From YalePressWiki

[edit] Chapter 3: What Drives Economic Growth?

Modern economics as a separate academic discipline began with Adam Smith’s The Wealth of Nations, whose central preoccupation was the question: what made economies rich? One of Smith’s most important insights was that specialization, and therefore trade, within and across a country’s borders was critical to growth. Individuals, he posited, would be far better off if each person specialized in what he or she did best and simply bought the things that other people could make more cheaply. In Smith’s view, even if you were a jack-of-all-trades, it would be to your advantage to concentrate on the one or two things you did best because there always would be someone who could do the other things better. Thus, rather than grow your own food, build your own house, or make your own clothes, it would be better to specialize in one activity, or to work for someone else who did, and then buy the rest of what you needed from others. Smith was optimistic about future economic prospects as long as individuals and firms could freely trade with one another across as wide a geographic area as possible (as long as transportation costs did not offset the advantage of trading from afar).

Only a few decades later, Thomas Malthus did his best to turn that optimism around. Malthus is widely known, of course, for his infamous prediction that population would grow faster than the food supply, thereby leading to mass starvation and death. With this one forecast, Malthus did much to cement the reputation of economics as “the dismal science.”

As the few statistics cited in the beginning of chapter 1 demonstrate, Malthus was wrong. And, as we discussed in chapter 2, Malthus failed to take into account the continued advances in the technology of food production that have made it possible to feed more and more people with the same amount of (or even less) land and far fewer people engaged in the production of food. Nonetheless, one could excuse the roughly two billion people in the world who today earn less than $2 per dollars a day for believing that Malthus was right to be so pessimistic. For this reason alone, one would think that economists would have been consistently interested in why some countries grow faster than others, as well as why individual countries grow faster or slower in different time periods. But after Malthus, interest in the topic of economic growth declined among economists and did not pick up again until the era of the Great Depression, when economies around the world not only were not growing but were actually contracting at historically unprecedented rates.

The renowned British economist John Maynard Keynes supplied the solution to the problem at that time. Keynes argued that the classical remedy—waiting for high and rising rates of unemployment to drive wages down to a level where it would be profitable for business firms to begin hiring workers again—would not work, or that it would take so long as to be practically useless. For one thing, there was a downward rigidity to wages; workers who still had jobs resisted efforts by employers to lower wages simply to create new jobs. Equally if not more serious, firms would have little or no interest in hiring any more workers—even at lower wages— without confidence that whatever goods and services they produced or delivered actually would be bought by consumers or other firms. In short, Keynes’s diagnosis of the Depression was that it was caused by insufficient demand for goods and services and could not be cured any time soon by waiting for wages and prices to fall.

A major economic field of study—macroeconomics—was born out of this basic insight. Associated with it was a set of straightforward economic prescriptions. If the private sector was generating too little demand, then government needed to come to the rescue, either by cutting taxes or increasing spending, or both. In other words, when the economy is weak, government deficits can help jump-start growth—but again, from the demand side of the economy. Conversely, if private sector demand growth was too strong, so strong that it was pushing up against the limited capacity of the economy to produce goods and services and thereby causing prices and wages to rise, then one appropriate policy response would be tighter fiscal policy, higher taxes and/or cuts in government spending. This latter problem of inflation would not become evident in U.S. experience until many years after World War II, but it was anticipated in Keynes’s thinking and, indeed, was the corollary of one of his prescriptions for getting an economy out of a recession or a depression (which is, in essence, a severe and prolonged recession).

Keynes’s emphasis on government responsibility for managing the economy—keeping it propped up when private sector demand was weak and dampening it when private sector demand was too strong—has survived him. Although some economists have since questioned the ability or wisdom of governmental attempts to smooth out economic fluctuations, the fact remains that in virtually all capitalist economies, macroeconomic policy management remains a central job of government. Understandably, therefore, to the average investor (and, indeed, the average citizen) economic growth is largely or only a demand-side phenomenon, driven by the growth in private sector and government demand for goods and services.

But although demand is certainly important, particularly in the short run, it cannot explain growth in the long run. Like any machine, the economy at any given point in time has a certain maximum capacity. Over the long run, economic growth is about the growth in that capacity, or what economists often call “potential output,” that is, the amount of goods and services the economy could produce if all its resources, people, and machines were fully utilized. In the 1980s, this focus on potential output was popularized under the rubric of “supply-side” economics and addressed the role that tax cuts play—or were alleged to play—in stimulating growth in economic capacity by encouraging individuals to work harder and to save more.

We will not wade into the controversy that continues to this day about how important taxes are in this process. The important point for our purposes is that supply-side economics was not new. A number of economists had theorized in previous decades about what determines growth of potential output. In this chapter, we want to review briefly what insights they had, then turn to recent empirical studies of growth, and finally conclude with some thoughts about what we believe has been missing from these efforts to understand the process of economic growth.

[edit] Explaining Economic Growth: The Theory

In one sense, understanding how economies grow is like understanding how to make a cake: one must simply find a recipe. Recipes for making a cake include some basic ingredients (sugar, flour, leavening, and so on), some labor (measured in minutes or hours), and some equipment (a mixer and an oven). For economies, there are as many recipes as products and services, but typically all of them require essentially the same three ingredients: raw materials, labor, and machines (also called physical capital).

Actually, there is a fourth ingredient for both cakes and economies: technological change. Just as the mixers and ovens today are more efficient and cook more evenly than those of yesteryear, technological advances in whole economies lead to new products and services that are more desirable than those already on the market, as well as to more efficient ways of generating and delivering all products and services, whether existing or new.

