Chapter

4 Sources of Growth

Author(s):
Chorng-Huey Wong, Mohsin Khan, and Saleh Nsouli
Published Date:
April 2002
Share
  • ShareShare
Show Summary Details
Author(s)
Xavier X. Sala-i-Martin

One of the most important questions that economists can ask is why do some countries grow over the long run whereas others do not. After all, a continuous growth rate over an extended period is what brings wealth to a nation. To give an example: in 1870, income per capita in the United States was equal to $2,244 in 1990 dollars. By 1990, income per capita had reached $18,258. Thus, real income grew at an annual rate of 1.75 percent over the course of those 120 years, and real income per capita multiplied by a factor of eight.

This growth allowed the United States to become the richest country in the world by 1990. To see the importance of slight variations in the growth rate over long periods, let us imagine how the United States would look today if, starting from the same $2,244 per capita income in 1870, the real annual growth rate had been 0.75 percent instead of 1.75 percent. Under that scenario, income per capita in 1990 would have been $5,519. In other words, instead of being the richest country in the world, the United States would have had a level of income similar to that of Mexico or Hungary, and a thousand dollars lower than that of Greece or Portugal. A real growth rate of 0.75 percent might seem ridiculously low, but it is similar to the growth rates of India, Pakistan, and the Philippines over a similar period.

Let us imagine now that the annual growth rate had been 2.75 percent instead of 1.75 percent. Real income per capita in 1990 would have been $60,841. We do not really know what an economy with that level of income would look like, since no country in the world has ever reached it, but it is certainly a lot more than the actual $18,258. Yet the difference is the result of only a 1 percentage point higher growth rate over a period of 120 years. A sustained annual growth rate of 2.75 percent may seem unrealistic, but it is exactly the rate experienced by countries like Japan and Taiwan Province of China over a similar period.

This example shows the importance of raising a country’s growth rate over an extended time. In terms of the effects on poverty and on income per capita, the elimination of the business cycle seems, by comparison, like an exercise of little importance. Yet attention to the study of the determinants of growth has been highly uneven.

The first classical economists—Adam Smith, David Ricardo, and Thomas Malthus—devoted most of their efforts to understanding the process of long-run economic growth. So did the classical economists of the first third of the twentieth century, most notably Joseph Schumpeter (1934) and Frank Ramsey (1928). With John Maynard Keynes (who is remembered, among other things, for saying “in the long run, we are all dead”), attention shifted to the determinants of the business cycle and the identification of policies to eliminate these annoying cycles.

Interest in long-run growth revived in the 1950s, and the growth theories of today are related to the neoclassical theories developed in those years, in particular the models of Robert Solow (1956) and Trevor Swan (1956), which are the cornerstone of those theories. The use of these models (described below) led neoclassical researchers to agree with the classical economists of the nineteenth century: the law of diminishing returns will eventually kill any growth based on physical accumulation; hence the only enduring source of growth must be technological progress.

After the prolific 1950s and 1960s, research on growth theory again declined. On the eve of the rational expectations revolution, economists’ attention returned to short-run business-cycle problems. One reason was the extreme technicality of growth research at the time and its dissociation from the real world.

Since the mid-1980s, with the publication of Paul Romer’s (1986) dissertation, growth theory has experienced a renaissance under the rubric of “endogenous growth.” This time researchers are more focused on empirical issues and real-world events. The availability of large-scale data sets and the extent of cooperation between academic researchers and economics practitioners have made the current round of economic growth research particularly exciting.

This chapter provides a short survey of the vast literature on growth theory.1 It introduces some of the tools used in analyzing the growth performance of a country, presents a sampling of the theoretical discussions of the possible sources of growth, and summarizes the findings from the vast empirical literature that examines the sources of growth in rich countries and the causes of slow growth in countries that have lagged behind.

The Neoclassical Production Function

The central element of the neoclassical theory of economic growth is the neoclassical production function. We assume that all the inputs to production can be aggregated into three basic ones: capital, labor, and technology. The production function describes how an economy can combine these three inputs to produce output (measured here as gross domestic product, GDP). For simplicity, we consider this GDP to be a single and simple good, for example, cookies. To produce cookies we need three basic inputs:

Capital. To produce cookies we need an oven, and we need flour, sugar, and cocoa. These inputs are called capital.

Capital consists of all physical inputs, such as machines, roads, computers, chalk, and buildings. These goods have one characteristic in common: they are all physical goods that need to have already been produced before we can produce the good in question (in this case, cookies). In other words, they are part of past production that was not consumed but, rather, was saved to be used as a factor in future production. The purchase of these physical goods is called investment. Capital thus consists of goods purchased by firms for use as inputs in future production processes. We will denote the input aggregate capital with the letter K.

Labor. To produce cookies we also need cooks. Cooks work to combine physical inputs, or capital, to convert them into cookies. We call this second important input labor.

We measure labor in terms of human beings residing in the economy who work for a certain number of hours a day and a certain number of days a year. Some people in the economy do not work in the production of GDP, possibly because they are too old (retirees) or too young (children), or because they cannot find a job (the unemployed), or because they simply do not want to work to produce GDP. In many economies, for example, women often work in production processes that are not directly computed in GDP, such as cottage industries or raising children.

Most economists compute labor by adding up the number of workers, or the number of hours worked, in the economy during a given period. This implicitly assumes that all workers and all hours worked are equally productive. But some workers are more productive than others. The productivity of a worker can be boosted by investing in that worker to improve his or her skills. Such investments in workers are called human capital rather than labor or work. These investments usually take the form of health care and education (although more educated workers are not necessarily more productive).2 Education is mainly acquired in school, of course, but it is also acquired through on-the-job training or work experience.

We will let L designate the labor input—that is, the aggregate of all the workers in the economy (possibly giving more weight to workers who are more skilled than to those who are less skilled). Following the discussion above, the labor input L can be disaggregated into the number of people working, N, times the skills of these people, e, so that L = Ne.

Technology or knowledge. The third important ingredient in the production of cookies is the formula. Before even skilled cooks (labor) can transform physical inputs (capital) into cookies, they need to know how to do that particular transformation. The formula or blueprint that tells them is called technology, or knowledge. We know there are different formulas, or recipes, for producing cookies, some better than others. Improvements in the formula are what we call technological progress.

Technology or knowledge is a “good” with special characteristics that distinguish it from the other two types of input. Economists call it a nonrival good, meaning that different people in different places can all use the same knowledge at the same time. This is not true for capital or for workers: a computer being used in New York cannot be used simultaneously in Washington. If a person is working in New York between 9:00 a.m. and 10:00 a.m. today, the same person cannot work at the same time in Buenos Aires. A formula or an idea, however, can be used simultaneously by many people in many places. For example, I can use the concept of addition to add 2 + 2 in my head. Yet my mental addition of 2 + 2 does not prevent any number of other people from adding 2 + 2 in their heads simultaneously. This is what we mean when we say that knowledge or technology is a nonrival good whereas capital and labor are rival goods.

