
0.1386 (0.3776)
14.6605 (3.8687)
0.1423 (0.3825)
Model 2
13.3870 (3.6968)
0.1384 (0.3773)
13.3870 (3.6968)
0.1384 (0.3773)
Model 3
0.2190 (0.4732)
0.2041 (0.4582)
0.1887 (0.4395)
0.2041 (0.4582)
Model 4
0.1513 (0.3945)
0.1513 (0.3945)
0.1407 (0.3804)
0.1513 (0.3945)
Out-of-sample causality test. Standard deviations of the forecast error are shown in parentheses
to consider summary measures of financial development based on several indicators (see, for example, Ang and McKibbin 2007). In our case, we construct two summary measures representative of financial sector development by taking the financial proxies used into account. In particular, we calculate the geometric mean of M3/GDP and Private credit/GDP (Summary measure 1), as well as the geometric mean of M2/GDP and Private credit/GDP (Summary measure 2). Tables 9, 10, 11 and 12 reflect the results of the modified form of the traditional Granger causality test by using these summary measures of financial development. They confirm that the models of poverty variations that include information on former variations of financial development (model 1) perform better at out-of-sample forecasting than the model that excludes financial development variations (model 2) in the 1970s–1980s period (Table 9), though the differences of performance are minor, and they are only somewhat statistically significant with respect to moderate poverty (Table 11). In any case, the discrepancies observed in the impact of financial development on poverty as a consequence of the financial development measure used are consistent with the results obtained by other researchers, although using different methodological approaches. Thus, for instance, Guillaumont and Kpodar (2008) find that M3/GDP level and instability are significantly correlated with the mean income of the poor, and draw attention to the positive direct effect of financial development on the standard
123
Financial development and poverty in developing countries
75
Table 11 Sum-difference test results for poverty variation (models 1 and 2) and financial development variation (models 3 and 4) for summary measures of financial development (1970s–1980s) Summary measure 1
Summary measure 2
Poverty headcount Poverty headcount Poverty headcount Poverty headcount ratio at $1 a day ratio at $2 a day ratio at $1 a day ratio at $2 a day N = 35 N = 35 N = 35 N = 35 Models 1 and 2 Constant −0.1985 (−0.15) −0.0515 (−0.32) −0.2293 (−0.17) −0.0563 (−0.34) DIFFi j
−0.0500 (−0.08) −0.6645 (−1.18)† −0.0018 (−0.00) −0.6174 (−1.02)‡
Models 3 and 4 Constant 0.0242 (0.23)
0.0063 (0.06)
0.0392 (0.35)
0.0188 (0.17)
DIFFi j
2.1626 (1.64)
2.6268 (1.52)
1.9071 (1.36)
2.5824 (1.58)
t Statistics shown in parentheses based on bootstrapped standard errors obtained from 100 bootstrap replications *** Significant at the 0.01 level; ** Significant at the 0.05 level; † Significant at the 0.25 level; ‡ Significant at the 0.30 level
Table 12 Sum-difference test results for poverty variation (models 1 and 2) and financial development variation (models 3 and 4) for summary measures of financial development (1980s–1990s) Summary measure 1
Summary measure 2
Poverty headcount Poverty headcount Poverty headcount Poverty headcount ratio at $1 a day ratio at $2 a day ratio at $1 a day ratio at $2 a day N = 35 N = 35 N = 35 N = 35 Models 1 and 2 Constant 0.6504 (0.48)
0.0191 (0.15)
0.6607 (0.47)
0.0182 (0.14)
DIFFi j
0.0084 (0.01)
1.4171 (1.20)
0.2513 (0.31)
−0.0221(−0.15)
0.0462 (0.37)
−0.0221(−0.15)
1.3920 (0.88)
3.2240 (2.33)
1.3920 (0.88)
1.7456 (1.37)
Models 3 and 4 Constant 0.0554 (0.44) DIFFi j
3.2584 (2.79)
t Statistics shown in parentheses based on bootstrapped standard errors obtained from 100 bootstrap replications *** Significant at the 0.01 level; ** Significant at the 0.05 level; † Significant at the 0.25 level; ‡ Significant at the 0.30 level
of living of the poor and the negative impact of financial instability on the income of this sector of the population. Nevertheless, they also confirm that the link between the mean income of the poor and the credit indicators (level and instability) related to Private credit/GDP is not significant. This suggests that in developing countries an increase in the private credit ratio does not necessarily translate into improved wellbeing for the poor. In consequence, these authors conclude that access to credit for the poor remains a challenge in the developing world and that the main channel for the impact of financial development on this segment of the population is the McKinnon conduit effect captured by the liquidity ratio. It is worth remembering that the so-called conduit effect presented by McKinnon (1973) assumes that even if financial institutions do not provide credit to the poor, who must self-finance investment, they are useful because they offer profitable financial opportunities for savings. McKinnon’s conduit effect must be especially taken