In chapter 1, we boiled down the recipes for economic growth into two broad categories, which we labeled “growth by brute force” and “smart growth.” By brute force, we meant the addition of more inputs—more labor and more capital that will lead to more output, although more capital alone will substantially raise output per worker. Yet one of the basic tenets of economics is that there are diminishing returns associated with the addition of any one factor of production. For example, with a given labor force, adding more and more machines will produce more output, but at a steadily declining rate. So although raising the share of output an economy devotes to both saving and investment can lead to higher growth for a while by providing more plant and machinery, it cannot do so in the long run. Put another way, in the long run, more investment can raise the level of total output but not its growth rate. This is one of many insights of one of the founding fathers of modern growth theory, MIT professor Robert Solow (1956), and of another growth model published at the same time by Trevor Swan (1956).

Our second category—smart growth, that is, technological advance—can rescue an economy from diminishing returns. Steadily equipping any given labor force with better machines or equipment, such as personal computers instead of typewriters, can raise both the level and growth rate of output. Indeed, a central contribution of Solow’s early work on growth theory (for which he was eventually awarded the Nobel Prize) is that technological advance (or increases in total factor productivity [TFP]) is the most important source of growth. Solow reached this finding for the United States using U.S. economic data through the 1950s and estimating an equation linking output to measures of capital and labor (Solow, 1957).^[1] Since the estimated equation explained only about 12.5 percent of the variation in output, Solow attributed the leftover, residual variance to technological change. Subsequent work by the late Edward Denison, for the Committee for Economic Development and later for the Brookings Institution, reached similar conclusions through a somewhat different procedure— growth accounting—which apportions out growth to a number of possible causes (Denison, 1962 and 1974). Other economists have since come to a similar conclusion, that technological change is a key driver of growth (Easterly and Levine, 2001).

The theoretical growth models constructed by Solow, Swan, and others since are shorthand ways of expressing in mathematical terms the relation between certain input variables—labor, capital, and technological advance—and the growth in the output of goods and services. Although abstract, such models can provide useful insights. For example, in one mathematical form, the models imply that responsiveness of output to changes in labor or capital (what economists call elasticity) is equal to the respective shares of labor and capital in overall output. Roughly speaking, therefore, since workers’ incomes typically account for roughly two-thirds of output in most capitalist economies, a one-percentage-point increase in the labor force (from some combination of population growth, immigration, and increases in the participation rate of individuals wanting to work) would, in this model, lead to a 0.67 percent increase in output.^[2]

But even the best mathematical models have their limits, and the post-war growth models were no exception. In the basic Solow Swan model, for example, technological change is considered to be exogenous—something that happens with some combination of serendipity and policies aimed at promoting it (for example, government spending on basic research or legal protection of intellectual property rights). As we discuss below, the statistical studies of economic growth that have been performed over roughly the last two decades are largely aimed at attempting to unravel the mystery of technological change, or what many economists call the Solow residual. Why does the pace of innovation speed up in some periods and in some societies, and why does it slow down at other times and in other places? To be able to answer these basic questions is, at bottom, to be able to explain what can speed up or retard economic growth itself.

A growing number of economists have wrestled with these questions over the past several decades. Most have followed in the model-building and testing tradition pioneered by Solow; we will discuss their efforts in the next section. A few others, however, have taken an entirely different and nonmathematical path, one that stresses the importance of institutions, that is, the rule of law and informal norms that ensure that productive economic behavior will be rewarded. The leader of this institutionalist school of growth is another Nobel Prize winner, Douglass North, although others have contributed to the field.^[3]

Economists who stress the importance of institutions typically point to the enforcement of rights to property (both physical and intellectual), contracts, and limited liability for investors in companies as being among the most important of these rules. Institutions take much time to develop, however, and generally cannot be copied or transplanted wholesale from some societies where they seem to work well into other societies that seem to be sorely in need of them. Instead, the institutions work most effectively, if at all, if they are home-grown. This can be frustrating to policy makers, whose time horizons are typically measured in years to the next election, not in decades—which may explain why the somewhat autocratic leaders bent on achieving economic reform (notably those in Korea and Singapore) have been so successful. The long time lags inherent in the development of institutions also frustrate the ability of economists to test their importance empirically, for lack of available data. But just because the contribution of these institutions cannot easily be validated by standard statistical tests does not mean they are unimportant. On the contrary, economists and policy makers who ignore the importance of institutions in economic growth run the risk of committing the proverbial lamppost fallacy: looking for one’s lost money under a lamppost because that is where there is light, not necessarily because that is where the money was lost.

As readers will see in subsequent chapters, our own thinking on the subject of economic growth has been strongly influenced by the institutionalist school of economic growth. This also explains our mode of argument, which is heavily historical, logical, and even anecdotal rather than statistical. We acknowledge the limitations of our work, which can be fairly described as informed guesswork. Some of our prospective critics (if there are any!) may emphasize the guesswork aspect of our work, but we hope most readers will recognize that our analysis is informed by a substantial body of facts.

[edit] Explaining Economic Growth: The Empirical Evidence

For roughly two hundred years, from the time of Adam Smith up through the contributions of Solow and Denison, the topic of economic growth was largely the stuff of abstract theorizing. All this has changed over roughly the past two decades for a simple reason: the historical data that economists need to run standard statistical tests have been generated and made available by several economists who pioneered this unglamorous, but very important, aspect of the field. Accordingly, growth theory has been elaborated and subjected to a wide number of statistical tests by various economists in recent years, the essence of which we will review now.

