The Difference Between Growth and Development

The link between foreign capital, governance, and development is complex and somewhat ambiguous. A first step toward uncovering it is to distinguish between “development” and “growth.” These two concepts are often used interchangeably; yet, while intimately related, they are distinct. Economic growth—increasing per-capita GDP—is of course a vital part of economic development. But economic growth itself is merely a measure of capacity; the extra money on its own does nothing to guarantee that a population is less impoverished, more educated, healthier, or, indeed, that the economy is in a better position to grow any further. To get closer to such guarantees, it is necessary to know, first, how the extra money is distributed, and, second, how it is used. A poor country with a growing economy may still develop little if the growth merely enriches a small élite, leaving the majority of the population without additional income. Nor does a poor country develop as much when its income is spent on, say, arms imports, rather than on public goods. The distinction between growth and development has a long history in development research; Sen gives one of its most articulate rationales: “It is as important to recognize the crucial role of wealth in determining living conditions and quality of life as it is to understand the qualified and contingent nature of this relationship. An adequate conception of development must go much beyond the accumulation of wealth and the growth of gross national product and other income-related variables.”⁵

Throughout this article, we use the concept of “human development” to distinguish between growth and development more generally. The concept of human development was established in the United Nation Development Programme's Human Development Reports and is meant to provide a more complete picture of development than economic growth. It includes not only income, but also measures of human capital—health and education—the tools that a person needs both to live a successful life in the modern world and to contribute to a country's economic progress.⁶

A distinction between growth and human development naturally produces a question: what is the relationship between the two? For an answer we turn to work by Ranis, Stewart, and Ramirez.⁷

From growth to human development

From human development to growth, and from growth to growth

human development contributes to economic growth by increasing the capacity of the workforce, which in turn alters the organization and adaptability of production and the range and complexity of economic output. This contributes to increasing national income through, among other things, social capital and the policy environment.

In turn, economic growth can contribute to human development by directly increasing government revenue, which, depending on government priorities, may then be reinvested further in human development. Growth can also contribute to human development by increasing household income (though the degree of the increase will depend on income distribution); households will then allocate a certain amount to investing in their own health, education, and welfare, according (as with the government) to their spending priorities.⁸

Lastly, economic growth can contribute to further economic growth directly, entirely bypassing human development. Increasing income can increase savings, which in turn leads to increases in the capital stock. Investment in the capital stock increases productivity, and thereby contributes to further growth. (But as mentioned above, this type of growth—growth absent progress in human development—is likely to lead a country into a vicious cycle of low human development leading to reduced growth.)

The contributors to economic growth and human development are laid out in Figures 1 and 2.⁹

External Funding for Development

In Figures 1 and 2, the only source of resources for use in development is domestic: economic growth. But developing countries often access a good deal of external funding, from states (official aid), and from private sources (individual remittances; private donations; commercial bank loans; and private foreign direct investment). Our concern is with aid and FDI, the two most prominent sources of external funding for economic growth and human development in developing countries.

While aid flows have remained largely stagnant in recent years, flows of FDI increased rapidly through 2000 (Figure 3). In the five years from 1995 to 2000, FDI in low- and middle-income countries grew at an average annual rate of 17 percent. While total flows have decreased in recent years, FDI continues to be the largest and most stable source of external finance for developing countries, exceeding by far the sum of commercial bank loans and official flows.¹⁰

Inflows of aid and FDI, 1970–2001

Aid—by which we mean Official Development Assistance (ODA)—enters the picture we drew in the first section directly as government revenue. Despite donors' best efforts to ensure that aid is spent according to donor, not recipient, priorities, the empirical evidence is overwhelming that, at the margin, aid is little different from any other source of revenue.¹²

The implication of this is that aid is likely to affect human development (and, indirectly, economic growth) via government spending priorities. In countries whose governments give high priority to human development, aid is likely to add to human development, for it will add to spending on human development. Where governments give human development a low priority, aid may just as easily do nothing for human development. In the extreme, aid may even impede human development, if given to a country whose government prioritizes things that work against human development—for example, a brutal dictatorship whose ruler spends the nation's wealth on tools of repression to maintain his hold on power. Figure 4 shows the possible linkages between aid and human development.

