Sie sind auf Seite 1von 37

JOURNAL OF

Journal of Economic Dynamics and Control 21 (1997) 205-242

Economic Dynamics & Control

The problem of population and growth: A review of the literature from Malthus to contemporary models of endogenous population and endogenous growth
Isaac Ehrlich* , Francis Lui'Department of Economics, State University of New York, Buffalo, NY 14260, USA ^Department of Economics,
HongKong University of Science & Technology, Kowloon, Hong Kong 1

Abstract This paper deals with the evolution of the literature on the problem of population and growth from the classical period to the recent literature on endogenous growth and development. The 'problem' concerns two distinct issues: 1. how to explain the observed covariation of the levels and rates of growth of per capita income and population size over time and space, and 2. how to improve the human condition represented by these variables through an accommodating social policy. The evolution of the literature we survey is reflected by the progressive treatment of key variables as endogenous, rather than exogenous to the growth process. It is also reflected by a shift from the historical concern about population explosion, and its implications for growth, to the more recent concern about the association between growth and population implosion in many developed countries.
Key words: Economic growth & development; Population; Fertility; Longevity; Human capital JEL classification: O10; O40; Jll; J13; J24

1. Introduction The literature on population and economic growth is about as old as economic science itself. Although the main focus of Adam Smith's inquiry has been

1"Corresponding author.
0165-1889/97/S15.00 1997 Elsevier Science B.V. All rights reserved SSDI 0 1 6 5 - 1 8 8 9 ( 9 5 ) 0 0 9 3 0 - T

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 3

the conditions governing wealth accumulation, or growth in aggregate production capacity, he and even others before him realized that the relevant growth measure concerns per capita, rather than aggregate output. Malthus dramatized the idea by identifying population as potentially detrimental to growth. Since that time, work on growth and development has been inextricably linked with population economics. This survey is restricted to studies concerning the 'problem of population and growth'. This term refers to two distinct but related issues. The first is the problem of explaining the observed covariation of per capita income and population levels, q = Q/N and N, or their respective rates of growth, g and n, over time and space. The second concerns the associated policy problem: how to improve the human condition represented by the levels of q or g and n through an accommodating public policy. Also, since population growth is determined by the difference between contemporaneous rates of fertility and mortality in the population, the twoprong 'problem' concerns the relationship between q or g on the one hand, and the two components of n on the other. Our survey follows the evolution of the main paradigms offered by economists pursuing these problems. This 'evolution' is reflected largely by the progressive treatment of key variables as endogenous, rather than exogenous to the growth process. The way different models identify key variables as either exogenous or endogenous affects both the insights they provide about the causal relationship at work and the role or relevance of government intervention. We start with the Malthusian theory of population and later refinements of it, which are known as the 'classical model', in Section 2. In this model, growth in food production is determined exogenously, and the long-run rate of population growth must simply adjust to it because of the tendency of population to explode at a dismally low level of consumption. Per capita income is trapped at a corresponding dismal level. In the short run, however, the model predicts a positive association between deviations of per capita income and the rate of population growth from their long-run values. In Section 3 we review empirical data bearing on the secular and cross-sectional association between levels or rates of growth of population and per capita income. The fact that these data, on the whole, are inconsistent with the basic propositions of the classical model may explain both the decline in its popularity over time and the revival of interest in a more systematic formulation of population and growth theory in the last few decades. Section 4 traces the start of this renewed interest. The first important breakthrough is associated with the development of the 'neo-classical growth model' by Solow (1956) and others. In this model the growth process is endogenized while population growth is taken to be largely exogenous to the economy. Population growth affects the level of per capita income, but not its long term rate of growth, which is entirely controlled by the exogenous rate of technological growth. The model does not offer a reformulated theory of population

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 4

based on explicit micro foundations. This task starts with the seminal work of Becker (1960) concerning fertility behavior. The literature following these important contributions has progressed along two main lines of inquiry. The first concerns a more complete treatment of population as an endogenous variable in the context of both static and dynamic settings, which we take up in Section 5. This literature has shown the relationship between economic variables and population to be a complex outcome of factors such as the incentives for having children, the cost of quantity vs. 'quality' of children (i.e., the amount of parental investments in the human capital of each child), the efficiency of private capital markets and intergenera-tional transfers within the family, and the possible existence of multiple equilibria. This branch of the literature still considers growth itself to be an exogenous variable along the lines of the neo-classical model. The second branch, which we survey in Section 6, has proceeded with the task of endogenizing both population and growth within a unified model of growth and development. The causal relationships governing the covariance of population and economic growth go neither from population to growth nor vice versa - both are determined as a result of differences in the initial conditions or changes in basic parameters of the model which may trigger the dynamic growth process. This literature, which is still developing, provides new insights concerning the problem of population and growth in both its positive and normative aspects. Recent studies concerning key policy issues, such as population control policies and mandated old-age income protection schemes and their relevance for the growth process, are then surveyed in Section 7.