However, a rival good is not necessarily an excludable good. Economists say that a good is excludable when people can be prevented (through the legal system or otherwise) from using it. For example, I can easily prevent other people from using my computer, simply by locking my office or setting up a password. But it is hard to prevent people from adding 2 + 2 in their heads once they have learned how. Table 4.1 classifies some examples of goods according to whether they are rival or excludable to show the important distinction between these two concepts.

Table 4.1.Classification of Goods According to Rivalry and Excludability
Type of GoodRivalNonrival
Highly excludableRice, pencils, workers, computersCable television
Partially excludableMicrosoft Windows 98
NonexcludableFish in the seaWatching the moon
Grass in public areasEinstein’s relativity theory

Most of the goods we deal with every day are both rival and excludable. These include not only items like rice, pencils, and computers, but also workers’ labor. Goods that are rival but nonexcludable are another interesting set. For example, fish in the sea and grass in a public commons fall in this category. No one can prevent me from setting out to sea to catch fish. Such goods are therefore nonexcludable, but the fish I catch are fish you cannot catch as well, so they are also rival.

Nonrival goods differ in their degree of excludability. At one extreme are such excludable goods as satellites or cable television. The cable company can prevent us from watching its channels by scrambling our screen; thus cable television images are excludable. However, millions of people in different places can watch the same images at the same time, so the images are entirely nonrival. At the other extreme, some nonrival goods, such as knowledge about addition, are nonexcludable, as already mentioned. No one can prevent us from adding (adding is nonexcludable), and many people can be adding at the same time (it is nonrival). Other goods in the same category are Tobin’s square root formula for money demand, Einstein’s relativity theory, and the moon. These goods are usually called public goods.

Some goods are partially excludable, for example, the programming codes for Microsoft Windows 98. In principle, the legal system prevents us from copying these codes unless we pay a fee to the inventor. So they appear to be excludable. However, once the program is purchased, the manufacturer has difficulty preventing people from passing copies on to their friends and relatives. Therefore these goods are not fully excludable. To provide an incentive for people to generate new knowledge and ideas, governments need to establish a legal environment that guarantees inventors’ right to collect money for the use of their inventions. In other words, governments need to create a system that makes sure that the invented goods become as excludable as possible (for a while, at least).

Now, back to the production function. The production possibilities of an economy are described through a formula that puts together the three inputs: capital and labor are combined with a formula (designated here as A) to generate cookies or output:

In neoclassical theory, the function F has a number of properties. One of these is that it must exhibit constant returns to scale. This means that if we double capital and labor, we should be able to double output. This assumption makes sense if we think of the principle of replication. Consider a plant in Iowa that combines capital and labor and uses a formula to produce cookies. If we reproduce the plant in New Jersey, with the same amounts of capital and labor, and if the New Jersey plant uses the same formula, we should be able to produce the same amount of output in New Jersey. Thus, by doubling capital and labor we should be able to double output. Notice that we do not also need to double the formula, because, as argued above, the formula is nonrival, so the same formula can be used in both Iowa and New Jersey.

Another important neoclassical assumption is that the production function exhibits diminishing returns to capital. If we add one more machine to our stock of capital without adding any more labor, the production of cookies will increase, but the amount by which it will increase depends on how much capital we had to begin with. If we are very poor and have few machines, the addition of an extra machine will generate a relatively large increase in output. But if we are rich and already have many machines, adding one more will trigger only a small increase in output. In other words, the more capital we have, the less productive an additional unit of capital will be to our enterprise.

An example of a neoclassical production function is the well-known Cobb-Douglas production function:

where α is the fraction of national income appropriated by the owners of capital, and 1 - α is the fraction appropriated by workers.3

According to this production function, the amount of cookies produced in our economy can increase for one of three reasons: capital increases, labor increases, or technology improves. We usually measure output in terms of output per capita or per worker.4 Using our notation above, let us decompose the labor input, L, into the total number of workers, N, and their quality or skill, e. If we divide both sides of equation (4.2) by N to get Y/N, or output per worker, we get

where y = Y/N is output per worker, and k = K/N is capital per worker. Equation (4.3) says that output per worker can increase in one of three ways: by increasing the capital stock per worker, k, through investment and capital accumulation; by increasing the skills of the workforce, e, primarily through investment in education; or by increasing the level of technology, A.

Neoclassical theory says that an economy cannot grow forever through investment in capital alone. In other words, if an economy seeks to grow by simply increasing its investment rate, it will not succeed in growing in the long run. In the short run, capital, and thus output, will increase. But as the stock of capital gets larger, the growth rate will diminish until it eventually stops.

The intuition behind this conclusion is simple and powerful: the law of diminishing returns. Consider a poor economy with a small stock of capital. Imagine that this economy experiences no growth in technology, education, or capital per worker, so it is not growing at all. Now imagine that the economy decides to increase its investment rate, so instead of investing 5 percent of GDP, it invests 20 percent of GDP in physical capital. Initially, the capital stock increases in response to this investment. Since the economy had a small stock of capital to begin with, the increase in k triggers a relatively large increase in GDP. Next year the country again invests 20 percent of GDP. However, the capital stock in place is now larger than it was the year before, because of the past year’s investment. Hence, the percentage increase in GDP (the GDP growth rate) will be smaller than it was last year. If the country continues to invest at the same rate, this process will continue until the law of diminishing returns eventually reduces the growth rate to zero. At that point the economy will be richer, but it will have stopped growing. Hence, a higher rate of investment in physical capital will not generate positive growth forever.

Another important concept from neoclassical theory is that of convergence. Imagine that two countries have identical production functions. One country, however, has a smaller capital stock than the other. Under these circumstances, the poorer country will have a higher marginal product of capital, because of the assumption of diminishing returns to capital: if we invest one unit of capital in the poorer country, we will get a larger increase in output than if we do the same in the richer country. Hence, in principle, the poorer country should grow faster than the richer one. This prediction is known as the absolute convergence hypothesis. This hypothesis can be tested with data, as will be done later in this chapter.

Clearly, the increase in output generated by investing in one more machine is larger for a country with small k (a poor country) than for a country with large k (a rich country). However, the growth rate is determined not only by the extra number of machines added when k is increased by one unit, but also by the number of units actually invested. A poor country may have a large marginal product, so that if it invests in one extra machine it will get a larger increase in output than will a rich country that does likewise. But the rich country may instead actually invest in more machines—more units of capital. Each country will get an increase in output, but the increase may be larger in the rich country. Hence, it is possible for the rich country to grow faster by investing more than the poor country. One way a rich country might be able to invest more units of capital than a poor country is if it has a higher rate of saving. Thus a key assumption underlying the absolute convergence hypothesis is that the two countries have identical saving rates. If instead we allow for differences in saving rates, the neoclassical model predicts what is known as the conditional convergence hypothesis (discussed further below).