From aid to human development

By contrast to government-centered aid, FDI is by nature private. Therefore, it enters our picture in a way quite different from aid; primarily, FDI is simply capital. Purchases of equity capital and reinvestment of profits by foreign investors act in the same way as domestic savings, by adding to the overall supply available to fund new investment. Insofar as these inflows lead to the creation of new fixed assets and/or the use of better technology, they may also increase the efficiency and productive capacity of a country's economy.¹⁵

FDI may also increase growth indirectly, through positive externalities. Foreign investors may introduce more-advanced technology and management practices; these may then spill over to domestic firms as they observe foreign-firm practices, or as labor—especially skilled labor or management personnel—moves between the two. In a similar way, domestic firms can absorb technical skills and quality-control techniques. Competition from foreign investors can also force domestic firms to increase their efficiency and their use of technology to keep pace. These externalities all may affect the organization and adaptability of production, and thereby play a role in increasing growth.¹⁶

FDI may contribute to human development too, if it is able to increase household income or tax revenue. If so, FDI may work in much the same way as aid: by increasing the resources on which either governments or households may draw for their human-development spending.

Yet FDI can also work against a country's development. Any of the channels by which FDI contributes to either growth or human development depends, first, on the sort of FDI a country receives, and, second, on the ability of the country to absorb the investment's benefits. Both will depend on human capital, labor and wage standards, and existing technology in the host country.¹⁷

Specifically, if a country attracts FDI only by granting foreign firms special incentives, FDI can cause serious economic distortions. For example, FDI may increase competition and thereby force domestic firms to become more efficient; but if foreign firms are too disproportionately advantaged, domestic firms may find it impossible to compete, and the FDI may result in a loss of indigenous enterprise. Along the same lines, if FDI is attracted only through tax incentives, then, far from raising government revenue, FDI may actually reduce it.

Another possibility is that, if FDI is heavily subsidized, domestic investors, crowded out by the FDI, may simply pretend to be outsiders by sending funds out of the country and then bringing them back so as to benefit from the subsidy for foreign firms.¹⁸

In addition, foreign investment in industries that serve protected domestic markets may ally foreign and domestic investors in pushing for continuing distortionary trade and investment policies. Lastly, heavy subsidies may lead to fierce competition between developing countries to attract foreign investors, with little or no aggregate benefit to the world economy, and the resulting proverbial “race to the bottom” among developing countries.

Even if not attracted through subsidies and incentives, FDI may not bring benefits to a country. Foreign investors will often repatriate funds close or equal to the amount they originally brought into the country, and foreign firms will frequently invest by borrowing heavily in the domestic credit market, so that the actual amount of FDI flows overstates the amount of incoming capital.¹⁹

The existing empirical work has only begun to sort out these complexities, and so far the evidence on the benefits and costs of FDI is mixed. In particular, no one is sure whether FDI increases or decreases household income, or what it does to tax revenues.²¹

From FDI to growth and human development

The analysis in this section points to a simple conclusion: aid should primarily affect human development; FDI may affect either growth or human development. But these effects are likely to be contingent: the effect of aid will depend on how a country's government spends its revenue, and the effect of FDI will depend on the type of FDI entering the economy and how well-equipped a country's economy is to harness its potential spillovers and make use of any extra capital it brings. We next examine each of these contingent relationships.

Aid and Government Preferences

Our analysis so far leads us to believe that aid will operate directly through the government. Thus to know aid's effect it is necessary to know government preferences on human-development expenditure.