Still, even with the best of data—and we will argue shortly that the data here have their limits—economists, like other social scientists, face obstacles that their counterparts in the physical sciences (physics, chemistry, and biology, for example) do not. Physical scientists generally are able to test their theories or hypotheses by running experiments, in which they can test one population that has been subjected to some intervention (such as a drug or a procedure) against a control group to see if that intervention makes the difference that theory suggests. These experiments often generate results very quickly, in a matter of days or months. In the case of highly sophisticated particle accelerators, physicists get results in literally a flash of a second (although it may take a bit longer to analyze the results of smashing atoms at the speed of light). Astrophysicists can also look backward— over many millions of years—by looking into space through increasingly powerful telescopes or probes launched into space to take advantage of the speed of light to find out what certain objects looked like or how they behaved many millions of years ago.

Economists do not have these luxuries for several reasons. For one thing, economists cannot run controlled experiments, with results observable only after a substantial delay, with entire economies, although in some rare cases, social scientists can conduct more modest experiments on selected populations (giving different groups various economic incentives or rewards for certain types of behavior, or providing groups of students different curricula or other educational interventions, for example).^[4] But no government will allow its country to serve as a control group or a guinea pig for a study on what encourages or inhibits economic growth, especially given the long time lags involved in collecting and analyzing sufficient data for economists to draw definitive conclusions. If some policy has at least a reasonable chance of raising growth, governments and the people they serve will or should want to implement it right away, not wait to find out many years later whether it might work (although interest groups in societies that might be hurt by growth-oriented policies, which inevitably create disruption, may be successful in resisting their adoption).

Accordingly, economists are almost always looking backward in an effort to develop policies for the future. They do this by applying statistical techniques to bodies of historical data to sort out one or more variables whose patterns might explain growth. If economists can do that with some grounds for confidence in the results, then they can offer prescriptions to government leaders with at least some hope that what has worked in the past has a reasonable chance of working in the future.

For example, in the case of economic growth, economists seek to find out which ones of some set of “independent variables”—such as capital, labor, and various other factors they believe might contribute to technological change—drive economic growth (which is the “dependent variable,” typically measured by per capita GDP or some variable designed to measure innovation or technological change directly). Once economists know, or believe they know, what factors have been most important in stimulating innovation in the past (ideally, factors over which governments have direct control, like spending on research and development, tax rates on income or sales, or openness to foreign trade and investment, for example), then they have some basis for proffering advice to political leaders that has some grounding in facts, not simply theory or, worse, political or personal bias.

Yet even in this endeavor, economic analysis has its limits. One problem is that in prescribing policies that have worked in the past, economists— and the politicians who listen to them—implicitly are assuming that the economies to which they are applying these policies will continue to behave or operate in the future in fundamentally the same way as in the past, or at least in similar fashion. This is equivalent to saying that the individuals and firms who make up these economies will act in the future much as they have in the past. While this is a plausible assumption, reality may intrude in some way or another, and this possibility at the very least raises questions about that assumption. This is especially true where some event—like a war, a major depression, or a sharp change in political or economic systems (the sudden transition from socialism to some form of capitalism in the former Soviet Union and Eastern Europe, for example)—has marked a sharp break between two historical periods. In such cases, people, firms, and even governments may behave very differently after the break than before.

A second limitation is that the statistical techniques that economists typically use (such as multivariate regression analysis) have their own shortcomings. For one thing, the results they generate are only as useful as the data to which they were applied, a limitation about which we have more to say in the following section. For another, statistical techniques often do not generate consistent or even clear answers, which is a limitation that we believe plagues the statistical work on growth in particular. There is always the problem of omitted variables or influences that really matter but which have not been included in the statistical tests, sometimes unintentionally or, more often, because the data to measure those influences do not exist or are highly imperfect.

And then there is the nagging problem of how to interpret the statistical results. Strictly speaking, regression analysis—which seeks to find the mathematical formula that best “fits” the behavior of some independent variables to the behavior of another dependent variable—usually generates at most what economists or statisticians call correlation. One variable is correlated with another if it moves in roughly the same direction as the other. For example, rainfall patterns are generally correlated with agricultural yields. Or the frequency of sunspots may be correlated with the ups and downs in the stock market. But correlation is not causation. The fact that two variables are highly correlated does not necessarily mean that one causes the other. The hypothetical sunspot example should be proof of that.

This distinction between correlation and causation is critical in social science, and in economics in particular, since political leaders who adopt a policy that economists recommend will generally assume that if they take that step they will get the positive results they desire—that adoption of a policy will cause some desirable outcome, like faster economic growth, to occur. But the regression results on which the policy recommendations rest may not justify such causal inferences. Or even if they do, when the policy is adopted, other forces—within or outside the economy (such as the weather)—may interfere with the experiment. Economists, politicians, and pundits will then debate for years thereafter about what truly caused what. The continuing debate in the United States over the impact of government budget deficits is one example of how controversies can seemingly go unresolved for years.