Growth Accounting

We are interested in knowing the sources of growth for a country because some types of growth may be healthier or more promising than others. For example, a country can grow rapidly (at least for a while) by increasing its capital stock through high saving and investment rates, by bringing all women into the labor force, or by increasing school enrollment rates. However, a country that follows this path is likely to see growth slow down in the future, because such events are not repeatable: once all women are working, we cannot bring more women to work; once the entire population has gone to college, there is no point in sending them to college again. At the other extreme, a country that has grown through large productivity improvements (increases in A) might be able to keep growing at similar rates, because knowledge has no frontiers.5

To show this, we need to measure the growth rate of technology (or the rate of technological progress). But since technology itself cannot be measured, we cannot measure the rate of technological progress directly. To see this, let us compute the growth rate of GDP from equation (4.3):

Equation (4.4) shows that the growth rate of GDP per worker can be decomposed into three components: the growth rate of technology, the growth rate of capital per worker, and the growth rate of education or skills. The immediate problem we face is that the growth rate of A cannot be directly measured. The other terms are easy: Δy/y is the growth rate of aggregate GDP per worker (this is measured in the national accounts), Δk/k is the growth rate of the capital stock (also measured in the national accounts), and Δe/e is the growth rate of labor education (or some other measure of skills), which also can be measured. As mentioned above, the coefficient 1 - α is the fraction of national income that accrues to workers (equal to the wage bill divided by national income), and therefore is the fraction of national income that accrues to capital. The only variable we cannot measure directly in equation (4.4) is the growth rate of A. Robert Solow’s central insight was that we can instead measure ΔA/A indirectly, by rearranging the equation as follows:

That is, the growth rate of productivity can be measured as the difference (or the residual) between the growth rate of GDP and the growth rate of the two inputs, each of which is weighted by its share or contribution to the production of GDP.

New Growth Theories

According to the neoclassical model discussed above, in a world with neoclassical technology and perfect competition, the main driving source of economic growth is technological progress. But the neoclassical researchers left unexplained the process by which technological progress occurs, by assuming that technology grows at an exogenous rate. This is clearly unsatisfactory theoretically because it is tantamount to saying that the ultimate source of growth is unexplained. Since the mid-1980s, in a departure from the neoclassical tradition, a large number of researchers have worked hard to determine the exact sources of growth. The resulting literature is known as the “new growth literature” or the literature of “endogenous growth.” Although it is difficult to summarize this vast literature in a short chapter, the research can be divided into three broad categories.

Human Capital

As mentioned earlier, the input labor can increase its productivity or skills through a process of investment. This process is often called education, although, as already noted, skills can also be acquired outside the formal education system. We can overcome the law of diminishing returns if we invest simultaneously in human and physical capital. Because the model assumes constant returns to scale (meaning constant returns to physical and human capital taken together), if we double both physical capital and human capital, we should be able to double output without diminishing returns.

The basic way to acquire human capital or labor skills is through the education system. People go to school for a number of years, and in doing so they sacrifice the wages they would have earned if they had worked instead of going to school. This sacrifice of wages (and therefore of consumption opportunities) is comparable to investment in physical capital through saving, and this is what has led some economists to call this input human capital instead of simply labor.

But what kind of education should a country provide its citizens? Specifically, which of the following two alternatives is more desirable for increasing the rate of economic growth? The first alternative, which we label elitist, educates a few people very well on the premise that a country performs according to the quality of its leaders, political, economic, and social. If the leaders are well educated, in this view, the country as a whole will do well.

The second alternative, which we label universalist, educates everyone to at least some minimum level. This alternative is based on the premise that economic growth requires workers to adopt new technologies rapidly. Over the past 30 years the high-growth countries of East Asia started their climb by producing textiles, then watches, then radios. They went up the quality ladder until eventually they were producing financial services and genetically engineered products.

The average worker in such an economy needs to learn several different jobs over the course of his or her productive life. Therefore such an economy needs a universalist educational system in which people not only learn specific skills but also learn how to learn. The experience of the past 30 years suggests that the universalist model of human capital accumulation has been more successful than the elitist model: countries that have adopted the universalist model have performed a bit better over the long run.

Technology

The economics of technology, or knowledge, is a bit different from the economics of most private goods. As mentioned earlier, knowledge is a nonrival good, because many people can use the same knowledge simultaneously. The nonrival nature of knowledge has a crucial implication: it needs to be produced only once. The spaghetti that you consume, I cannot consume, so firms need to produce spaghetti for you and spaghetti for me. But a formula, a concept, or an idea needs to be produced only once, because once it is created, many people can use it simultaneously. This means that the production of technology has fixed costs (often called research and development, or R&D costs), which are often high but are paid for only once.

For example, the production of new programming codes for the first version of Microsoft Excel required many people working day and night to create the codes. But once the codes were created, they were easily reproduced on CD-ROMs or diskettes. Similarly, Black and Scholes (1973) worked for several years to produce a formula for pricing financial options. But once the formula was found, it was relatively easy to reproduce it from book to book and from calculator to calculator. The initial cost was high, but the marginal cost of producing a new copy of the formula was near zero.

The fact that knowledge goods are associated with large fixed costs means that firms will tend to produce them only if they are in imperfect competition. This is the main departure of the second strand of the new literature from the neoclassical framework: the world is not neoclassical because some firms are not perfectly competitive.

We know from microeconomics that when there are fixed costs, marginal costs tend to be below the average cost. Under perfect competition, the price would tend to be equal to the marginal cost. Consequently, any firm inventing a new piece of knowledge would lose money. Hence, firms tend to produce new goods only when there is a legal system to guarantee inventors the right to sell their goods at a profit to recover the initial cost of their inventions. This is where the other main characteristic of knowledge, nonexcludability, comes into play. The government must provide the legal framework that makes the technology goods as excludable as possible, so the inventors can prevent people from using the goods unless they buy the goods from them at a profit. The concept of excludability is thus associated with the legal system: the codes for Microsoft Excel could be easily reproduced and sold by pirates if a legal system protecting Microsoft did not exist. The government may introduce patent laws and other institutions that grant inventors monopoly power over their inventions—at least for a while. But the development of goods that cannot be made fully excludable, such as basic knowledge, will have to be financed through direct government subsidies for research.

Some economists argue that the technological progress spawned by the industrial revolution took place mainly because national governments had become strong enough to guarantee that private inventors would have some kind of market power over the sale of their goods. Thus, to induce the creation of knowledge and the generation of long-run economic growth, a country must establish a legal system that protects inventors and gives them intellectual property rights to the sale of their inventions. Without this guarantee, much technological innovation would never take place.