For a variety of reasons, we believe that these preferences will be related to the level of democracy. The tendency of democracies to spend more on human development than autocracies is well-established.²²

Thus our intuition is that, all else equal, democracies will invest more in human development than autocracies. Since aid operates directly through government preferences, democratic governance should be an important determinant of whether aid is spent in a way that drives human development. Recently several authors have found that aid usage does seem to be contingent on the level of democracy. Svensson shows that aid given to democracies increases growth, while aid given to autocracies does not; Kosack finds that aid to democracies hastens human development.²⁷

FDI and Human Capital

In contrast to aid, FDI's impact on development will not depend directly on government policy preferences. As with aid, FDI may contribute to human development by increasing government revenue, but this is a tenuous channel. FDI's primary effect is through the market: the capital stock, household income, and technological spillovers. It is the character of the economy that will be important in determining how FDI affects development.²⁸

With FDI, then, the question is: what sort of economy is likely to both attract the right sort of FDI, and allow that FDI to have a positive aggregate effect on growth and development? This sort of economy will be one that already has a high level of human capital. The reason is straightforward. As we discussed above, we know from past research that whether countries win or lose from FDI will depend, among other things, on human capital. Human capital will affect the nature and type of FDI that a country attracts, and to what extent the foreign investment augments or replaces domestic sources. In addition, the beneficial externalities from FDI, discussed earlier, are likely to be highly dependent on the level of human capital in a country. For example, if a country has a high enough level of human capital, it may be able to absorb a new technology brought by a foreign investor, where a country with a low level of human capital may not benefit at all from the same technology.²⁹

Hypotheses

The implications of our theory are clear: aid should stimulate or depress human development; in turn, it may affect growth by adding to workforce capacity. FDI may directly affect both growth and human development. But our theory also suggests that the effects of both aid and FDI should be contingent: aid's on government human-development policy preferences; FDI's on the existing level of human capital.

Our theory also gives us some insight into whether these effects should be positive or negative. Our hypotheses are the following:

H1: The rate of human development will be positively associated with the democratic accountability of the government.

H2: The greater the commitment of a country's government to human development, the more aid given to that country will accelerate the rate of human development.

Corollary: Where aid accelerates human development, growth should increase as well.

H3: In countries with a high level of human capital, FDI should accelerate growth and human development.

In the following sections, we find considerable support for our Hypothesis 1 on democracy: in most countries, the more democratic a country's government, the faster is its rate of human development. But the relationship between democracy and human development decreases with development: it is strongest in the least-developed countries; weaker in more-developed countries; and may actual impede human development in the most-developed countries. We also find considerable support for our Hypothesis 2 on aid, but despite our confirmation of Hypothesis 1, our results show that the level of human capital, not democratic accountability, directly determines aid's effectiveness: in countries with extensive human capital, aid hastens human development and growth, but in countries with limited human capital aid has no effect on growth and can actually decelerate human development. Lastly, we find partial support for our Hypothesis 3 on FDI: FDI never increases growth or accelerates human development, but can decelerate human development in countries with limited human capital.

In the following sections we first specify the empirical model we use to test these hypotheses, and then discuss the results of our estimation.

Models of Growth and Human Development

To test our hypotheses we need to estimate models of the determinants of economic and human-development growth. Theories of growth have emphasized a range of determinants of growth, including capital accumulation, human capital, research, development and innovation, infrastructure, management, and organization.³¹

Our base model takes the following form: Economic growth (y) or the rate of human development (h) depend on the log of the average level of income (i), the level of human capital (c), the level of democracy (p), aid receipts relative to GDP (a), FDI relative to GDP (f), various other exogenous variables that may affect y or h (z′), fixed time (τ) and country (η) effects for y or h, and an error term (ε).

Each of the variables is indexed by country (i) and time (t). The vector z′ contains six variables that may affect y or h:

• inflation and openness (the ratio of exports plus imports to GDP): proxies for sound economic policy;
• arms imports (as percentage of total imports): the proportion of international trade devoted to repression or war-making capabilities;
• gross national savings: a proxy for domestic investment;
• natural-resource exports (as a percentage of total exports): a proxy for the sophistication of the economy; and
• battle deaths: a proxy for the human costs of war.

In these regressions we measure h, the rate of human development, with growth in the United Nation Development Programme's (UNDP) Human Development Index (HDI). The HDI consists of three elements: life expectancy; knowledge (two-thirds literacy and one-third combined primary, secondary, and tertiary education enrollment); and wealth. The variable c—human capital—is simply the index of education and health from the HDI; it therefore differs from the HDI only in that it does not include income.