123
76
S. Perez-Moreno
into account in financially underdeveloped economies and environments with credit constraints. In this line, our results suggest that in developing countries, financial development can favor the population in a situation of moderate poverty as progress in financial intermediation provides genuine opportunities for their savings, facilitating investments in physical and human capital in environments with high liquidity constraints. By contrast, financial development in the developing world does not seem to necessarily involve better access to credit for poor people, in accordance with the private credit ratio used, even for the segment of the poor with a certain level of income/expenditure. Other authors, however, have obtained results somewhat different from ours. Thus, for example, Beck et al. (2007) prove that financial development has a significant impact on poverty reduction by using a private credit ratio. A key explanatory factor might be the composition of the sample, as theirs comprises both developing and developed countries (advanced economies have bigger and more diverse financial systems), while our sample, as well as that of Guillaumont and Kpodar (2008), only consists of developing economies. In turn, Ang (2010) observes that financial development helps reduce income inequality (he does not examine poverty) and emphasizes that the results for the case of India are robust to the use of different measures of financial development. In any event, as the financial sector plays a multidimensional role in the process of development and financial sector development involves the combination of many activities and institutions, we must bear in mind that the traditional measures used in this article are only proxies that enable capturing some aspects of financial development. Moreover, Honohan (2008) even questions the correlation of these commonly used indicators with key issues as household access to bank accounts. In this sense, it would be very convenient, to the extent permitted by data availability, to also consider other measures particularly suitable for developing nations, such as the number of bank branches per inhabitant, which can be an effective proxy for the accessibility of banking services and very relevant for the rural areas where most poor people live in the majority of developing countries.
5 Conclusions The study of the linkages between financial development and poverty has received relatively limited attention in development economics literature. Nonetheless, the views and theoretical and empirical conclusions reached by development researchers are considerably varied and, on occasions, contradictory. In this article, we carry out a causal analysis of the relationship between financial development and poverty in developing countries, in which we apply a modified form of traditional Granger causality tests to suit the short times series that are available. In order to assess the possible causal relationships, we use panel data model evaluation techniques to test the out-of-sample forecasting performance of competing models rather than relying on in-sample fit or coefficient estimates. Among the conclusions obtained from our analysis, we can highlight the following:
123
Financial development and poverty in developing countries
77
(1) The results mainly depend on the nature of the financial development indicator used. In particular, when financial development is measured by the liquid assets of the financial system as a share of GDP or the ratio of money and quasi money to GDP, forecasts of future poverty variations are significantly improved when information on past financial development variations is included. (2) When financial development is measured by the value of credits granted by financial intermediaries to the private sector as a share of GDP, the empirical evidence shows that there is no causal link from financial development to poverty. (3) The detected causal link only occurs in the first period analyzed, the 1970s– 1980s, but not in the second one, the 1980s–1990s. It highlights that a general conclusion on such a link in developing countries cannot be adopted as a rule, since the causality relationship examined is influenced by the particular historical context and the economic, political and social circumstances existing in each period. (4) The mentioned relationship is especially significant between financial development and moderate poverty rather than extreme poverty. It indicates that financial development does not primarily benefit the poorest, but instead poor people with a certain level of income/expenditure. (5) The use of summary measures of financial development (the geometric mean of M3/GDP and Private credit/GDP, and of M2/GDP