When designing the institutions that protect property rights, governments need to weigh two factors. Ex post—-that is, once the good has been invented—it is optimal for the government to disallow the monopoly power of the inventor. For example, once a pill that cures cancer has been invented, it would be best for society if the government simply forced the inventing firm to give the pill away for free, so that everyone in the world could have it. But if drug companies know that the government will force them to give away their new inventions, or sell them at a price equal to the marginal cost, thus preventing them from collecting profits, they will never invent a pill for cancer. Consequently, although it would be optimal ex post to set a low or zero price on the invention, it is optimal ex ante to promise a large profit for the inventors. Governments have to find a balance between the two objectives of promoting innovation while limiting monopoly. In the pharmaceutical industry in the United States, for example, this balance has been achieved through the issuance of temporary patents: the pharmaceutical firm that invents a new product enjoys a patent of seven years, during which only that firm may sell the product, at monopoly prices. After the seventh year, any company that wishes to do so may produce the good, so perfect competition drives prices down.

To achieve global growth, the patent laws that protect inventors in one country need to protect them worldwide as well. In other words, the world as a whole needs to establish international protection of intellectual property rights, so that goods invented in, for example, the United States can be sold at a profit worldwide without fear of another country copying them for free. This is one of the goals of the World Trade Organization.

Government

Many economists have studied the ways in which a national government can affect the rate of economic growth. Obviously, the government is the single most powerful economic agent in an economy, so its actions are quite important. The government can affect the economy in many ways. For example:

  • It sets up the legal system to allow private individuals and firms to produce and to reap the rewards of their production. If it fails to do its job here—if it over- or underregulates, if it is corrupt, or if it fails to protect property rights—the growth rate of the economy will be lower.

  • The government influences the political environment in which firms operate. For example, sharp disagreements between political factions create uncertainty about the future direction of policy, which reduces investment and growth.

  • The government can affect the macroeconomic environment in which firms operate. For example, firms may hesitate to do business in a country with high inflation, large fiscal deficits, large swings in the value of the currency, high unemployment, serious external sector distortions, and financial repression.

  • The government can affect the growth rate through the imposition of taxes and through its own purchases of part of the country’s output (government spending). The composition of spending may also be important, because some types of spending are productive whereas others are not. For example, public investment is probably better for growth than are social security transfers to the elderly. The kind of taxes that are imposed can also have an impact on the rate of growth.

The economic growth literature has generated many models for analyzing the effects of these issues on growth. The next section analyzes the empirical evidence to show which channels are more likely to allow the government to affect growth.

Channels for Growth: The Empirical Evidence

Growth Accounting

Tables 4.24.7 (from Bosworth and Collins, 1998), grouped at the end of the chapter, show the breakdown of growth for 88 economies over four periods: 1960–70, 1970–80,1980–86, and 1986–92. For example, the first row in Table 4.2 shows that the average annual growth rate of output per worker in China over 1960–70 was 1.7 percent. The contribution of capital to growth—the term αΔk/k in equation (4.4)—was zero, and the contribution of education—the term (1 – α)Δe/e in equation (4.4)—was 0.4 percentage point. The rest of the growth rate, 1.3 percentage points, came from the increase in technology (shown in the last column).