Contributors to human development—schools, hospitals, better-trained teachers and doctors, and so on—do not have an instantaneous effect on a population's level of human development, and will therefore take some time to show up in a country's score on the HDI. To account for this lag, the dependent variable h in equation (2) is the rate of human development in the period t + 1. For the same reason, any effect that aid will have on growth is likely to be delayed, since the only way that aid can affect growth is via human development (see Figure 4). Therefore, we also lag the variable a in equation (1).

All variables are described in detail in the section below on data.

Equations (1) and (2) are useful for examining the independent effects of aid and FDI on economic growth and human development, but our theory leads us to believe that these effects may be conditional. To examine this possibility, we next estimate our base models with a series of interactions.

In the section above on aid, we concluded that governments with higher preferences for human-development spending would tend to be more democratic. Yet when we interacted aid with the level of democracy, it was never significant. Because the level of democracy exerts an independent effect on the rate of human development, democracy does play a role, albeit an indirect one, in aid's effectiveness. But clearly democratic accountability is not the only way of increasing the priority governments put on human development. One observable implication of a government with a strong commitment to human development is, of course, higher human capital; therefore, human capital is one proxy for a government's human development priorities. It is this interaction that we include in our model.³³

Equations (3) and (4) account for the possibility of contingency in the effects of aid, FDI, and democracy:

These equations are identical to equations (1) and (2), but each contain three interactions: aid × human capital, fdi × human capital, and democracy × human capital.

Initially, we estimated equations (1) through (4) using ordinary least squares (OLS), but immediately we ran into a problem: aid, FDI, and human capital may all be endogenously determined. We may, for example, expect that countries with lower levels of income and human capital should receive more aid. Similarly, we may expect that FDI is attracted to countries with higher growth (though not necessarily to countries with faster human development). Lastly, we know from our theoretical model that growth may add to human capital if it adds to government revenue and/or household income. It also seems reasonable that faster human development will be associated with lower human capital, as certain elements of human development may be progressively harder to achieve (for example, it might be easier to move from literacy of 50 to 60 percent than from 90 to 100 percent).³⁴

Our data support these fears. Table 1 shows average values of aid, FDI, and human capital for each quartile of our dependent variables: economic growth and the rate of human development. We see from the table that aid goes disproportionately to countries with faster human development, though not to countries with faster economic growth. FDI flows disproportionately to countries with faster economic growth, though not to countries with faster human development. Finally, countries with the most limited human capital seem to have the fastest human development, and the slowest economic growth.

Average values of aid, FDI, and human capital

As a result, it is likely that OLS estimations of equations (1) and (2) are biased and inefficient. Political scientists typically deal with endogeneity in panel data with two-staged least squares-instrumental variable estimation (2SLS-IV). Initially we took this approach, but it too proved insufficient. First, the instruments available in the literatures on aid, FDI, and human capital are all weak and of questionable validity, and we were unable to construct more reliable instruments. Second, 2SLS-IV does not deal with the very real possibility that our model is serially correlated.

Thus instead of 2SLS-IV, we decided to use a panel estimator proposed by Arellano and Bond, and updated and extended by Arellano and Bover, and Blundell and Bond.³⁵

In the way it is applied by Arellano and Bond, GMM also allows us to control for endogeneity. Arellano and Bond's GMM estimator uses moment conditions to derive a set of valid instruments for our endogenous variables. The Arellano-Bond GMM technique therefore allows us to control for endogeneity in our explanatory variables, as well as to address the possibility of serial correlation.