and Private credit/GDP) seems to strengthen the conclusion that financial development entails moderate poverty reduction in the 1970s–1980s period, though in this case the statistical significance is reduced. (6) According to our analysis, there is no evidence of Granger causality from poverty to financial development, since information on poverty variations does not in any case improve forecasts of future financial development variations. In short, in the period of the 1970s–1980s, the empirical evidence supports the hypothesis that financial development leads to the reduction of moderate poverty, whereas we do not find that the level of financial development is altered as a consequence of changes in poverty. In other words, the results indicate that financial development, particularly measured by the ratios M3/GDP or M2/GDP, may cause unidirectional moderate poverty reduction, in a Granger-causal fashion. These findings involve significant implications for development policies. As financial development seems to contribute to the reduction of moderate poverty, we can conclude that financial development may not only be a pro-growth action, but also pro-poor. Our empirical analysis suggests that a key element that explains the beneficial effect of financial development on the moderately poor is the improvement in savings opportunities, which can help them face their liquidity constraints and increase their physical and human capital investments. Moreover, the poor can gradually start improving their access to credit as well, because of savings accumulation and greater credit availability in the financial system. The former considerations are in tune with the political recommendation of liberalizing the interest rate in developing countries so that it can be determined by market forces. This process of financial liberalization often involves an increase of real interest rates, as a result of removing ceilings on nominal rates or of reducing inflation. Higher real interest rates are supposed to raise real savings and financial forms of saving,
123
78
S. Perez-Moreno
specifically savings in bank deposits (Ruehl and Hakimian 2002). Therefore, we can state that establishing a process of effective liberalization of the interest rate—with a proper regulatory and supervisory system—and improved savings mobilization—supported by an adequate geographical distribution of the financial system (Honohan and Beck 2007)—may contribute to alleviate poverty in developing countries, particularly in economies with a poorly developed financial sector and severe credit constraints for the poor. In any case, in order to advance substantially in the fight against poverty and in favor of economic growth, it is necessary to take into account that said financial reforms must be coordinated with other development policies. In fact, empirical evidence shows that development is characterized by interconnected economic, social, political, cultural, religious and ideological factors, and it requires the implementation of a wide variety of economic and social policies that promote a consistent macroeconomic framework and an equity-enhancing productive development strategy which enable less-developed countries to access and ascend on the ladder of development.
Appendix See Table 13. Table 13 Countries by region
123
Country
Region
Barbados
Latin America and Caribbean
Burkina Faso
Sub-Saharan Africa
Burundi
Sub-Saharan Africa
Colombia
Latin America and Caribbean
Costa Rica
Latin America and Caribbean
Cote d’Ivoire
Sub-Saharan Africa
Dominican Republic
Latin America and Caribbean
Ecuador
Latin America and Caribbean
El Salvador
Latin America and Caribbean
Gabon
Sub-Saharan Africa
Gambia, The
Sub-Saharan Africa
Ghana
Sub-Saharan Africa
Guatemala
Latin America and Caribbean
Honduras
Latin America and Caribbean
India
South Asia
Jamaica
Latin America and Caribbean
Kenya
Sub-Saharan Africa
Madagascar
Sub-Saharan Africa
Malaysia
East Asia and Pacific
Financial development and poverty in developing countries Table 13 continued
79
Mexico
Latin America and Caribbean
Nepal
SouthAsia
Niger
Sub-Saharan Africa
Nigeria
Sub-Saharan Africa
Pakistan
South Asia
Panama
Latin America and Caribbean
Paraguay
Latin America and Caribbean
Philippines
East Asia and Pacific
Rwanda
Sub-Saharan Africa
Senegal
Sub-Saharan Africa
Sierra Leone
Sub-Saharan Africa
South Africa
Sub-Saharan Africa
Sri Lanka
South Asia
Thailand
East Asia and Pacific
Trinidad and Tobago
Latin America and Caribbean
Venezuela
Latin America and Caribbean
Acknowledgements The author gratefully thanks two anonymous referees for their useful comments, as well as Dr. Diana Weinhold, the Development Studies Institute (London School of Economics and Political Science) and the Centre for Financial and Management Studies (SOAS, University of London) for the support received.