Table 4.2.East Asia: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
China
1960-701.7-0.00.41.3
1970-803.21.90.50.8
1980-867.12.50.44.0
1986-926.23.10.52.5
Indonesia
1960-701.80.50.50.8
1970-805.03.50.31.1
1980-862.63.20.5-1.1
1986-923.92.60.50.8
Korea
1960-705.13.50.90.6
1970-805.94.50.50.8
1980-866.22.90.72.5
1986-926.63.90.71.9
Malaysia
1960-703.72.70.40.6
1970-804.02.80.40.8
1980-861.52.80.6-1.9
1986-925.41.90.62.8
Philippines
1960-702.31.60.50.2
1970-803.31.90.50.8
1980-86-3.01.30.44.6
1986-920.70.30.4-0.0
Singapore
1960-705.65.20.30.1
1970-804.33.9-0.00.4
1980-863.63.70.7-0.8
1986-927.42.60.64.0
Taiwan Province of China
1960-706.54.50.51.4
1970-806.14.10.71.1
1980-864.52.10.51.8
1986-925.92.80.54.0
Thailand
1960-705.23.90.01.2
1970-803.82.70.10.9
1980-863.11.90.90.3
1986-928.33.20.84.0
Table 4.3.South Asia: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
Bangladesh
1960-702.21.00.01.2
1970-80-0.5-0.10.4-0.8
1980-862.20.30.31.6
1986-920.90.00.30.5
India
1960-702.41.60.20.5
1970-801.31.10.4-0.2
1980-863.21.10.31.8
1986-923.31.40.31.5
Myanmar
1960-700.70.60.10.1
1970-802.40.50.21.7
1980-861.81.40.6-0.3
1986-92-1.90.30.6-2.8
Pakistan
1960-705.14.40.60.1
1970-801.91.0-0.00.9
1980-863.31.00.22.2
1986-921.40.80.22.4
Sri Lanka
1960-702.40.60.61.2
1970-802.21.90.10.2
1980-863.42.60.10.6
1986-921.91.40.10.4
Table 4.4.Africa: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
Cameroon
1960-700.21.20.1-1.1
1970-806.02.70.32.9
1980-865.53.90.41.2
1986-92-7.00.70.4-7.9
Congo, Dem. Rep. of
1960-702.2-0.20.22.2
1970-80-1.40.60.3-2.3
1980-86-0.01.40.4-1.8
1986-92-5.50.40.4-6.3
Côte d’Ivoire
1960-706.32.70.13.4
1970-803.33.00.20.0
1980-86-0.70.10.3-1.1
1986-92-3.7-1.20.3-2.8
Ethiopia
1960-702.22.60.0-0.4
1970-800.90.50.10.4
1980-86-0.82.30.1-3.1
1986-92-1.61.30.1-3.0
Ghana
1960-701.32.10.5-1.2
1970-80-1.9-0.00.1-2.0
1980-86-2.0-1.10.6-1.5
1986-921.60.20.60.8
Kenya
1960-701.2-0.40.11.5
1970-804.30.40.43.4
1980-86-0.1-0.70.40.2
1986-92-0.1-0.70.40.2
Madagascar
1960-701.10.70.00.3
1970-80-1.20.30.2-1.6
1980-86-3.0-0.60.3-2.7
1986-92-1.20.00.3-1.5
Malawi
1960-702.53.20.1-0.7
1970-803.93.00.10.7
1980-86-0.9-0.40.2-0.7
1986-92-0.6-0.40.2-0.3
Mali
1960-701.40.90.00.5
1970-802.50.40.12.0
1980-860.00.20.2-0.3
1986-92-0.50.70.23.6
Mauritius
1960-700.1-0.70.40.4
1970-802.70.60.41.6
1980-862.3-0.60.22.8
1986-924.01.50.22.3
Mozambique
1960-703.21.80.21.1
1970-80-4.6-0.10.1-4.6
1980-86-6.2-0.70.2-5.7
1986-923.6-0.20.23.6
Nigeria
1960-700.61.70.2-1.3
1970-801.53.80.1-2.4
1980-86-4.1-0.10.3-4.3
1986-922.31.20.43.1
Rwanda
1960-700.10.50.1-0.4
1970-801.81.30.20.3
1980-860.53.3-0.1-2.6
1986-92-2.61.2-0.1-3.6
Senegal
1960-700.2-0.10.00.2
1970-80-1.20.10.1-1.5
1980-861.2-0.10.31.1
1986-920.60.20.30.0
Sierra Leone
1960-704.22.50.11.5
1970-800.70.70.2-0.2
1980-86-0.80.50.1-1.3
1986-920.8-0.20.11.0
South Africa
1960-702.31.10.21.0
1970-802.12.0-0.10.2
1980-86-1.60.50.2-2.3
1986-92-1.8-0.50.2-1.5
Sudan
1960-70-0.73.20.1-3.8
1970-801.11.60.2-0.6
1980-86-1.41.00.3-2.7
1986-92-0.5-0.40.3-0.4
Tanzania
1960-703.00.40.22.4
1970-800.51.10.1-0.6
1980-86-1.2-0.1-0.1-1.0
1986-922.40.1-0.12.4
Uganda
1960-701.31.1-0.00.3
1970-80-4.30.20.2-4.7
1980-860.8-0.40.21.0
1986-923.00.80.22.0
Zambia
1960-701.00.50.20.4
1970-80-1.1-0.20.4-1.3
1980-86-2.9-2.00.5-1.4
1986-92-2.6-2.30.4-0.7
Zimbabwe
1960-702.7-0.40.13.0
1970-80-0.10.70.2-1.1
1980-860.9-0.10.20.8
1986-92-1.3-0.30.2-1.2
Table 4.5.Middle East: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
Algeria
1960-702.70.7-0.02.0
1970-802.42.10.5-0.2
1980-860.61.10.5-1.0
1986-92-2.8-0.70.5-2.6
Cyprus
1960-706.51.80.54.1
1970-802.91.40.60.9
1980-864.01.6-0.02.3
1986-925.71.5-0.04.1
Egypt
1960-703.31.30.21.8
1970-805.83.00.32.5
1980-863.63.20.20.1
1986-92-0.20.80.2-1.3
Iran, Islamic Rep. of
1960-706.03.60.31.9
1970-80-2.43.60.5-6.3
1980-86-2.10.30.8-3.1
1986-920.6-0.80.70.7
Israel
1960-704.91.30.33.2
1970-802.81.50.60.7
1980-861.00.40.10.5
1986-922.90.70.12.1
Jordan
1960-702.23.30.1-1.1
1970-807.63.80.63.0
1980-860.52.01.1-2.6
1986-92-3.9-0.61.0-4.3
Malta
1960-703.72.20.11.4
1970-808.20.90.37.0
1980-861.01.80.4-1.2
1986-925.22.00.42.7
Morocco
1960-705.91.00.24.6
1970-802.02.00.2-0.2
1980-860.70.80,2-0.3
1986-92-0.60.40.2-1.2
Tunisia
1960-703.91.90.21.7
1970-801.51.30.51.8
1980-860.11.10.5-1.6
1986-921.8-0.10.51.4
Table 4.6.Latin America: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
Argentina
1960-702.81.40.31.1
1970-801.71.50.3-0.1
1980-86-2.0-0.20.2-2.0
1986-920.6-0.70.21.1
Bolivia
1960-703.51.20.22.0
1970-802.41.70.20.5
1980-86-4.6-1.20.3-3.7
1986-920.7-1.40.31.8
Brazil
1960-703.11.30.11.6
1970-804.92.40.12.4
1980-86-0.00.70.4-1.1
1986-92-1.90.30.4-2.6
Chile
1960-702.61.00.21.8
1970-800.1-0.20.30.0
1980-86-1.9-0.30.3-1.9
1986-925.11.00.33.8
Colombia
1960-702.50.50.21.8
1970-802.90.90.51.4
1980-860.10.80.3-0.9
1986-921.30.40.30.7
Costa Rica
1960-702.51.30.11.1
1970-801.71.80.6-0.8
1980-86-1.80.10.5-2.3
1986-922.00.90.50.7
Dominican Rep.
1960-703.51.40.31.8
1970-803.63.00.30.3
1980-86-1.60.40.4-2.5
1986-92-1.01.10.4-2.4
Ecuador
1960-701.80.80.10.8
1970-806.51.80.93.7
1980-86-0.70.20.3-1.3
1986-92-0.4-0.30.3-0.4
El Salvador
1960-702.11.20.20.7
1970-800.21.60.4-1.8
1980-86-4.2-0.60.2-3.8
1986-92-0.5-0.40.2-0.3
Guatemala
1960-703.01.10.21.6
1970-803.41.60.31.4
1980-86-3.7-0.30.2-3.6
1986-920.6-0.50.20.9
Guyana
1960-701.40.5-0.21.1
1970-80-2.3-0.50.3-2.2
1980-86-5.8-1.30.4-5.0
1986-92-3.0-1.10.4-2.3
Honduras
1960-702.21.20.10.9
1970-802.31.20.30.8
1980-86-2.7-0.20.8-3.3
1986-92-0.00.10.7-0.8
Haiti
1960-70-1.10.10.1-1.2
1970-803.42.70.20.4
1980-86-2.71.40.1-4.2
1986-92-4.30.00.1-4.4
Jamaica
1960-703.91.80.31.7
1970-80-3.8-0.30.2-3.7
1980-86-2.7-1.30.5-1.8
1986-920.9-0.40.40.8
Mexico
1960-704.22.10.41.7
1970-802.11.60.10.4
1980-862.20.70.7-3.5
1986-92-0.20.10.7-0.9
Nicaragua
1960-703.82.00.11.6
1970-80-2.50.80.2-3.5
1980-86-3.6-0.20.9-4.3
1986-92-5.6-1.30.8-5.1
Panama
1960-704.82.60.21.9
1970-803.02.20.60.2
1980-86-0.00.50.3-0.8
1986-92-1.5-0.60.3-1.2
Paraguay
1960-701.91.00.20.7
1970-805.03.00.41.7
1980-86-1.62.00.1-3.7
1986-921.10.90.10.0
Peru
1960-703.21.10.31.8
1970-800.20.30.7-0.9
1980-86-1.20.00.3-1.5
1986-92-5.4-0.60.3-5.1
Trinidad and Tobago
1960-702.91.30.01.5
1970-803.92.50.60.7
1980-86-4.91.2-0.0-6.0
1986-92-3.2-0.4-0.0-2.8
Uruguay
1960-700.7-0.20.20.6
1970-802.70.90.41.3
1980-86-2.30.00.5-2.8
1986-922.5-0.30.52.4
Venezuela
1960-702.2-0.10.22.0
1970-80-2.10.60.9-3.5
1980-86-3.1-0.50.4-3.0
1986-920.7-0.70.40.9
Table 4.7.Industrial Countries: Sources of Growth, 1960–92(Annual percentage change)
Contribution of:
Region/PeriodOutput per WorkerPhysical capitalEducationFactor productivity
Australia
1960-702.91.10.71.1
1970-801.70.9-0.00.8
1980-861.20.50.10.5
1986-921.10.40.10.6
Austria
1960-705.22.40.82.0
1970-803.31.60.11.5
1980-860.90.80.3-0.2
1986-921.70.60.30.8
Belgium
1960-704.31.30.12.8
1970-803.11.20.51.4
1980-861.40.60.30.4
1986-922.00.60.31.0
Canada
1960-702.20.40.21.6
1970-801.50.50.70.2
1980-861.70.90.20.6
1986-920.71.10.1-0.5
Denmark
1960-703.52.00.11.4
1970-801.81.20.30.3
1980-861.80.20.21.3
1986-921.50.70.20.6
Finland
1960-704.71.60.52.5
1970-802.61.00.90.7
1980-862.10.7-0.21.6
1986-922.41.4-0.21.2
France
1960-704.92.10.32.5
1970-802.81.50.60.8
1980-861.80.90.50.5
1986-922.10.80.50.8
Germany
1960-704.31.80.32.2
1970-802.51.10.21.2
1980-861.40.70.10.6
1986-921.80.30.11.4
Greece
1960-708.53.10.44.8
1970-804.01.80.71.5
1980-860.20.50.2-0.5
1986-921.40.60.20.6
Iceland
1960-702.91.10.31.4
1970-803.70.90.42.3
1980-86-0.20.30.4-0.8
1986-920.90.90.4-0.3
Ireland
1960-704.21.70.02.5
1970-803.81.50.51.7
1980-863.11.50.41.2
1986-924.30.50.33.4
Italy
1960-706.12.10.33.6
1970-803.11.10.31.7
1980-861.40.70.40.2
1986-921.80.70.40.7
Japan
1960-708.93.80.05.0
1970-803.62.50.70.5
1980-862.61.20.31.1
1986-922.61.30.31.0
Netherlands
1960-703.91.61.11.1
1970-802.61.10.31.2
1980-861.30.60.30.4
1986-920.30.00.3-0.1
New Zealand
1960-701.20.50.10.6
1970-800.60.61.2-1.2
1980-861.70.6-0.11.3
1986-921.31.0-0.10.4
Norway
1960-703.51.00.61.8
1970-803.21.01.40.8
1980-862.00.60.31.4
1986-921.40.8-0.10.7
Portugal
1960-706.42.3-0.54.5
1970-803.01.21.20.5
1980-860.90.80.6-0.5
1986-923.10.90.61.6
Spain
1960-706.62.40.63.4
1970-804.02.00.21.7
1980-862.71.10.31.3
1986-921.90.80.30.8
Sweden
1960-704.01.50.02.4
1970-801.00.80.8-0.5
1980-861.80.60.01.1
1986-921.10.80.00.2
Switzerland
1960-703.21.6-0.31.9
1970-801.11.11.5-1.4
1980-860.40.5-0.40.3
1986-921.41.0-0.40.8
Turkey
1960-705.01.50.03.4
1970-803.21.70.41.0
1980-864.00.90.72.3
1986-923.10.90.61.6
United Kingdom
1960-702.61.5-0.01.1
1970-801.71.00.60.2
1980-862.70.70.31.6
1986-921.00.80.30.0
United States
1960-702.00.50.60.9
1970-800.40.20.7-0.5
1980-861.10.3-0.00.9
1986-921.00.4-0.00.6