With these many caveats in mind, we now briefly describe the various statistical tests economists have deployed to unlock the puzzle of growth. As we have already suggested, these tests rest on the availability of statistical data on levels of output and other variables in different countries that might contribute to economic growth. Why different countries? Because the reliability or “confidence” of statistical tests improves as the quantity of data analyzed increases, especially if one wants to test the presence and magnitude and influence of many variables at the same time. As statisticians like to say, the more data they have relative to the number of variables tested, the more “degrees of freedom” they have. When statistical tests are limited to one country, the statistician only has data for that country for a given number of variables of interest over as long a period as they have been collected. In the United States, this is probably since 1950, and be- cause the measures we are interested in are released annually, the data base can cover about fifty-five years or data points, at maximum. For other countries, the time series—the available set of statistics—may be even shorter. But when time series data for different countries are pooled together, the number of observations is greatly magnified and so is the power of the statistical tests, at least in principle.

These fine points of statistical testing were not an issue in the first generation of post-Solow statistical tests of growth, which used the data series on output, output per worker (or work-hour), and output per capita that were compiled by Angus Maddison (1982), who is one of the leading figures in the highly specialized field of cross-country data collection, and Matthews, Feinstein, and Odling-Smee (1982). The tests asked a seemingly simple question: have standards of living, as measured by productivity (output per hour of work) or output per capita in different countries, converged over time? In other words, do advances in leading countries spill over to a set of follower countries, through exports of goods, capital, and ideas from the advanced guard to the followers? And does this spillover and imitation process happen in such a way that the follower countries catch up to the leaders by growing more rapidly for a time (perhaps by investing and saving greater fractions of their output while adopting the leaders’ technology)?

Several early studies of different groups of countries confirmed that this had indeed happened. Matthews and colleagues found it to be the case over the 1870–1973 period for seven countries that were industrialized by the early 1970s (Matthews et al., 1982). One of the authors of the present volume reached a similar finding, using Maddison’s data, for a larger sixteen-country group over a slightly longer period, 1870–1979 (Baumol, 1986), but found that convergence had not occurred among the much larger set of countries for which the requisite data were provided by Summers and Heston (1991). That is, for the converging countries one could explain the growth rate of their productivity over a little longer than a century almost entirely by knowing only one thing: their initial level of productivity in 1870. If a country started out far behind the productivity leader (which, in 1870, was Australia), it grew much more rapidly than if its productivity level was already at or close to the frontier. This simple proposition, that the further behind the leader a country was in 1870, the faster it grew later, explained the very rapid growth of Japan, Sweden, France, and Germany over this long period, and the relatively slower growth of the United Kingdom and the leader itself, Australia.

Yet even the author of one of these studies cautioned that too much should not be read into this apparent finding of convergence, noting that the 1870 productivity levels were measured with considerable error and that Maddison constructed them using a method of backward extrapolation that would have biased the finding toward convergence (Baumol, 1986, 1076). Baumol could have added that the 100 years covered by the data series included two world wars, and that after World War II, in particular, one of the countries in the data set (the United States) provided ample financial and technical assistance to both Europe and Japan that should have enabled them to catch up to U.S. productivity levels after the war.

Thus, a more interesting question is whether, since World War II, convergence has occurred among a larger group of countries, including many that were once or still are less developed. Baumol (1986) used a data set of per capita incomes (which provide a rough approximation to productivity data) compiled by University of Pennsylvania professors Robert Summers and Alan Heston for that larger group of countries (these statistics have since become the data set of choice of a large body of researchers).^[5] Unlike a similar set of statistics assembled by the World Bank at that time, the Summers and Heston data for output in different countries are adjusted for differences in the relative purchasing power of currencies, not just for differences in exchange rates between countries. This distinction is very important because the prices of the same commodities or services may be very different in different countries. Measures of output that do not take purchasing power differences into account do not capture the true disparities in standards of living among countries.

When Baumol analyzed the Summers-Heston data for seventy-two countries over the 1950–80 period, he found a very different set of results from those he had reported for the narrower set of industrialized countries over a previous longer period: for the entire group of countries, convergence had essentially disappeared. Indeed, there was even a mild positive relationship between a country’s initial level of productivity and its subsequent growth: that is, the countries that were richer to begin with tended to grow a bit faster than other countries. Baumol did find, however, various country clusters where convergence seemed to take place within (but apparently not across) those groups, between 1950 and 1980, among the (then) centrally planned economies (the Soviet Union, China, and Eastern Europe) and again among the industrialized countries. This convergence clustering did not appear to take place within developing economies as a whole, although we know from subsequent experience that at least one group of developing countries, notably those in Southeast Asia, has displayed rapid convergence among themselves and relative to the world’s leading countries.

Figure 4. Cross-Country Initial GDP per Capita vs. Average Growth Rates, 1980–2000. Source: Alan Heston, Robert Summers, and Bettina Aten, Penn World Table Version 6.1, Center for International Comparisons at the University of Pennsylvania (CICUP), October 2002 (106 observations: countries having complete data set). Available at http://pwt.econ.upenn.edu/php_site/pwt61_form.php.

Baumol’s finding of a lack of overall convergence in the postwar era through 1980 has continued to hold up. Figure 4 displays the growth rates in per capita income over the 1980–2000 period, together with initial per capita incomes for 106 countries in a more recent version of the Summers- Heston data set (with coauthor Bettina Aten). The figure clearly fails to support the convergence conjecture (the tendency for initially poorer countries to grow more rapidly than countries with initially higher incomes, as catch up would require). Indeed, if anything, simple visual inspection of figure 4 suggests that initially richer countries may have grown faster than initially poor countries, a result consistent with figure 1 in chapter 2.