2. Malthus and the classical theory of population and growth A key principle of Malthusian population theory is the dependence of population growth on the economy's material conditions, especially food supply. In Malthus (1798), mankind's biological capacity to reproduce is assumed to exceed its physical capacity to produce. Without restraint, population would increase geometrically, but food production would increase only arithmetically. Since survival requires a minimum level of consumption, population growth would eventually be checked by the growth in food production, and per capita consumption would fall to a dismally low level - an outcome commonly called the 'Malthusian trap'. Deviations from this trap could occur only by negative, or preventive checks on the birth rate through 'moral restraint' (the postponement of marriage due to fear of hunger) or by positive checks on the death rate through 'misery and vice' (the periodic outbreak of wars, pestilence, and disease). Another important, but often neglected aspect of Malthusian theory is its assumed motive for having children. Children are not treated as objects of

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 5

psychic rewards for parents, but as capital goods yielding future labor services, which are produced at constant costs. An increase in the demand for labor would generate a stream of expected returns in excess of costs and consequently result in a rise in the birth rate (see Blaug, 1962, p. 78). This, then, is the source of possible short-run deviations from the population's equilibrium level. The Malthusian paradigm can be restated more formally in terms of both its static and dynamic components. The long-run population supply schedule is assumed to be infinitely elastic at an arbitrarily low wage level, w*, commonly identified with 'subsistence' (but see Becker, 1988); hence the 'subsistence theory', or 'iron law', of wages. The static demand schedule, representing labor productivity at food production, can be depicted as an inverted [/-shaped curve intersecting the horizontal supply schedule from both below and above. Under static conditions the stable equilibrium population level, N*, is determined at the point where the demand schedule intersects the supply schedule from above. (The point where the former curve intersects the latter from below would be an unstable equilibrium.) Any level of wage, or per capita income Q/N, higher than w* would diminish the negative checks on marriage, hence fertility, while any level below w* would trigger the positive checks on morbidity and mortality. These positive and negative checks are not effective in the long run, however. The ultimate restraint on population is that people cannot live either below or above subsistence for too long. Malthus's dynamic equilibrium can thus be deduced as a steady state in which the population's growth rate is determined by the arithmetic rate of increase in food supply, which determines the location of the demand curve for labor. Malthus's ideas were hardly novel. Adam Smith (1976, Book I, Ch. 8) already outlined all the fundamental propositions of the Malthusian theory. Indeed, as Schumpeter (1954, pp. 254255) pointed out, the same propositions can be traced to a much earlier economist - Giovanni Botero. In Botero (1589), populations tend to increase to the full extent made possible by human fecundity, and their actual increase is limited, therefore, by the limited possibilities of increasing the means of subsistence. Malthus's negative and positive checks are also found in Botero. Furthermore, the 'Malthusian trap' hypothesis - that per capita income tends to stay at the subsistence level because of the tendency of population to explode - is ascribable not just to Botero. David Hume also alluded to it before Malthus (see Rostow, 1990). Malthus's main innovation may thus have been his mathematical rules of geometric and arithmetic progressions. But even these had already appeared in the works of Petty and others several decades earlier (Schumpeter, 1954, p. 255). Whatever its originality, Malthus's voluminous work has had a unique influence on the major classical economists following him, and even on contemporary writers. Ricardo and his students adopted Malthus's horizontal long-run labor (and population) supply schedule, but formalized more systematically the 'demand side' of the equation. Malthus did not provide an adequate explanation

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 6

of why food production can increase only arithmetically. The Ricardian model ascribed this to the scarcity of land and the law of diminishing returns to labor farming fixed land. By assuming that labor and the working capital required for farming always combine in fixed proportions, the model deduced that population would continue to expand just as long as the residual profits of capitalists (but not the rent on land going to landlords) remained positive, fueling additional investments. This would be the case as long as the marginal product of labor on fixed land were above labor's subsistence pay. While the Ricardian model recognized the possible role of technological progress and industrialization in shifting upwards the demand schedule for labor, such shifts were not presumed to modify the iron law of wages unless workers' expectations concerning the 'subsistence' or 'natural' wage were themselves modified in the process. John Stuart Mill's ideas were also rooted in this paradigm. He accepted Ricardo's argument that, ceteris paribus, a market wage rate higher than the 'natural rate' would induce an increase in population, discourage investment, and consequently reduce the demand for labor. This would eventually bring the market wage rate back to its natural rate and population would decline to its equilibrium level. Mill (1965) also proposed that the way to avoid 'an over peopled state' or lift workers wages above misery was to implement a public policy of birth control and to expand efforts on popular education. Even Marshall (1930) echoed Malthus's concept of negative checks by saying that society should exercise pressure on the individual by religious, moral, and legal sanctions, sometimes with the object of quickening and sometimes with that of retarding the growth of population to achieve progress. Malthus's pessimistic predictions have been at odds with historical evidence concerning population and growth (see Section 3). Nonetheless, it is relevant to assess the more lasting effects of the Malthusian or 'classical' model on the development of modern population and growth theory. Perhaps the most important of these has been the treatment of population growth as endogenous to the economy, resulting both from the prevailing economic conditions and the motives for having children. This approach has been revived and greatly developed, however, only in recent decades. For many decades following the classical period, economists had simply abandoned the field of population studies (Blaug, 1962, p. 69). The decline of fertility in the last half of the nineteenth century, for example, was explained as a change in the 'taste for procreation.' To some extent, Malthus's own limited reliance on the role of incentives may have influenced this outcome. Moreover, his proposition that population growth would eventually converge to an exogenously given growth rate of food supply may also have contributed to the failure to treat population as endogenous. Other aspects of the classical theory find expression even in the contemporary economic literature on endogenous growth: the possibility of multiple equilibria,

I Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

including a poverty trap, the distinction between static and dynamic equilibria, and the relevance of both fertility and longevity in economic growth. Adherents of the Malthusian theory have interpreted it to mean the dependence of economic growth on population control measures imposed by the government - an idea which still resonates among contemporary advocacy groups. Malthus himself, however, was consistently opposed to government interference in decision making by individuals, including any private birth control measures, on the grounds that these would stifle the enterprising spirit of individuals, thus resulting in adverse effects on productivity.