Alwyn Young (1994) was the first to arrive at the important finding that the growth performance of the East Asian “miracle” economies (including, most prominently, Singapore) could be explained by factor accumulation. For example, the growth rate of GDP per capita in Singapore during 1960–70 averaged 5.6 percent a year. Of this, physical capital accumulation accounted for 5.2 percentage points, and education or human capital accumulation accounted for 0.3 percentage point. Once we account for physical and human capital accumulation, the share of Singapore’s growth rate remaining to be explained (the growth rate of technology) is close to zero (0.1 percentage point). Similar results are found for the periods 1970–80 and 1980–86. These findings led Young to conclude that Singapore’s economic growth was due not to “inspiration” (getting better and better at producing goods with the same amount of effort) but to “perspiration” (saving and the accumulation of physical capital).

This interesting and important conclusion contradicted the previous, widely held belief that the growth of these East Asian countries was due to substantial improvements in technology. In fact, their growth was almost entirely due to the old-fashioned accumulation of physical capital. An important implication was that if these countries failed to increase their level of technology and instead continued to accumulate predominantly physical capital, the law of diminishing returns would eventually reduce their growth rates to zero. This was the devastating picture that Young painted in 1994, predicting that the East Asian miracle would soon be over. Luckily for Singapore, the growth rate of productivity picked up between 1986 and 1992, averaging an annual 7.4 percent. Only 2.6 percentage points of this could be accounted for by capital accumulation and 0.6 percentage point by education. Technology, therefore, grew at an impressive rate of 4 percent a year.

The Convergence Debate

Convergence across economic units, whether households or countries, is an important economic issue. As already mentioned, the neoclassical model predicts conditional convergence between rich and poor countries. Equally important, labor, public finance, and growth economists want to know whether today’s relatively poor families will remain poor for many generations, and whether the dynasties that are rich today are the same ones that will be rich a hundred years from now. They also want to know whether the degree of income inequality across families increases or falls over time. These are important questions for anyone interested in a society’s general welfare, and for policymakers who want to pursue redistributive policies to achieve social peace.

Macroeconomists and theorists of economic growth are interested in exactly the same questions. For them, however, the relevant unit of analysis is not the family but the country or region within a country. They want to know, for example, whether rich countries will remain rich and poor countries poor for many decades, and whether the distribution of world income and output across countries is becoming increasingly equal over time. These important questions lie at the heart of the convergence debate. Although economists have been interested in these issues for decades, the debate has captured the attention of mainstream macroeconomic theorists and econometricians only since the end of the 1980s.