If countries are not converging in their standards of living, then what explains the continuing economic differences across countries? Attempts to answer this simple, but vital, question have spawned a separate industry within the economics profession. These research efforts would not be possible, of course, without the Summers-Heston-Aten data, which contain information only on the variables to be explained—the levels and growth rates of output (per capita or per worker or per work hour). A variety of data sources have since arisen for variables that might “do the explaining,” such as measures of physical and human capital (labor force and education levels), institutional variables (law and corruption, for example), international trade, financial indicators, government and private investment in research and development, and measures of climate and geography provided by such organizations as the World Bank, the United Nations, and individual researchers.

The search for answers to the growth puzzle—and, specifically, the causes of the variation in the Solow residual (the rate of technological advance)— has engaged some of the most distinguished figures in economics, including two Nobel Prize winners (Stanford University’s Kenneth Arrow and Robert Lucas of the University of Chicago), as well as many leading lights in the profession (such as Robert Barro, Greg Mankiw, Andrei Shleifer, and Edward Glaeser of Harvard University; Jeffrey Sachs and Xavier Sala-i-Martin of Columbia University; Stanford University’s Paul Romer; Barry Bosworth and Susan Collins from the Brookings Institution; Yale University’s William Nordhaus; Ross Levine from the University of Minnesota; Steven Durlauf from the University of Wisconsin; Elhanan Helpman of Harvard and Tel Aviv Universities; and William Easterly of New York University, among others). It is difficult (if not impossible) to summarize all of this work in a short space, but certain broad generalizations are possible. (Readers who want a more thorough guide to this research, and indeed to the evolution of the discipline of economics in general, are strongly encouraged to read Helpman, 2004, and Warsh, 2006.)

First, many of the economists who have conducted these studies now believe that if the right model of growth can be identified, it will show that there is a fundamental dynamic toward conditional convergence. That is, if one controls for the right variables, it remains true that countries with low initial levels of productivity will have faster growth in productivity and economic output than richer countries. Of course, this conditional convergence process may occur slowly—or rapidly—depending on one’s patience or expectations. Some of the cross-country statistical tests suggest that on average throughout the world, the gap between the richest and poorest countries closes at the rate of about 2 percent annually.^[6] At this rate, it takes about a generation (thirty-six years) for a lagging economy to close half the gap between its per capita income and that of the leading economies. Those looking for miraculous turnarounds in a short span of time will be disappointed by this figure. But for others, the prospect for closing this much of any income gap in just a generation may seem remarkable.

Second, despite the substantial statistical investigations of growth that have been undertaken over the past two decades, economists who believe that the statistical work has helped to unlock the growth puzzle (as we will note shortly, this includes most, but not all, economists who pursued this line of work) still fall broadly into two groups. In one camp are those whose views adhere closely to the assumption built into the initial Solow- Swan growth model: that technological progress is primarily determined by forces—such as climate (which affects the rates of disease), geographic location (which determines costs of transportation and thus propensity to trade), and institutions (which are man-made but may take decades, if not centuries, to change)—that are outside the economic system and over which policy makers have little or no immediate control (see Bosworth and Collins, 2003, and Frankel, 2003). To this list some add culture, which is difficult to incorporate in formal statistical tests, but which some economic historians argue is the dominant driving force behind growth, a subject we will explore further in chapter 5.

In a second camp are economists who contend that the statistical studies lend support for the view that active policy intervention, in the short to intermediate run, can accelerate the growth of either or both labor productivity (output per worker or hour of work) or technological advance (measured by additions to output that arise even if labor and capital investment are held constant). Growth-enhancing policies can include governmental decisions to open up an economy to more trade and foreign in- vestment, to support more research and development (through direct spending or tax incentives), to increase human capital by broadening the availability of primary and secondary education, and to conduct sound macroeconomic policies (avoiding consistent and large budget deficits or inflationary monetary policies). At its core, economists who fall into this second camp are more optimistic about the ability of governments to encourage more rapid improvements in living standards than what might otherwise occur naturally.

In the technical language that economists often use, economists in this second camp are suggesting that technological advance is endogenous, that is, it is determined by forces within the economic system itself rather than such exogenous factors outside the system as climate and geography. Much of the intellectual impetus for this way of looking at growth was provided in the 1980s through the work of Paul Romer, then at the University of Chicago and currently at Stanford (Romer, 1986).^[7] Romer (and others who followed in his wake, including Robert Lucas and William Nordhaus) built on the earlier insights of Kenneth Arrow (1962) and Eytan Sheshinski (1967), who believed that the ideas that underpin technological advance are the unintended by-products of investment in new equipment that spilled over and thus benefited the rest of the economy. In this way, more investment would lead to more technological advance, suggesting that the latter somehow depended on the former.

One unspoken policy implication of this view that investment generates beneficial spillovers is that governments do indeed have a potentially important role to play in encouraging growth. To the extent that governments can stimulate investment, through tax incentives in particular, and also to the extent that they can encourage more domestic saving, which should enlarge the pool of funds available for financing investment (thus bringing down its cost), government can enhance the long-run prospects for growth. This implication sharply departs from the investment pessimism of the Solow-Swan growth model, which implies that additional investment eventually stops adding to growth because of diminishing returns. But if investment can actually enhance technological advance, this pessimism may be misplaced.