The Arellano-Bond GMM technique has a number of other benefits. As with 2SLS-IV, it allows us to control for the fixed effects of time and of each country on our parameters, leaving only effects that are true across countries and across time. This frees our results from the effects of country-specific factors—for example, that Bolivia is land-locked—and time-specific factors—that, for example, there was an oil crisis in the 1970s.³⁷

Readers familiar with the A-B technique will note that it was originally designed to estimate models that, unlike ours, include a lagged dependent variable as an independent variable. We do not include a lagged dependent variable because there is little theoretical justification for it—while growth or human development may persist year-to-year, there is not likely to be persistence across five-year periods—and Achen notes that including a lagged dependent variable without theoretical justification can artificially dominate a regression regardless of the variable's explanatory power.³⁹

The Arellano-Bond GMM technique

Because the Arellano-Bond GMM technique is relatively new to political science, and because the verisimilitude of our results rests heavily on the validity of the technique, it is worth exploring further precisely how GMM operates in our case. Readers already familiar with the Arellano-Bond GMM technique may, of course, skip to the next section.

To start, we can rewrite our equations (1) and (2) as

where y is our dependent variable (economic growth or the rate of human development); x is the vector of explanatory variables defined above; η represents fixed country-specific effects; τ represents fixed time-specific effects; and ε is a time-varying error term. Subscript t indexes for time period, and i for country. For simplicity, we omit τ_i, our six time-period dummies, from the following equations.

The next step is to eliminate fixed country-specific effects (leaving only effects that are true across countries). To do so, we first-difference equation (1a):⁴¹

Arellano and Bond 1991.

As we see in equation (1b), first-differencing eliminates country-specific effects (η). But the first-differenced equation introduces a new form of bias, because the error term (ε_i,t − ε_i,t−1) may now be correlated with the endogenous variables, which is to say that the expected value of the error term given the set of independent variables is now no longer equal to zero (E [Δε_it /Δx_i,t] = E [ε_i,t − ε_i,t−1 /x_i,t − x_i,t−1] ≠ 0).

To correct this problem, Anderson and Hsiao propose instrumenting for the endogenous variables with the second lagged difference of the dependent variable.⁴²

Anderson and Hsiao 1982.

Once this is done, the change in y (Δy_it−2 = y_it−2 − y_it−3) is correlated with the dependent variable (y_i,t − y_i,t−1) but uncorrelated with the error term (ε_i,t − ε_i,t−1). Thus the Anderson-Hsiao estimator is consistent. For example, at t = 3, equation (1b) would take the following form:

and y_i,3 is a valid instrument for y_i,3 − y_i,2(Δy_i,2), because it is highly correlated with Δy_i,2 but uncorrelated with (ε_i,3 − ε_i,2). At any time t, the valid instrument set is (y_i,1,y_i,2,…,y_i,t−2).⁴³

When lagged instruments are correlated with the fixed effects, future values of x_i,t will be correlated with the current error term. These lagged instruments will only be valid instruments in periods prior to the current period, but not in the current period or any future period. But strictly exogenous instruments are uncorrelated with the fixed effects and so are valid instruments in every period.

However, Arellano shows that using the lagged difference as an instrument results in an estimator that has a very large variance, because Δy_i,t is a linear combination of y_i.⁴⁴

Arellano 1989.

So, while consistent, the Anderson-Hsiao estimator is inefficient. To increase the efficiency of the estimator, Arellano and Bond propose using additional instruments based on all of the available moment restrictions.⁴⁵

Arellano and Bond 1991.

This is possible with a GMM estimator. The estimator uses the sample moments that restrict the covariance between the regressor and the error to zero. Under the standard assumptions that, first, there is no serial correlation in the error term (ε_i,t − ε_i,t−1) and, second, the explanatory variables are weakly exogenous, we can use the following moment restrictions:

With these restrictions, we can create (T − 2)(T − 1)/2 moment restrictions for T ≥ 3 and then create an estimator based on these restrictions. Arellano and Bond and Kiviet use simulation methods to prove that this method—“the GMM first-difference estimator”—is more efficient than the traditional Anderson–Hsiao method outlined above.⁴⁶

See Arellano and Bond 1991; Kiviet 1995.

Here we do not derive the estimator; for that, see Bond, or Arellano and Bond.⁴⁷

See Bond 2002; Arellano and Bond 1991.