3. Some basic facts concerning demographic trends and growth Despite its intellectual appeal, the classical theory of population and growth has had little success in explaining related empirical evidence. In this section, we shall review some of this evidence - in particular, the covariance of the level or growth rate (g) of per capita income (q = Q/N) and the population growth rate (n) or its component parts: the birth rate (b = B/N) and the death rate (d = D/N). By definition, n = b-d. According to the study by Hagen (1959), the average growth of world population from the beginning of the Christian era to 1650 was in the neighborhood of percent per year. It then began to rise, first in Western Europe. In England, population doubled in the two centuries prior to the Industrial Revolution, yielding an average rate of increase of 0.35 percent (Coale, 1986). During the second half of the nineteenth century, the increase occurred also in the peasant societies of the colonies. A recent study by Kremer (1993), using very long time series, argues that historically, when the level of per capita income, q, was low, the population growth rate, n, was proportional to its level, N, until world population reached about 3 billion. Kremer also argues that a high N may be conducive to income growth (see Section 5.4 below). This is hardly indicative of any direct causal relationship between N and Q/N or between n and g, however, partly because of the need to distinguish movements in n that are caused by those in either b or d. For example, historical evidence indicates that the level of per capita income in peasant societies had not risen before the rise in the population growth rate. In these societies, death rates had been as high as birth rates. Where population growth occurred, it occurred because death rates fell (see Hagen, 1959). In the modern developed economies, where per capita income is relatively high, n, appears to be negatively related to Q/N essentially because a higher Q/N is associated with a larger decline in b than in d (see the studies surveyed in Lee, 1987). Further insight concerning the association between Q/N and n, or its components, can be derived from the historical trends during the pre- and post-

/ Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

Industrial Revolution era. Population was increasing at a rate unprecedented in European history during the last decades of the 18th century. This was due to both a significant decline in mortality in the 1780s and 1790s and an increase in the birth rate, with much of the increase in life expectancy probably due to a decline in infant mortality rather than an increase in longevity (Blaug, 1962, p. 69). Perhaps these trends gave rise to the Malthusian preoccupation with the prospect of population explosion. But birth rates, which had been rising in the 18th century, actually began to fall around 1800. The first transitional reductions in the national birth rates seemed to have occurred in France and the United States. The latter had experienced rapid increase in the birth rate prior to 1800, and thus the start of the decline in fertility was not much noticed. The decline in France was well recognized, however. By 1900, the total fertility rate (TFR), which is defined as the average number of children who would be born per woman during her reproductive period (age 15 to 50), based on the age-specific birth rates in the population, was only 2.8, a level attained in other European countries only around the 1920s or later. From British census data we know that the population growth rate in Britain was 1.3 percent for 1801-11, 1.7 for 1811-21, 1.5 for 1821-31, 1.2 for 1831^1, and 1.2 for 1841-51 (Rostow, 1990). Although the size of the population continued to increase throughout the period, the annual rate of increase reached a maximum in the 1811-21 period. It should be emphasized, however, that these increases in population were accompanied by declines in both the birth and death rates. The general pattern of demographic trends in countries which have experienced significant declines in fertility as well as significant increases in per capita income can be characterized as follows: In the first stage mortality rates fall, fertility then rises for awhile, both contributing to a rise in population growth. Eventually, however, this trend is followed by a b steady and continuous decline in fertility. For example, in 1965 the overall TFR in the developed countries was 2.8, while in 1991 it dropped to 1.8. Life expectancy at birth over the same period climbed from 71 to 77 in the same countries (World Bank, 1988, 1993). This pattern is generally called the 'demographic transition', a term first introduced by Notestein (1945). The pattern is quite common, and is not confined only to European countries. Singapore, Hong Kong, Taiwan, and South Korea, which have experienced a more than 50 percent reduction in fertility in recent decades, have also gone through a similar process. Hong Kong's TFR, for example, fell more than 70 percent from 4.7 in 1965 to 1.4 in 1991 - a period of only 26 years. There are exceptions, however. In the case of

b'For systematic documentation and analysis of these empirical regularities see Dyson and Murphy (1985), Coale
(1987), and Easterlin (1987).