In his modeling, Romer went one step further, observing that technological advances often were not simply by-products, but were the objects of economic activity itself—the products of deliberate investments of time and money by individuals and firms seeking to improve on what already exists and ultimately to commercialize any successful results. In this sense, business firms’ investment in knowledge creation is analogous to their investment in new equipment that promises to make employees more productive. But unlike investment in a new machine, which has more or less predictable productivity-enhancing consequences, investment in knowledge discovery (and, if successful, its subsequent commercialization) is fraught with uncertainty. It is not surprising, therefore, that the statistical work that has gone into trying to explain the sources of technological advance has come up with varied answers, and some controversy over certain variables (such as openness to foreign trade) still continues.

One other policy implication stands out from Romer’s work, however: that technological advance is not likely to occur, at least in economies at the frontier where imitation is not an option, unless those who undertake it are assured of some reward. Hence the importance of imperfect competition, or something other than the perfectly competitive ideal where so many firms are making an identical product that they compete away any excess profits. If some extraordinary profits are not available to the individuals or firms who leap into the unknown, taking the risks to develop and commercialize something new, then technological advance would not occur. That is why economists typically defend the importance of an effective system of intellectual property rights that confers monopoly status on innovators for some limited period of time, or why market structures should not be perfectly competitive in dynamic industries, at least in the short to intermediate run. Continuing technological advance, however, competes away any short-run profits so that over the long run they disappear. ^[8] We draw on these key insights in our discussion of what is essential to entrepreneurial capitalism in subsequent chapters.

Third, there seems to be some rough consensus among economists in both camps that institutions—such as well-enforced property rights and the absence of corruption—are important to growth. But debate still continues over whether Anglo-Saxon or so-called civil code legal systems are more effective in advancing growth.^[9] The key challenge with respect to institutions is how best to create them. Must countries wait decades, or even centuries, for institutions to evolve naturally? An open question is whether the right institutions can be manufactured or transplanted in short periods of time.

Fourth, included implicitly if not explicitly in the view that institutions matter is the rough consensus that one of those institutions is the development of human capital, which is the steady improvement in the skills of the labor force. In their empirical work, economists have typically measured human capital by years of education, although they admittedly recognize this to be an imperfect proxy for skills. A number of statistically based studies of growth find a strong link between human capital measured in this fashion and economic growth.^[10] That link can arise through two channels. A more educated workforce has a larger effective labor supply, since an hour of work by a more skilled individual is equivalent to more than hour of work supplied by an individual with lesser skill. In addition, as a society’s workforce becomes more educated, the greater is the likelihood that some of its members will contribute to technological advance in some way, by inventing or commercializing inventions or somehow assisting others who do. Here the possibility of reverse causation constitutes a key problem: may things not work the other way, with growth providing the resources needed to expand education so that growth stimulates education rather than the other way round? The answer is far from certain.

Finally, the debate is perhaps most contentious over the role of foreign aid: whether it enhances, has no effect on, or even may detract from growth. We will discuss this subject at length in chapter 6.

[edit] Limits of Statistical Analyses of Growth

Laymen and political leaders can be forgiven for wondering how very smart economists can analyze seemingly the same bodies of data and come up with very different conclusions about the impact of such governmental policies as foreign aid (among others). Do these statistical tests affirm nothing more than the old saw that there are three kinds of untruths: lies, damn lies, and statistics?^[11]

One answer is that the economists who have argued over the role of aid have not used the same bodies of data, nor the same models to analyze or test them. Another reason for the differences is that analysts have conducted statistical tests over different time periods, examining different sets of countries. Indeed, the mini-industry of economists running “cross- country regressions” has grown as new economists come into the field, finding or constructing new data series to add to those already available.

To broadly generalize, a wide range of results has been obtained from the statistical tests that have been reported in the leading studies. Essentially, one can pretty much find whatever result or results one is looking for, depending on what variables, countries, and time periods one wants to include in the regressions. This state of affairs is hardly comforting to policy makers and others outside the profession. But it is the reality, and, to some extent, it should be expected. After all, a number of the data series constructed to represent some of the more qualitative variables thought to influence growth—such as the “rule of law,” corruption, and openness to trade, among others—are indexes compiled either by the researchers themselves or some organization or body interested in the subject (such as Transparency International, a nongovernmental organization that measures corruption). As a result, these data series have an element of subjectivity that is not present in the more objective variables, such as investment expenditure and hours worked (although even these standard variables have their own measurement problems, especially for developing countries, where resources for economic data collection are less plentiful than in richer countries).

It is not surprising, then, that some of the economists who have carried out these statistical tests have questioned their usefulness and reliability. Ross Levine and David Renelt were early skeptics (see Levine and Renelt, 1992). More recently, Easterly has suggested that no standard variables, even including such theoretical stalwarts as investment in equipment, are consistently and reliably linked to economic growth (Easterly, 2001). But Easterly and Levine are in the minority of economists in this area. Most other economists who have studied growth believe, to one degree or another, that at the very least the statistical tests help identify which variables contribute to growth, although admitting that much uncertainty remains about the relative and absolute magnitudes of each contribution. It is difficult to believe, for example, that investment in physical and human capital has made no difference toward increasing output. Similarly, we know at the very highest level of abstraction that incentives matter for growth, as Easterly recognizes. It cannot be an accident that countries that have allowed individuals and firms to own their own property and to reap the rewards of their efforts have enjoyed much more prosperity than cen- trally planned economies where individuals and firms did not enjoy these rights. The challenge for economists, policy makers, and citizens around the world is to see if more definitive statements can be made about the factors that are most important for growth. We take up that challenge in chapters 4 through 8 of this book.