Adding additional instruments

As we noted above, the GMM first-difference technique already instruments for all the explanatory variables with their lagged values. Blundell and Bond show through myriad Monte-Carlo simulations that in the presence of weak instruments (instruments that are only weakly correlated with the endogenous variables) the GMM first-difference technique can be severely biased. But we can reduce the possibility of bias by including additional instruments for variables that may be endogenous.⁴⁸

Blundell and Bond 1998.

We discuss our specific instruments in the next section, but first we describe how the model uses them.

Arellano and Bover, and Blundell and Bond show that it is possible to include additional instruments when at least part of the vector of explanatory variables is uncorrelated with fixed country and time effects.⁴⁹

See Arellano and Bover 1995; Blundell and Bond 1998.

We can add these additional instruments to our model alongside the lagged instruments (those calculated from the moment restrictions). Arellano and Bover and Blundell and Bond show that we can combine into a single system a regression in differences (using lagged values as instruments) and a regression in levels (using the additional instruments) with a different vector of instruments for each type of regression.⁵⁰

Ibid.

Blundell and Bond show that this system estimation technique minimizes the problem of weak instruments to the point that we are able to use common tests of over-identifying restrictions to test the validity of our instruments.⁵¹

Blundell and Bond 1998.

This is the technique we use in our regressions, and we will refer to it as “system GMM.”⁵²

Bond 2002 offers an excellent overview of all variations of the estimator.

The equations take the form:

where the instruments for the regression in differences (x_i,t²) are lagged levels of the explanatory variables, while the instruments for the regression in levels (x_i,t′²) are the lagged differences of the explanatory variables.

Put another way: endogenous explanatory variables in the level equations are instrumented with lags of their own first differences plus any additional instruments.⁵³

See Blundell and Bond 1998.

Arellano and Bover and Blundell and Bond show that by estimating the two equations in a single system we reduce the potential bias and imprecision associated with the traditional first-difference GMM estimator.⁵⁴

See Arellano and Bover 1995; Blundell and Bond 1998.

Variables and Data Sources

Our data cover an unbalanced panel of a maximum of 103 countries—both developed and developing—from 1970 to 1999. To smooth year-to-year variation, we average the data into six five-year periods. Data on all variables are not available for all periods, however; in the end our available observations range from a maximum of 363 over 90 countries to a minimum of 236 over 77 countries.⁶⁰

Regressions

Base models

We begin by estimating our base equations (1) and (2) by system GMM with and without the vector of control variables z′. The results are in Table 2. At this point, FDI and the level of democracy both seem to have no relationship with economic growth, but aid has a strong (positive) relationship once we include controls—columns (1) and (2). Aid also seems to have a strong positive relationship with the rate of human development, but only until we add our vector of controls. The level of democracy, too, seems in column (3) to have a weak, negative association with human development, but this relationship fades once we include control variables—column (4). As with economic growth, FDI is unassociated with the rate of human development.⁶⁶

To ensure that the change in our results between those equations with and those without controls are the result of selection bias, not the change in sample size, we reran all of our equations, limiting the sample size to the model with the fewest observations (N = 269). None of our results changed substantively.

Equations (1) and (2) by system GMM

Accounting for contingency

As noted at the beginning of this article, we strongly suspect that the effects of both FDI and aid may be contingent: aid's rests on government human-development policy preferences; FDI's, on the existing level of human capital. Thus estimations of equations (3) and (4), which account for contingency through a series of interactions, are the most important evidence of the theoretical arguments in this article.

Columns (5) and (6) of Table 3 show estimates by system GMM of equation (3) (dependent variable: growth). As before, we first estimate the equation without controls, in column (5), then with controls, in column (6). In columns (5) and (6) neither FDI nor the level of democracy show a contingent relationship with economic growth (or any relationship at all). But there does indeed seem to be some contingency in the impact of aid, though one that at first glance seems to disappear when we add our controls—column (6).