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 9

the French transition, the fall in the death rate throughout the 19th century closely paralleled the fall in the birth rate rather than preceding it. As a result, the population increase in France was much slower than in other European countries (Coale, 1986; Coale and Watkins, 1986). The stylized facts of the demographic transition are manifested not just in time-series data for specific countries, but also through cross-sectional comparisons. Rostow (1990), using data for the year 1965 from 76 countries, found that birth rates are negatively and significantly correlated with per capita GNP. He also found a similar negative correlation between death rates and per capita GNP. These observations suggest that as a country develops, both fertility and mortality rates tend to go down. That the decline in mortality can in the longer run cause a reduction in the fertility rate, contrary to the Malthusian theory, is supported not only by casual observations, but also by econometric studies. Schultz (1985), using Swedish pooled cross-county data from 1860 to 1910, has shown that a quarter of the decline in TFR in that period was due to the 50 percent reduction in child mortality. Another quarter of the decline of TFR was caused by the increase in the urban share of the population, while the third quarter could be attributed to the 10 percent increase in the female-to-male wage ratio. The latter finding points to the role of the opportunity cost of time of the parent mostly involved in child care in determining the association between population and growth. The importance of the parent's market productivity is also echoed in Lee and Gan (1989) who use modern Japanese time series data for their analysis. They show that higher wages for women may have lead to lower fertility because of the dominance of the substitution over the family-income effect generated by the female's wage rise. An increase in males' wages, in contrast, has a positive effect on fertility, because the family-income change generated thereby 2 is higher than that caused by a rise in females' wages. Ehrlich and Lui (1991) have investigated the empirical relationship between young-age longevity, old-age longevity, the overall growth rate of per capita income, g, and alternative measures of fertility (TFR and b), using a cross-section of international data for the period 1960-1985. They find that young-age longevity, measured by the probability of survival in the age group 0-25 (7ii), has a negative effect on both the birth rate and the TFR, while old-age longevity, measured as the conditional probability of survival in the age group 50-75 (n2), has virtually no effect on these two variables. They also find that a 1 percent increase in Tii has a significant and more sizeable effect on the growth rate of per capita income, g, than a corresponding increase in n2. In a related study, Meltzer

2The argument concerning the potentially different income and substitution effects associated with higher wages
of males and females within a family unit was originally made by Mincer (1962) in the context of labor-force participation decisions.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 10

(1992), using a more detailed age group classification in which probabilities of survival are weighted by the number of years within each age class, finds results generally similar to those in Ehrlich and Lui (1991). Both studies, however, treat longevity measures as pre-determined variables. That the decline in mortality can bring down fertility has also been shown in a study that treats both as endogenous variables. In Yamada (1985), the causal relationships between these two variables are examined using time series data and causality testing techniques. He concludes that infant mortality and fertility are jointly determined. A decline in infant mortality due to an increase in per capita real income brings about a subsequent decline in fertility. Other studies have dealt more directly with the causal relationship between population and per capita income. According to Coale and Hoover's pioneering study (1958), a high rate of population growth is not supported by a corresponding increase in investment that maintains per capita income intact. A high fertility rate increases the dependency burden and lowers private savings and investment rates. There is therefore a fall in the 'per capita consumer equivalent' income. Enke (1960, 1976), who advocates population control policies, also comes up with a similar result. But these conclusions have been conceptually and empirically challenged in a study by Ram and Schultz (1979). Using Indian data from the 1950s to the 1970s, they conclude that over that period, there were large increases in life expectancy of youth and adults as agespecific death rates declined sharply. The observed sequence was improvement in health, fall in mortality, increase in life span, modest decline in fertility, and increase in population. During that period, India also experienced an impressive increase in investment in schooling. Treating the latter as part of the economy's overall capital formation, investment and savings were found to have risen appreciably faster than national income. Improvements in health also raised productivity. The Indian data thus support the proposition that, partly because of the large increase in health and longevity, the growth in population does not result in a decline in 3 savings and capital formation.

4. The renewed interest in population and growth The first significant break with the influence of the classical theory of population and growth came about with the development of the so-called neo-classical

3The empirical regularities surveyed in this section concern mainly secular trends or cross-sectional variations in
fertility and growth. Over time there have also been deviations from the long-term trends which lasted over short periods. Examples of such deviations are the decline in fertility during the Great Depression and World War II and the subsequent Baby Boom in the 1940s and 1950s in the USA and other European countries.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 11

growth models of Solow (1956), Nelson (1956), Denison (1962), Koopmans (1965), Cass (1965), and others. By denying the significance of fixed land as an overwhelming constraint on production, and replacing it with an endowed level of production technology exhibiting constant returns to scale for labor and capital inputs, the neo-classical model has shown that even with population increasing geometrically over time, the level of per capita income need not adjust to an exogenously restricted (and lower) rate of production growth, let alone be trapped in a dismal level of subsistence. Under a competitive market system, individuals' incentive to save will assure a rate of capital formation which, in the long run, will match or surpass the rate of population growth, depending upon the rate of advance in production technology. The 'golden rule' of balanced growth would thus involve a constant ratio of capital to output over time, with technology as well as population and physical capital rising at constant geometrical rates. These models, however, have ignored a critical feature of the classical theory by treating population growth (ri) as an exogenous variable, or at least ignoring its micro-foundations. Also, the magnitude of n is assigned a role in determining only the level of per capita income, but not its steady-state growth rate, g. The latter is assumed to be influenced solely by the growth in technology, which itself is taken to be exogenously determined. The models are thus silent on the possible influence of economic conditions on demographic variables. They cannot address the observed diversity in n or its components, b and d, across different economies or in the same economy over time, especially during different stages of economic development. The seminal paper by Becker (1960) is the first serious attempt to fill the void in this literature by reformulating the classical theory of population in modern economic terms. This paper offers a theory of fertility in which children are regarded essentially as durable 'consumption goods' which obey the basic laws of demand, although they may also share some characteristics of a producer good. Formally, Becker introduces the quantity of children as an argument in parents' utility function, where they compete with all other commodities. He recognizes, however, that parents have a demand for both quantity and quality of children, the latter being defined by the amount of resources they spend on educating and nurturing each child. Decisions concerning the quantity of children are partly affected by parents' ability to exercise birth control, which is determined in turn by their own education level, subject to the uncertainty of conception and birth. The demand for fertility is thus specified in this paper as a function of family income, costs of children, and parental knowledge. Although the paper does not deal explicitly with the interaction between the quantity and quality of children, Becker expects both to be 'superior goods'. He also expects the quantity income elasticity of demand for children to be smaller than the quality income elasticity, as is the case for most durable goods. He