[edit] Growth and the “Washington Consensus”

In 1989, well before most of the empirical tests of the determinants of economic growth were conducted, John Williamson (an economist who has worked at both the World Bank and the Institute for International Economics) attempted to resolve the growth puzzle by way of another technique. He asked a number of economists and policy experts in Washington, D.C. (including people working for think tanks, the United States government, and the international financial institutions and whom he thought were expert in economic growth), what policies they thought would contribute most to growth in Latin America in particular (Williamson, 1994). The top ten answers are displayed in table 1 and have since come to be known as the “Washington Consensus” set of policy prescriptions.

In the 1990s, the Washington Consensus became more than just a list compiled by one economist. It evolved, largely by accident, into a recipe for growth and financial stability that the world’s major international financial institutions, especially the International Monetary Fund, imposed in the 1990s on a number of developing countries that required bridge financing to enable them to weather financial crises (such as suddenly falling exchange rates, shortages of domestic currency reserves, and shaky financial systems). Presumably, officials thought these conditions were necessary for both growth and stability for at least two reasons.

First, many of the policies that made up the Washington Consensus— fiscal discipline, open markets, competitive exchange rates, and privatization, among others—already were largely in place in developed economies. If the policies were good for the rich, then by implication they must be good for the poor. Further, presumably some participants in Williamson’s survey listed policies they thought foreign investors were most interested in seeing before committing their funds to developing countries.

The Washington Consensus list has since become a topic of much discussion and controversy among economists and policy makers in developed and developing countries alike. While some notable scholars and policy makers generally have supported the policy prescriptions on the list (see Fischer, 2003), others have argued that, in the fifteen or so years since the list was compiled, experience has not borne out the validity of its prescriptions (Rodrik, 2003). As Williamson himself has noted, the regions of the world that have done the most to stabilize, liberalize, and integrate their economies with the rest of the world economy were Latin America and the transition economies of Eastern Europe and the former states of the Soviet Union (Williamson, 2004). Yet the growth record of Latin America since the early 1990s has at best been relatively poor, and the growth of the transition economies has been uneven.

At the other extreme, the fastest growing economy of the past twenty years has been China, which admittedly has moved in the pro-market direction suggested by the Consensus list, but only in a gradual fashion. Nonetheless, in criticizing the list, Harvard University’s Dani Rodrik has asked a provocative question: if the best economic minds of the late 1970s had been surveyed about what policies China should have adopted to stimulate economic growth, they almost certainly would have given some variation of the “Big Bang”—that is, the simultaneous adoption of all of the reforms on the Washington Consensus list. Yet, as we will discuss in more detail in chapter 6, China pursued a very different course with much success, retaining its state-owned enterprises but gradually encouraging them to shrink while at the same time privatizing the Chinese economy “at the margins” by gradually allowing individuals to own their own plots of land for growing crops and allowing villages to own and operate new firms. In chapter 7, we will suggest a similar incremental strategy for promoting entrepreneurship in Europe.

One unfair, though highly publicized, criticism of the Washington Consensus is that Argentina, which was supposed to be a shining example of the success of the Consensus policy prescriptions, suffered one of the worst financial crises of any country in the world in 2001. But this criticism is misplaced. Again, as Williamson (2004) and others have noted, Argentina may have followed some of the prescriptions on the list—notably, privatization, openness to foreign direct investment (until recently, Argentina had the largest share of foreign banks of any Latin American country) and property rights protection—but ignored two other critical items on the list, fiscal discipline and a competitive currency.

Nonetheless, the Argentine and Chinese experiences, among others, highlight one of the central problems of the Washington Consensus list. The list provides no guidance to countries about the relative importance of the different prescriptions on the list or about their timing or sequence. In fairness, that was never the point of Williamson’s exercise and he himself has since expressed some surprise (and regret) about the extent to which the list has become the centerpiece of debate in economic policy circles around the world. Furthermore, although it was not Williamson’s intention, the IMF and others confused the purpose of the list. Over time, the policy prescriptions came to be viewed as more essential because of their contribution to financial stability (and some of them, such as fiscal discipline and competitive exchange rates, surely are) than for sustained economic growth.

Accordingly, any consensus about the right set of policy prescriptions for growth in particular has broken down. Indeed, analysts since have moved in two opposite directions. Williamson, together with Peruvian economist Pedro-Pablo Kuczynski (2003), has proposed a sharp narrowing of the list to just four key factors:

1. Policies aimed at avoiding financial crises, especially by avoiding fixed exchange rates, which clearly can derail a country from its long-run growth path for a very long time;
2. Liberalization of domestic markets, not just product markets (by lowering trade protection measures such as tariffs) but also labor markets, which impede the growth of rising industries and firms and inhibit the necessary shrinkage of uncompetitive industries and firms;
3. Strengthening of domestic institutions that foster growth, an insight that the two economists assert was one of the most important changes in the thinking of development economists in the 1990s; and
4. Recognition that the distribution of economic rewards is a subject that cannot be ignored when a country is trying to promote growth, if only because highly inequitable distributions of income can give rise to political pressures that inhibit or defeat growth (in this regard, the two analysts put greater weight on assuring widespread educational opportunities than on redistributive tax policies).

Rodrik, a noted critic of the original Washington Consensus, proposes moving in a very different direction: augmenting Williamson’s initial list with another ten factors that he believes are central to growth. Table 2 lists Rodrik’s ten additional policy prescriptions.