Two main concepts of convergence appear in the classical literature. They are called β-convergence and σ-convergence. We say that there is β-convergence if poor economies tend to grow faster than rich ones. Evidence on β-convergence can be found when the coefficient of a regression of the growth rate for a cross section of families or economies is negative on the initial level of income. For example, the left-hand panel of Figure 4.1 shows a hypothetical set of economies that display β-convergence. For each economy the growth rate between 1960 and 1997 is computed. In this hypothetical example, the economies that were poor in 1960 grew faster than other economies over the following 37 years. In contrast, the right-hand panel of Figure 4.1 shows a hypothetical example in which the poor grew less rapidly than the rich, so there is β-divergence.

Figure 4.1.Convergence and Divergence of Growth, 1960–97

(Growth, percent a year)

Note: Each point represents a country.

Economies display σ-convergence when the dispersion in their levels of real GDP per capita tends to decrease over time. That is, if we compute the coefficient of variation for year 0, then for year 5, then for year 10, and so on and observe that the variation declines over time, we say that there is σ-convergence.

These two convergence concepts are, of course, related. Intuitively, we can see that if GDP per capita of two economies becomes similar over time, it must be that the poorer economy has grown faster. In the section that follows, we analyze the two concepts in terms of a sample of economies.

World Distribution of Income

Are we living in a world in which poor economies are growing faster than rich ones, so that there is catching up, or β-convergence? Or are we living in a world in which the poor are becoming poorer and the rich richer? In other words, is world income inequality rising or falling?

Figure 4.2 shows, for a sample of 110 countries over 30 years beginning in 1960, the absence of any relationship between the initial level of income and the subsequent growth rate. Some poor countries grew very little, and some rich countries grew very little. On the other hand, some poor and middle-income countries, as well as some rich countries, experienced considerable growth. We must conclude that the economies of the world did not exhibit β-convergence during this period, since in general the poor countries did not grow faster than the rich ones.

Figure 4.2.Absence of Growth Convergence Across Countries Worldwide, 1960–90

(Growth of GDP, percent a year)

1Sample of 110 countries.

Figure 4.3 analyzes the evolution of the dispersion of income across these same economies over time. It shows that, during these 30 years, income inequality across these 110 economies increased. Figures 4.2 and 4.3 also suggest that if we look at the subsample of richer countries that are members of the Organization for Economic Cooperation and Development (OECD) instead of at the world at large, we will see a slightly different picture. For the OECD countries, the relationship between growth and the initial level of income is significantly negative, as depicted by the downward-sloping regression line in Figure 4.2. Thus, the OECD economies have converged in the sense of β-convergence. Similarly, income inequality among the OECD economies has declined since 1950 (Figure 4.3), so these economies also exhibit σ-convergence.

Figure 4.3.Dispersion of GDP Worldwide and in OECD Countries

(Growth of GDP, percent a year)

1Sample of 110 countries.

Examples of Regional Distribution of Income

It is important to know whether, in the course of a country’s economic growth and development, its different regions tend to converge or not (in the sense of β- and σ-convergence). If the tendency is toward divergence, it means that regional inequalities are increasing as the country’s economy progresses, and policymakers may want to do something to reverse this trend.

The states of the United States. Figures 4.4 and 4.5 display the behavior of different regions within the United States. Figure 4.4 is a scatterplot of the average annual growth rate of income per capita in the U. S. states between 1880 and 1990 against the logarithm of income per capita in 1880. The poor states in 1880 (that is, shortly after the Civil War, a time when the poor states were mainly in the South) are also the ones that experienced the fastest growth over the 110 years. The opposite is true for the richest states. Hence, the states of the United States enjoyed β-convergence between 1880 and 1990.

Figure 4.4.Convergence of Personal Income per Capita Across U.S. States, 1880–1990

(Growth of personal income per capita, percent a year)

Figure 4.5 shows the evolution of income dispersion across U.S. states over the past century. Although some ups and downs are evident (inequality rose during the 1920s and again during the 1980s), the long-run tendency is toward a decline in income dispersion. Hence, there has been σ-convergence across the states of the United States as well.

Figure 4.5.Dispersion of Personal Income Across U.S. States, 1880–1992

(Standard deviation of log of personal income per capita)

Japanese prefectures. Figures 4.6 and 4.7 repeat the analysis described above for Japan’s 47 prefectures. Here also the tendency is toward convergence, in the sense of both β- and σ-convergence.

Figure 4.6.Convergence of Personal Income per Capita Across Japanese Prefectures, 1930–90

(Growth of personal income per capita, percent a year)

Figure 4.7.Dispersion of Personal Income Across Japanese Prefectures, 1930–90

(Standard deviation of log of personal income per capita)

European regions. Figures 4.8 and 4.9 repeat the exercise for different regions of a sample of five European countries (France, Germany, Italy, Spain, and the United Kingdom). Figure 4.8 shows the type of negative relationship that is familiar from our discussion of U.S. states and Japanese prefectures. The correlation between the growth rate and the logarithm of initial GDP per capita in Figure 4.8 is not as high as those observed for Japan and the United States, but it is still strongly and robustly negative. Finally, Figure 4.9 shows that income dispersion within each of the five European countries has declined systematically. Hence, there is both β- and σ-convergence across regions in Europe as well.

Figure 4.8.Convergence of GDP per Capita Across European Regions, 1950–90

(Growth of GDP per capita relative to country mean, percent a year)

Figure 4.9.Dispersion of GDP per Capita in Five European Countries

(Standard deviation of log of personal income per capita)

Reconciling Cross-Country and Cross-Regional Results

At first glance, it may appear that the lack of convergence across countries worldwide stands in contradiction to the strong evidence for convergence across regions. One might think that the cross-country results also contradict the neoclassical theory, which embraces the assumption of diminishing returns to capital, whereas the regional results support the theory. But when the results are considered in the context of conditional convergence, both sets of results can be reconciled with the neoclassical model. The reason that poor countries grow more slowly than rich countries might be that the growth rate of an economy depends not only on its marginal product but also on the number of machines the country actually invests in. It is perfectly conceivable that a poor African country could have a very large marginal product (for each additional machine in which it invests, it gets a large increase in output) but a low growth rate, because the number of machines actually invested in (the investment rate) is very small. Thus, a rich country with a small marginal product could grow faster than the poor African country because it invests in more machines than does the African country. For this reason, if it is true that poor countries tend to have smaller investment rates, it is conceivable that the relationship between growth and level of income is not negative—that is, that there is no convergence.

Regions within a country display a lot more similarities than do countries across the world. In particular, it is reasonable to think that regions within a country have similar saving rates. As already mentioned, if two countries have the same saving rate, then the poorer country will tend to grow faster. Hence, the cross-country evidence can be reconciled with the cross-regional evidence if there are larger differences in saving rates across countries than within countries.