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 12

argues, however, that these behavioral propositions must be tested only after accounting for the effectiveness of parents at exercising birth control. As evidence he cites his own empirical analysis, based on some US and Swedish data, showing that when contraceptive knowledge is held constant as one of the explanatory variables in a regression analysis, a positive relationship emerges between income and the rate of completed fertility. The literature following this phase of renewed interest in population and growth can be divided into two complementary streams. The first has pursued a more complete treatment of population as an endogenous variable, but has generally taken growth itself to be an exogenous variable in the tradition of the neo-classical growth model. This literature is reviewed in Section 5 by basic analytical themes which have also influenced the second, and more recent stream, which treats both population and growth as endogenous variables. We review the latter in Section 6.

5. Endogenizing population A complete treatment of population as endogenous to the economy requires that both fertility and longevity be treated as endogenous components. The literature dealing with the latter topic has so far developed mainly as an application and extension of human capital theory and health economics under static conditions (see, for example, Grossman, 1972; Ehrlich and Chuma, 1990). A related comment applies to the theory of marriage (see Becker, 1973). We shall therefore limit the scope of the following survey to studies of fertility behavior and its relation to major aspects of population dynamics. 5.1. The motive for having children A key question in establishing the micro-foundations of a theory of population is the parents' objective function. As both the original Malthusian specification and Becker's (1960) reformulation indicate, two basic motives are generally recognized: having children as a means of providing some material benefits to the family unit or as pure 'consumption goods'. Material benefits are of two kinds: children may contribute labor services to a family business, as is generally the case in farming, or provide old-age security to parents. Psychic benefits can also be further distinguished by the way children enter a parent's utility function: either as durable goods, which confer benefits on parents through the latter's interaction with overlapping generations of children, or as an infinite 'dynastic' set of utility functions of all future generations of offspring descending from the 'dynasty head'. While both specifications can be viewed as an expression of altruism on the part of parents, the latter represents a 'pure' altruistic motive, because it implies that the pleasure parents

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 13

derive vicariously from children exists independently of whether parents know, let alone interact with their future progeny. Whether one of these motives dominates the other has potentially important implications concerning the relationship between population and growth and even the relevance of population control policies (see Sections 6 and 7.1). We shall therefore review studies that have attempted to determine the relative strength of these motives analytically or through some indirect empirical tests. One line of attack has been to test the strength of the old-age security motive by examining the effect of social security on fertility. Leibenstein (1963) was one of the first to argue that social security programs may reduce fertility. He suggested that if old-age support is indeed an important motive for having children, then one may expect that a pay-as-you-go social security system, which is a state-provided alternative to the family as a source of support, will generally lower the fertility rate. Additional assumptions are needed, however, for the argument to go through. A common justification is the absence of a capital market that allows people to save efficiently (Nerlove, Razin, and Sadka, 1987). As we shall see below, this is hardly the only way to justify the proposition. There are numerous empirical studies that appear to partially support this proposition, (see Nugent, 1985, for a review of this literature). For example, Entwisle and Winegarden (1984), studying a cross-section of 48 less developed countries, find that the total fertility rate in 1975 varies inversely with government expenditures on pension benefits as a proportion of GDP in 1970, after controlling for other explanatory variables, including the countries' GDP, life expectancy, and the population literacy ratio. Nugent and Gillaspy (1983) use a cross-section of 34 counties in the principal sugar-cane growing area of Mexico and find that changes in fertility between 1960 and 1970 are negatively related to a social security proxy variable. Ehrlich and Lui (1994) study a cross-section of 45 countries and find that, after controlling for GDP and measures of both young-age and old-age longevity, social security benefits as a percentage of GDP in 1960 have a negative and significant impact on the birth rate in 1965 and the total fertility rate in 1975 in poor countries, but no such significant effects in relatively rich countries. Cain (1981, 1983) develops a 'lexicographic safety first' theory of parents' fertility behavior. By this theory, the paramount concern of parents is their old age support. Parents' perception of the minimal number of children necessary for such support depends on (1) the likelihood that a child survive to adult age and provide support, (2) the potential level of support they can expect from their children, and (3) the availability of alternative means of support. Parents are unresponsive to any other motive for having children until that minimal number of children is attained. A switching regression model by Jensen (1990), using Malaysian data, is employed to test Cain's theory. The evidence appears to support the hypothesis that old-age security is the principal concern of those parents who have not yet attained a sufficient number of children.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 14