We agree that many, if not all, of the additional items on Rodrik’s list are desirable, not just for growth but also for financial stability and the equitable distribution of income. But the problem with lists of this type is that they give readers, let alone policy makers, no sense of proportion. The question is, given the limited attention spans of leaders and their citizens, as well as the many competing pressures and objectives, which items on the original or additional lists should policy makers implement first? The lists do not provide answers to this vital question.

Indeed, as we suggested in chapter 1, laundry lists of desirable policy prescriptions can be very problematic, especially if they do not provide some sense of priorities. Faced with a daunting lineup of ten or twenty “musts,” policy makers can easily suffer from information or obligation overload. Feeling that they must make progress toward carrying out most, if not all, of the prescriptions, political leaders and their advisers can too easily be tempted to throw up their hands and say “it’s not possible” and ignore the lists altogether. Like students who are given a mass of assorted facts to memorize but have no structure or context in which to place them, the consumers of these policy lists may simply look at them, perhaps memorize the items for a short while, but then quickly forget them when confronted with the everyday challenges of having to run governments and meet the unceasing and often conflicting demands of their citizens.

In short, while there may at one time have been a consensus at least among Washington-based policy makers and economists about policies that are most conducive to growth, that consensus no longer exists. This is evident among the policy analysts (such as Williamson and Rodrik, among others) who write about growth, as well as in the different, and sometimes inconsistent, results of the growing body of statistical studies that attempt to explain the great difference of patterns of growth among countries.

[edit] The Four Faces of Capitalism: A New Way to Look at Growth

The time is ripe, in our view, for some fresh thinking on the subject of growth. In particular, may there be a different way of thinking about this vital subject that policy makers in various countries can actually use to accelerate the pace of improvement in living standards of their populations? We believe the answer is yes, and that is what we undertake to supply in the rest of this book.

We begin with the fundamental proposition that economies are complicated systems that cannot be reduced to one or two central driving forces, and cannot be turned around by applying one or even a few of the policy prescriptions various development economists or institutions have recommended over the years. To return to the analogy in chapter 1 of the economy as a well-oiled growth machine, the economic machine has a number of parts that are interconnected and that work together. Likewise, if economies are to grow at their maximum possible rate, they must have in place at least some elements of four basic characteristics that we outlined in chapter 1 and will elaborate further in chapter 5.

We say “at least some elements” because the specific policies that are appropriate will vary for different countries at different times. Context, culture, and history all matter. There is no single detailed blueprint that can or should be imposed on every country. The fact that various countries have achieved rapid growth rates with somewhat different institutional structures is testament to that fact. Yet before we address these four key characteristics of the well-oiled growth machine, we believe it is useful to examine growth through a different lens. Specifically, in our view, one can learn much about what it takes for economies to generate sustained growth by keeping in mind what we believe are the four broad types of capitalism that have been and currently are in place in different parts of the world.

These archetypes of capitalism admittedly are very rough generalizations. Furthermore, few economies fit neatly into any one category. More commonly, economies possess different elements of these archetypes at any point in time, and the composition of these elements varies over time. Even more to the point, these archetypes are not handed down from some higher authority, though there is some cultural inertia behind any one. History shows, however, that through deliberate actions, sometimes with unintended consequences, economies can move from one archetype to another, and in shorter order than many people may commonly believe.

Having these archetypes in mind serves as a useful reference point for our discussion later about how countries, through their leaders, can in fact choose different paths to growth. Ultimately, however, we will argue that there is one path—actually, the right blend of two of these archetypes— that promises the most rapid and sustained path for growth.

[edit] Notes

1. ↑ Solow built on earlier growth models developed by Roy Harrod and Evsey Domar, and by Nicholas Kaldor, that emphasized the difficulty economies had in staying on their long-run potential growth path. The Harrod-Domar model, in particular, implied that economies constantly were poised on a knife edge between growth and collapse, which Solow and others have since rejected as unrealistic. For a useful survey of the growth models of the 1950s and beyond, see Solow, 1994.
2. ↑ A model that has this characteristic is called a Cobb-Douglas production function, named after economists Charles W. Cobb and Paul H. Douglas (the latter was a United States senator from Illinois after having had a distinguished academic career as an economist).
3. ↑ See, e.g., North, 1981, 1990, and 2005; North and Thomas, 1973; and Baumol, 1952.
4. ↑ There are ethical and other technical issues associated with conducting such experiments. For one early guide to this subject, see Rivlin, 1971.
5. ↑ The Summers and Heston data project stems from an effort by them and the late Irving Kravis (also of the University of Pennsylvania) that started in the 1970s, expanded in the 1980s, and has since come to be known as the Penn World Tables (see Heston, Summers, and Aten, 2002).
6. ↑ These studies are summarized in Barro and Sala-i-Martin, 2004, 14.
7. ↑ Romer advanced his theory of endogenous growth in his Ph.D. dissertation and has since refined it in a number of ways. He reviews much of this work, as well as the work of others who have followed in a similar vein, in Romer, 1994.
8. ↑ For a more formal analysis of this “intertemporal price discrimination,” see Baumol, 2006.
9. ↑ For a thorough review of the studies on each side of the debate, see Dam, 2006.
10. ↑ The most prominent study to reach this conclusion is Barro and Sala-i-Martin, 2004. See also Bosworth and Collins, 2003.
11. ↑ This saying is usually attributed to the British statesman Benjamin Disraeli and was popularized in the United States by Mark Twain.

This site made possible through the generous support of the Ewing Marion Kauffman Foundation