Equations (3) and (4) by system GMM

It is always risky to put too much trust in coefficient estimates that are at best rough, but there is still value in examining the magnitude of effects, especially when they are interactive. Table 4 helps us to make sense of the interactions; it shows the conditional effect of each variable (aid, FDI, or the level of democracy) at various levels of c (human capital), using unique coefficient estimates and standard errors for each level of c generated from the coefficient estimates and the variance-covariance matrices from the regressions in columns (6) and (8). In column (9) of this table, we can see the interactive effect of aid on growth. Aid appears to have no effect on growth in countries with human capital of lower than 60, but in countries with human capital of 60 or higher its effect is large and positive (60 is roughly the level of Nigeria or Pakistan in the period from 1995 to 1999). The higher a country's level of human capital, the more aid contributes to growth. For example, in a country with a level of human capital of 75 (a level equivalent to Brazil in 1980–84, China in 1975–79, or Colombia or Thailand in 1995–99), an increase in aid of 1 percent of GDP, according to our estimation, would increase expected growth in the following period by 1.22 percent; and in a country with a level of human capital of 85 (a level equivalent to Albania or Mauritius in 1995–99), next-period growth would increase by 1.67 percent.

Conditional effects of aid, FDI, and the level of democracy

Turning back to Table 3, we see in columns (7) and (8) system-GMM estimates of Equation (4) (dependent variable: human development), with and without controls. In these results, aid and the level of democracy show strong associations with the rate of human development both with and without controls; FDI, though, has only a weak association, and only with controls.

Aid's effect on the rate of human development is contingent on human capital: from column (12) of Table 4, we see that aid's effect is large and positive at higher levels of human capital (55 or higher), and that this effect grows larger as the level of human capital increases. However, aid's effect is negligible at lower levels of human capital (50 or below) and turns negative at the lowest levels of human capital (below 30, a level equivalent to Burkina Faso in 1985–89 and Niger in 1995–99). Of course, few countries in our data set have such low levels of human capital (see Table A1 in the Appendix 2), so these precise cut-offs are no more than approximations. If the coefficients in column (12) are taken literally, when given to a hypothetical country with a level of human capital of 60, aid equivalent to 1 percent of GDP would increase the expected rate of human development by 0.19 points per period. To a country with a level of human capital of 75, the same amount of aid would increase the expected rate of human development by 0.37 points per period. But an increase in aid of 1 percent of GDP to a country with a level of human capital of 25—again, a level for which we have little data—would decrease the expected rate of human development by 0.22 points per period.

Turning now to the level of democracy, we see that it too has a contingent effect on the rate of human development: column (14) of Table 4 shows that this effect is positive at mid to low levels of human capital (65 or below), negligible at levels of human capital between 70 and 80, and it is actually slightly negative at the highest levels of human capital (85 or higher). That is, democracy adds to human development, but the more developed the country, the less important is democracy, and democracy may hinder human development in very developed countries. Put another way, the less developed the country, the more its development is helped by democracy.

FDI's negative association with human development contrasts with its lack of any association in column (4) of Table 2. Thus FDI's effect, like aid's, does seem to depend to some extent on the level of human capital in a country. We can see FDI's effect more clearly in column (13) in Table 4: FDI has a negligible effect on human development at levels of human capital above 55 (Bangladesh in 1990–94 or Rwanda in 1995–99); but at 55 and below, FDI has a negative effect on human development, and this effect grows stronger, the lower the level of human capital in the country. For example, an increase in FDI of 1 percent of GDP in a country with a level of human capital of 50 (a level equivalent to Senegal or the Côte d'Ivoire in the period 1995–99) would lower the expected rate of human development by 0.39 points per period.

Table 5 contains a summary of our regressions.

Summary of the regressions

To see our results at work, consider a few examples. If our results reflect accurately the way aid and FDI affect development, we should see that, in a country with a demonstrated commitment to human development, lots of aid is good for development. One of the most striking examples of this is China, whose level of human capital in 1980–84, 77.9, was not much higher than the mean level for that period, 74.1. Ironically China is often a poster case for the benefits of foreign investment, not aid. Yet in the 1980s and early 1990s, despite this reputation, China received a tremendous amount of aid, far more than it received in foreign investment (due mostly to Japan and Germany, who together provided an average of almost $1 billion in aid annually from 1980 to 1999). Only in the later 1990s did FDI overtake aid. Yet China achieved spectacular growth and extremely fast human development over the entire period from 1980 to 1999. Table 6 shows China's aid and FDI receipts alongside its economic growth and the rate of its human development. For Indonesia, another conventional poster case, the situation is similar; it is also shown in Table 6. (Indonesia's level of human capital in 1980–84 was 69.4, slightly below the mean for that period.)