The above tests, however, do not settle the issue. Becker and Barro (1988) [BB], basing their analysis on a purely altruistic motive for having children, argue that a permanent increase in social security taxes can be expected to either temporarily or permanently discourage fertility. Given that the increase in social security taxes and benefits raises the net tax on the current generation, the effective cost of children is increased. In the context of BB's dynastic optimization analysis, this implies a necessary reduction in fertility in the generation experiencing the increase in social security taxes. Whether such a reduction is temporary or permanent depends on whether the wage and interest rates ( w and r) are endogenously determined. In an open economy (Barro and Becker, 1989), the latter are assumed to be constant, and the consumption path for later generations, and hence the corresponding relative price of fertility in their model, remain unaffected. In future generations, therefore, fertility 4 adjusts to its previously optimal level. In a closed economy (Barro and Becker, 1989), where both w and r are affected by changes in the capital-to-labor ratio, the steady-state level of r falls, causing a permanent decrease in desired fertility. Lapan and Enders (1990) argue on the basis of a somewhat more general specification of a dynastic utility maximization that outstanding government debt in general raises the full cost of child bearing and causes a permanent reduction in the steady-state fertility rate. A similar result has also been obtained in a related earlier paper by Batina (1987). He argues that an increase in government debt due to a reduction in consumption tax will reduce the steadystate number of children. Like BB (1989), both of these papers treat the interest rate and capital accumulation as endogenous variables. Cigno and Rosati [C&R] (1992) attempt a comprehensive test of the relative strength of self-interest vs. altruism through the impact of social security. In addition to fertility, they also focus on savings as a jointly determined variable. Using a three-period overlapping generations framework, C&R consider a model of old-age support whereby young working parents 'lend' resources to their children at a given interest rate and then collect on these loans for old-age support (cf. Ehrlich and Lui, 1991). Under a model of altruism, children's utility is specified as part of the parents' utility. There is no human capital investments or growth in these models. Using Italian time series data, C&R derive three sets of results bearing on the relative strength of the selfish old-age security motive and intergenerational altruism. First, they consider the effects of a fully funded increase in social

4A temporary increase in the effective cost of children would by the same analysis lead to a reduction in fertility
in the current generation, but a significant increase in fertility which may offset the previous reduction in the following generation. BB use this argument to explain the Baby Boom experienced in the US and other countries following the temporary reduction in fertility during the great depression and World War II. An alternative explanation is given by Butz and Ward (1979).

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 15

security coverage. If self-interest is the motivating force for parents, they expect a negative impact on fertility, but an ambiguous effect on savings. In contrast, if altruism is important, they expect social security to have no effect on fertility. But since the fully funded increase in social security coverage is a form of forced saving for old age, private saving (for, say, middle age) will go down. Since their empirical results are that fertility falls while savings goes up, they argue that the old-age security motive is paramount. Second, C&R examine the implications of social security deficits. Their model implies that larger deficits will cause parents to demand fewer children if they are motivated by altruism but not by self-interest. In the nonaltruistic case they also predict a decline in savings because an increase in social security deficits means a rise in benefits for the current generation. Under altruism, the effect on savings is ambiguous. On the one hand, parents wish to leave larger bequests to alleviate the future tax liability on their children. On the other hand, since the lifetime consumption of parents also rises because of the increase in social security benefits, this may have a negative effect on saving. In their empirical analysis they find the deficits to have no effect on fertility but a negative effect on saving, again supporting their self-interest hypothesis. Third, C&R find that greater access to capital markets exerts a negative effect on both saving and fertility. Their self-interest model does not predict an unambiguous effect on savings. But they expect that under this assumption children will become less attractive as a form of investment for old-age support. Under altruism, however, they expect that both savings and fertility should increase. Willis (1989) develops another theoretical approach that may be used to test the old-age security hypothesis. He expects such motive for childbearing to arise when there is uncertainty concerning one's longevity and annuity markets are imperfect. In this case, parents depend partly on their own savings and partly on their children for old-age support. Survival to older age would make old parents deplete a greater amount of own savings and leave smaller bequests for their children. Hence, under imperfect annuity markets, the old-age security hypothesis would imply a negative correlation between parents' survival to old age and the welfare of their mature children. Willis further hypothesizes that imperfect annuity markets contribute to greater fertility. In a perfect market, survival risks across individuals of the same generation can be pooled via life insurance or annuity contracts. But if the latter do not exist, intergenerational transfers within the family offer a partial substitute. Reliance on intergenerational risk pooling will cause an increase in fertility relative to the situation under a perfect market. An important factor that has been ignored in some of these studies is that the quantity of children is not the only variable under parents' control. Once we introduce the quality of children and the availability of savings opportunities as complementary choice variables, it is no longer clear that social security will