Evidence on the Determinants of the Growth Rate

Robert Barro (1991) expanded the Summers and Heston (1991) data set to estimate cross-sectional equations of the form:

where γ is the vector of rates of economic growth and x1,…,xn are vectors of explanatory variables, which vary across the research. The number of countries in a typical regression was close to 100. One of Barro’s original explanatory variables was the logarithm of the initial level of income, and in a regression like equation (4.6), its coefficient was negative and statistically significant. Barro’s initial interpretation of this regression was that the variables xi were the determinants of long-run economic growth, whereas the initial level of income was a proxy for some “relative income variable” that would capture the different levels of technological progress. Four lessons emerged from Barro’s study. First, education was an important determinant of growth rates. Second, the investment rate was strongly and positively correlated with growth (although the cause of this relationship was far from clear). Third, the coefficient of the initial level of income was significantly negative once other variables were held constant. Finally, different measures of political instability and market distortions also seemed to matter, in varying degree.

After this initial study, the empirical literature blossomed with new results that used the Barro approach. The more robust of the variables from a sample of the studies are summarized below. (The findings are based on Sala-i-Martin, 1997.)

Variables Strongly Correlated with Growth

Political variables. The rule of law and protection of property rights appear to promote growth, and political instability (revolution, military coup, war) appears to hinder it. This means that an important empirical determinant of the rate of economic growth (according to the experience of the past 30 years) is the quality of the government. Governments that maintain the rule of law and guarantee property rights tend to perform better. Thus, an important component of success is government’s efforts to maintain a good political and legal environment to give private investors the incentive to invest.

Market distortions and market performance. Economies with greater market distortions—as measured, for example, by real exchange rate distortions and the standard deviation of the black market premium—tend to have poorer growth performance.

Investment. Countries that invest more tend to grow faster. Saving and investment rates in many African nations with slow growth over the past 30 years have been less than 10 percent of GDP, whereas those rates in many high-growth Asian economies have been more than 20 percent and at times closer to 50 percent of GDP. Economists have emphasized the distinction between equipment and nonequipment investment. Although both types of investment help the economy grow, investment in equipment tends to have a much stronger effect on growth than investment in structures.

Primary sector production. Empirical results tend to confirm some researchers’ argument that a disproportionate prominence of primary products in total exports has a negative effect on growth, since it tends to generate “Dutch disease.” (In Dutch disease, a country with large primary-product exports sees an appreciation of its currency that makes its other exports uncompetitive. This happened to the Netherlands after large natural gas deposits were exploited there in the 1950s.) Countries that are better endowed in the primary sector tend to do much worse in the long run.

Openness. Many people have argued that a secret to growth is the elimination of trade barriers, exchange controls, and capital controls. Economies that are more open to trade and foreign investment are predicted to grow faster. The empirical results tend to confirm this hypothesis. A substantial literature emphasizes the role of openness in promoting growth: openness leads to larger markets, increased international trade, and greater specialization, as well as the adoption of technologies developed abroad and the production and export of goods using the imported technologies.

Type of economic organization. A variable measuring the degree of capitalistic (as opposed to statist) organization in an economy has been found to relate positively to economic growth. In one study, countries were scored from 0 to 5 according to how important private enterprise is in the organization of the economy. The categories and their corresponding values were as follows: 0, statist (Iraq and Ethiopia belong in this category); 1, mixed statist (Egypt, Rwanda); 2, mixed capitalist-statist (Malta); 3, capitalist-statist (Italy, India); 4, mixed capitalist (Greece, Senegal); and 5, capitalist (United States, Botswana). The results suggest that the closer the economy is to capitalism, the faster it grows. The variable is good in explaining the growth of an economy: economies tend to grow faster in countries where there is more freedom.

Education. Countries with high primary and secondary school enrollment rates tend to grow faster. This suggests that the creation and accumulation of human capital have an important role in economic growth.

Macroeconomic stability. Although no single measure of macroeconomic stability (such as low inflation, small public deficits, small current account deficits, or exchange rate stability) appears to be robustly related to growth, the data show that, in the aggregate, poor macroeconomic policies are bad for growth. It may be that no single variable picks up this effect because governments that adopt one bad macroeconomic policy tend to adopt many others at the same time: countries with high inflation tend, for example, to have large deficits and currency depreciation. Thus, although a single bad policy might not limit growth, countries that follow poor macroeconomic policies in general tend to experience slow growth.

Africa. One recurrent finding of studies of the determinants of growth remains to be explained: a dummy variable that takes the value 1 for sub-Saharan Africa and zero for other countries has strongly negative significance in almost all regressions. Africa is predicted to grow very slowly because it scores low on all the variables that are good for growth, and high on all of the variables that harm growth. But even so, the significance of the African dummy implies that Africa has grown even less than its performance on those variables would indicate. That is why the finding of a negative coefficient on the Africa dummy variable is sometimes referred to as a “confession of ignorance.”

Variables Not Strongly Correlated with Growth

Some variables appear unimportant for growth: no measure of government spending (including investment), for example, appears to affect growth significantly. In other words, the size of government does not seem to affect the growth rate much, possibly because government spending needs to be financed with taxes, which have distorting effects. Even if all government spending were productive (and we know that this is not the case, because some types of spending are directed to welfare or social service activities that, however necessary, are not productive), we would still have to finance it with taxes. But higher taxes tend to reduce the incentive to invest, since they tend to reduce the ex post return to investment. Hence, whether a large government is good or bad for growth depends on whether the positive effects of productive investment outweigh the negative effects of the taxes needed to finance such spending.

A variety of measures of financial sophistication also appear insignificant. For example, the inflation rate and its variance do not appear to influence growth much. Although one can arrive at this result through several regressions, the finding does not appear robust. A possible explanation is that some authors consider inflation to adversely affect growth only when inflation reaches very high levels, suggesting a nonlinear relation between growth and inflation. That may be why the negative effects of inflation are not being picked up. But the current analysis allows only for a linear relationship.

Measures of scale effects (measured by total area and total labor force) and of ethnolinguistic fractionalization (supposed to capture the level of internal strife among ethnic groups) also appear to have little impact on growth. Finally, having had a colonial past might be thought to hinder growth, because colonial structures and bureaucracies leave a country in disarray, but there is no evidence to support this hypothesis: former colonies tend to do roughly as well as countries that have never been colonized.

Bibliography

A larger, more complete, and technically more demanding survey can be found in Barro and Sala-i-Martin (1995).

For example, a U.S. resident named Tiger Woods makes a relatively large contribution to world GDP, even though he has only 12 years of schooling. Most people with 12 years of schooling do not contribute as much as Tiger Woods does. This example illustrates that not all the productivity of workers comes from the accumulation of education.

This is exactly true if markets are competitive.

If everyone in the economy were working, the population and the workforce would be identical, so output per capita and output per worker would be the same.

This statement may not be completely true if the country has based its growth on the adoption or adaptation of technologies invented elsewhere. In such a case, growth will stop after all available technologies have been copied.

    Other Resources Citing This Publication