Aid and FDI in the development of China and Indonesia

On the other hand, only a handful of countries have achieved rapid human development with a heavy reliance on FDI, and the vast majority of these countries already had extremely high levels of human capital. In fact, of the countries in our data set that had human capital of less than 85 (Malaysia in the early 1990s) and that received substantial amounts of FDI (more than 3 percent of GDP), only seven—Benin, Botswana, Egypt, Malaysia, Papua New Guinea, Singapore, Swaziland, and Togo—achieved at least the mean rate of human development (around 2 points each period). Of these only Botswana and Swaziland achieved this rate for more than one period,⁶⁷

Botswana achieved this rate for the two periods from 1975–84; Swaziland achieved it for the three periods from 1970–84. Singapore and Malaysia are both close to the threshold level of human capital of 85 when they achieved this rate: Singapore achieved it in the period from 1970–74, whereupon its human capital becomes greater than 85; the same is true of Malaysia, which after one period of rapid human development from 1980–84 had human capital of more than 85.

and Egypt achieved the rate only with the addition of substantial amounts of aid (about 2.5 percent of GDP, mostly from the United States).

The results in this section provide important insight into the way that aid, FDI, and democracy operate. Our theoretical discussion in the beginning of this article provides us with intuition into how democracy and human development interact: the priorities of democratic governments should give more weight to human-development spending than autocratic governments. The tendency of human development to be faster in most of the democratic countries in our data set is probably the result of a higher priority accorded to human development by democratic governments. Yet this tendency is not equally present in all countries: democracy appears to accelerate human development less as a country acquires more human capital, and democracy actually slows human development in countries with the most extensive human capital. Undoubtedly several factors are at work here, the most obvious of which is that countries that already have extensive human capital already possess governments that, for reasons other than democracy, have high human-development spending priorities, so that adding democracy to the mix does not do as much. A more interesting possibility is the one identified by Huntington: that because democracies are unable by their nature to resist public demands, they create waste and inefficiency and so may be inimical to economic development.⁶⁸

Huntington 1968.

In countries with very low levels of human development, these pressures may do a great deal of good by focusing the government's attention on badly needed public goods; but in highly developed countries, Huntington may have been right.

We also know from our theoretical discussion that aid will affect human development via government revenue. Thus in countries with more extensive human capital—where governments are clearly more committed to human development—aid seems to increase government revenue, and, as a result, government spending on human development. In countries with limited human capital, however, government spending on human development may actually decrease with aid. Our results also show that, in countries with more extensive human capital, aid increases growth. The implication is that aid has the ability to help countries into the virtuous cycle from Ranis, Stewart, and Ramirez, where increasing human development reinforces economic growth, and so on.⁶⁹

Ranis, Stewart, and Ramirez 2000.

We showed earlier that FDI affects human development in two ways: through household income and through government revenue. Thus FDI's inability in our results to accelerate human development probably stems from a failure of FDI to increase either household income or government revenue. In countries with limited human capital—where FDI decelerates human development—FDI may even decrease household income (for example, by depressing wages) and government revenue (for example, if the foreign investment is only attracted through generous tax breaks). Similarly, we know that FDI can affect economic growth through two channels: technological spillovers and the capital stock. Thus, because FDI does not seem able to increase economic growth, it is likely that FDI does not in the aggregate increase the capital stock, or produce technological spillovers sufficient to alter the organization and adaptability of production.⁷⁰

It is important to realize that our results might be true only of FDI generally. If we were able to separate FDI into sectors or types, we might see that certain types do, for example, produce significant technological spillovers, or that other types actually reduce the capital stock in a country. It could also be that FDI does affect human development or growth, but only in the long run.