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 16

necessarily decrease the fertility rate even if old-age security is a major concern. At the same time, even a permanent reduction in fertility following an increase in social security taxes may be compatible with altruism in this case (see Ehrlich and Lui, 1994, and Section 6 below). While some of the empirical tests which appear supportive of the dominance of the old-age security as a motive for having children may also be consistent with an altruistic motive, the same criticism applies to empirical tests which appear unsupportive of this hypothesis. For example, Mueller (1976) concludes that the rate of return to children is negative in most peasant societies. This suggests that old-age security is not the only motive for raising children. But this conclusion may be premature. Willis (1980) argues that the combination of a negative rate of return and a positive rate of population growth may actually indicate a strong selfish motive for child bearing because parents must have a large number of children in order to secure a subsistence consumption level during old age. Population growth in such circumstances would be inefficiently high since low levels of transfers to the elderly per child will be associated with high levels of fertility. This is consistent with studies in less developed countries where the rate of return to investments in children is negative when the fertility level is high. Attitudinal surveys in these countries indicate that parents expect support from their children at old age, and this expectation is one of the important reasons for the existence of large families. Contrary to models which adopt the old-age security motive, papers treating children as objects of intergenerational altruism, especially Barro (1974), have generally reached different conclusions concerning the effect of social security on individual savings. A basic implication of Barro's Ricardian Equivalence theorem, given altruism, is that a pay-as-you-go social security scheme will have no effect on the savings rate. While one may be tempted to use this hypothesis to test the importance of the altruistic motive, this would not be a valid test. As Wildasin (1990) has shown, the neutrality result concerning social security's effect on savings no longer holds when fertility is endogenous. Moreover, altruism is not even a necessary condition for obtaining the neutrality result. It is possible to show that in a model that recognizes old-age security as a significant motive for having children, social security may have no effect on either the savings or fertility rates in a steady-state growth equilibrium 5 (Ehrlich and Lui, 1994). As indicated at the beginning of this subsection, the relevance of the material vs. psychological motive for having children can also be tested by reference to the economic value of children in a family business. In general, measures of child

5Another indirect way to assess the relative importance of the old-age support motive vs. pure altruism is by
reference to the success of models incorporating the alternative motives to explain the stylized facts of the demographic transition referred to in Section 3. See our discussion in Section 6.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 17

labor-force participation appear to be positively correlated with birth rates (for example, see DaVanzo, 1971). In models that treat both child labor-force participation and fertility as endogenous variables which are responsive to the pecuniary returns from child labor, variables such as the size of land-holdings, agricultural productivity, child wage rates, and cropping patterns that are better suited for child labor appear to be positively related to both fertility and child labor-force participation and negatively related to child schooling (see Rosenzweig and Evenson, 1977; Makhija, 1977; Levy, 1984). 5.2. Quantity versus quality of children Regardless of whether altruism or an expectation of material benefits is the major motivating force for bearing children, the quality of children as measured by their human capital is a distinct choice variable for parents. Quantity and quality of children are jointly determined - an idea well recognized by Marshall (1930, Book IV, Chs. 1-3). Addressing the former without the latter would generate incomplete, and even misleading propositions about the correlation between economic growth and the components of population growth. The first systematic analysis of the interaction between the quantity and quality of children is offered in Becker and Lewis (1973). They see parents as maximizing a utility function where both quantity and quality enter as distinct arguments, subject to a specific budget constraint. The formal problem they analyze is max U = U(x, q, y ) subject to I = xqP + yPy, where x is the number of children, q is their quality, y is the rate of consumption of all other commodities, / is full income, P is the price of a unit of child quality, xq, and Py is the price of y. It is easy to see that the shadow price of the number of children, x, is greater the higher their quality, q. Similarly, the shadow price of q is higher the greater x is. Following Becker (I960), Becker and Lewis further assume that the income elasticity of demand for quality is substantially higher than that for quantity. Consequently, when income increases, the shadow price of quantity must go up more than the shadow price of quality. This model thus provides a simple explanation for the observation that the fertility rate may be negatively correlated with the level of income even though both the 6 quantity and quality of children are superior goods. The negative correlation between quantity of children and income level can be explained even if there were no child quality component to parents' fertility

6Hanushek (1992) reports empirical evidence showing a negative effect of family size on children's scholastic
performance in support of the existence of a distinct tradeoff between quantity and quality of children in intrafamily allocation decisions.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 18

behavior. Instead of stressing the role played by the desired human capital of children, one may focus on the role of the human capital stock of working adults. The Eckstein and Wolpin (1985) paper takes this approach. This paper assumes that the cost of children is a function of the parent's wage rate - itself an endogenous variable - affecting the time cost of rearing children. Under the assumption that per capita income (and the wage rate) rises continually because of human capital accumulation, their model can generate a pattern of first rising and then falling fertility rate, thus explaining a basic feature of the demographic transition. Growth itself, however, is not an endogenous variable in this model. Becker and Lewis (1973) is a static model, and as such, it cannot be used to analyze dynamic aspects of the interaction between quality and quantity. Many of the more recent papers viewing growth in per capita income as an endogenous variable identify human capital as the engine of growth. Thus, when parents choose to emphasize quality over quantity of children by investing in the latter's human capital, they raise both the level and long-run rate of growth of per capita income while lowering the fertility rate. A further discussion of the tradeoff between quantity and quality will therefore be pursued in Section 6 where these models are surveyed. 5.3. Intergenerational transfers Social security is just one example of intergenerational transfers. Such transfers have long been recognized as a pivotal determinant of the fertility rate. Caldwell (1976, 1982), in particular, argues that in any society and any stage of economic development, whether fertility is high or low is determined primarily by the direction of the intergenerational flow of wealth. In all traditional societies, the flow has been from younger to older generations. However, a number of economic or social changes, such as greater mass communication or education, which reduce people's willingness to place family interests above their own, or a shift from familial to nonfamilial modes of production, may reverse the direction of the flow. When such changes take place, the family is largely nucleated, and the 'great divide' is said to occur: Instead of children contributing to their parents, parents begin leaving larger bequests or invest more in the human capital of children. Since the costs of children are raised in the process, fertility goes down. In Willis (1986), fertility and investment decisions made by altruistic parents will be socially optimal under a perfect market. The existence of voluntary private transfers from children to parents need not have anything to do with the old-age security motive. These transfers may occur because under an imperfect capital market, the parent's wealthmaximizing investment in the child's education, in addition to rearing costs, may exceed what is optimal for parents to invest. Without a truly complete market, the child cannot borrow from a bank. Parents are substituting for banks as a source of finance. Old-age transfers are