References Adams RH (2004) Economic growth, inequality and poverty: estimating the growth elasticity of poverty. World Dev 32(12):1989–2014 Aghion P, Bolton P (1997) A theory of trickle-down growth and development. Rev Econ Stud 64:151–172 Ang JB (2010) Finance and inequality: the case of India. South Econ J 76(3):738–761 Ang JB, McKibbin WJ (2007) Financial liberalization, financial sector development and growth: evidence from Malaysia. J Dev Econ 84(1):215–233 Arestis P, Demetriades PO (1997) Financial development and economic growth: assessing the evidence. Econ J 107:783–799 Arestis P, Demetriades PO, Luintel KB (2001) Financial development and economic growth: the role of stock markets. J Money Credit Bank 33:16–41 Banerjee AV, Newman AF (1993) Occupational choice and the process of development. J Political Econ 101:274–298 Beck T, Demirgüç-Kunt A, Levine R (2000) A new database on the structure and development of the financial sector. World Bank Econ Rev 14:597–605 Beck T, Demirgüç-Kunt A, Levine R (2007) Finance, inequality and the poor. J Econ Growth 12(1):27–49 Calderón C, Liu L (2003) The direction of causality between financial development and economic growth. J Dev Econ 72:321–334 Christopoulos DK, Tsionas EG (2004) Financial development and economic growth: evidence from panel unit root and cointegration tests. J Dev Econ 73:55–74 Clarke GRG, Xu LC, Zou HF (2006) Finance and income inequality: what do the data tell us?. South Econ J 72(3):578–596 Demetriades PO, Hussein KA (1996) Does financial development cause economic growth? Time-series evidence from sixteen countries. Journal of Development Economics 51:387–411 Dollar D, Kray A (2002) Growth is good for the poor. J Econ Growth 7:195–225 Foster J, Szekely M (2000) How good is growth. Asian Dev Rev 18:59–73
123
80
S. Perez-Moreno
Galor O, Moav O (2004) From physical to human capital accumulation: inequality and the process of development. Rev Econ Stud 71:1001–1026 Galor O, Zeira J (1993) Income distribution and macroeconomics. Rev Econ Stud 60:35–52 Goldsmith RW (1969) Financial Structure and development. Yale University Press, New Haven Granger CWJ (1969) Investigating causal relations by econometric models and cross-spectral methods. Econometrica 37:424–438 Granger CWJ, Huang LL (1997) Evaluation of panel data models: some suggestions from time series. Discussion Paper 97-10. Department of Economics, University of California, San Diego Greenwood J, Jovanovic B (1990) Financial development, growth, and the distribution of income. J Political Econ 98:1076–1107 Guillaumont SJ, Kpodar K (2008) Financial development and poverty reduction: can there be a benefit without a cost? IMF Working Papers No. 08/62. International Monetary Fund Gupta KL (1984) Finance and Economic growth in developing countries. Croom Helm, London Honohan P (2004) Financial development, growth and poverty: how close are the links. World Bank Policy Research Working Paper No. 3203. World Bank Honohan P (2008) Cross-country variation in household access to financial services. J Bank Finance 32:2493–2500 Honohan P, Beck T (2007) Making finance work for Africa. World Bank, Washington Jung WS (1986) Financial development and economic growth: international evidence. Econ Dev Cult Change 34:333–346 Levine R (1997) Financial development and economic growth: views and agenda. J Econ Lit 35:688–726 Levine R, Loayza N, Beck T (2000) Financial intermediation and growth: causality and causes. J Monetary Econ 46:31–77 Lustig N, Arias O, Rigolini J (2002) Poverty reduction and economic growth: a two-way causality. Sustainable development department technical paper series. Inter-American Development Bank, Washington, DC McKinnon RI (1973) Money and capital in economic development. Brookings Institution, Washington, DC Perez-Moreno S (2009) An assessment of the causal relationship between growth and inequality in Spanish regions. Eur Plan Stud 17(3):389–400 Rajan RG, Zingales L (2003) Saving capitalism from the capitalists. Crown Business, New York Ravallion M, Chen S (1997) What can new survey data tell us about recent changes in distribution and poverty?. World Bank Econ Rev 11(2):357–382 Ruehl S, Hakimian H (2002) Development finance, principles and experiences. The Centre for Financial and Management Studies, University of London, London Sala-i-Martin X (2002) The world distribution of income (estimated from individual country distributions). NBER Working Paper No. 8933 Schumpeter JA (1911) The theory of economic development. Harvard University Press, Cambridge Shaw ES (1973) Financial Deepening in economic development. Oxford University Press, New York Weinhold D, Reis E (2001) Model evaluation and testing for causality in short panels: the case of infrastructure provision and population growth in the Brazilian Amazon. J Reg Sci 41(4):639–657 World Bank (2007) World development indicators 2007. World Bank, Washington
123