19

/ Ehrlich, F. Lux j Journal of Economic Dynamics and Control 21 (997) 205-242

therefore simply 'child debt transfers'. In this case, fertility is not affected. When intergenerational transfers shift from the family to the public sector, however, this is no longer true. Introduction of a pay-as-you-go social security system causes a divergence between the social and private costs of children and makes the fertility rate suboptimally low. Moreover, Willis believes that public transfers will cause equal and opposite changes in private transfers within the family. This proposition appears to follow the Ricardian Equivalence argument. Given that fertility is endogenous, however, it is unclear why private transfers would completely offset the effects of public ones. The argument that improved capital markets are a hindrance to fertility has also been used by some authors to reach an opposite conclusion. For example, Nerlove, Razin, and Sadka (1986,1987), postulating that parents can both treat their children as a means of providing oldage security and be altruistic toward them, argue that the introduction of a capital market may possibly increase the number of children. The existence of a capital market facilitates the transfer of resources from current to future consumption. This may reduce the demand for children if the latter are desired only for old-age security. There are two reasons, however, why the effect could also be in an opposite direction. First, better access to the capital market improves welfare and thus may generate a positive income effect on the desired number of children. Second, the market rate of interest may be higher than the rate of return on investing in children for some families, but it may also be lower for other families. The latter would have an incentive to borrow in order to invest in children. Even though a capital market provides better savings opportunities, which are substitutes for intergenerational transfers, the overall impact of the emergence of capital markets may still be to raise the fertility rate. Intergenerational transfers may exist entirely independently of imperfections in capital markets, simply as a result of the productivity of roundabout methods of maximizing the joint incomes of parents and their offspring. This theme is discussed in Section 6.2 below. 5.4. Multiple equilibria In the recent literature on endogenous growth (see Section 6), some papers resort to the use of multiple equilibria models to explain the diversity of growth among countries. In such models there is typically more than one stable steady-state equilibrium, and possibly also some unstable ones. Which equilibrium the economy will converge to depends more on historical or initial conditions rather than on the underlying parameters of the model. Once an economy settles on one of these equilibria, it needs a 'big push' in initial conditions, or major parametric changes, to move the economy from one steady state to another. The distinct contribution of these models is the implication that the same set of basic parameters, such as the utility function, the rate of time

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 20

preference, or the technology of production, could be consistent with very different states or stages of economic and demographic development. Under specific conditions, they may therefore be used to justify activist government policies. Multiple equilibria are not a new feature of growth models. Solow refers to the possibility of multiple equilibria in his original neo-classical growth model (1956, pp. 71, 90). Nelson (1956) provides a good illustration. His model is an extension of the basic neo-classical model with one twist. He assumes that, at a very low level of income, the fertility rate is also very low. But as income goes up, he assumes that fertility would start increasing rather rapidly. Finally, at an even higher level of output, fertility would slow down once again. Given these assumptions, it is easy to show that multiple equilibria will exist in this economy. In particular, if initial income is above the threshold level required to move to the third stage of development alluded to above, then fertility would become low and per capita income would keep increasing. But if the initial income level is low enough, the economy will move towards the so called low-level equilibrium trap where fertility is high and people remain poor. Similar 7 analyses are pursued in Buttrick (1958, 1960), Tsiang (1964, 1988), and Leibenstein (1963). In a recent paper by Galor and Weil (1993), multiple equilibria can also be generated through another mechanism. Higher wages of women reduce fertility by raising the opportunity cost of children more than household income. The resulting lower fertility, in turn, raises the level of per capita physical capital. They also assume that physical capital is more complementary to women's labor input than to men's, so women's wage rate is then increased relative to that of men. Their inference is that countries with an initially high level of capital will converge to a high income-level equilibrium with low fertility and high relative wages for women. Just the opposite would be true under these assumptions for countries with a low level of initial capital. Whether multiple equilibria models are accurate in describing reality is a controversial issue. Simon (1992, Ch. 7), using calibrated simulations of his model of population, argues that a one-time increase in income does not stimulate a nearly sufficient increase in fertility that would 'eat up' the additional income to leave per capita income unchanged even in less developed countries. Kremer (1993) also argues that many development economists and historians see development as a continuous process, rather than the result of discontinuous shocks, such as the Industrial Revolution. Kremer's model, which attempts to support the pattern of observed historical data on population, also produces theoretical results that are compatible with continuity. His crucial assumption is

7Nelson's work was developed independently of Solow's, and the idea of multiple equilibria was also pursued
independently by Buttrick and Tsiang. For a discussion of the evolution of this idea, see Tsiang (1964).

/. Ehrlich, F. Lui / Journal of Economie Dynamics and Control 21 (1997) 205-242

21

that each person's chance of invention or scientific discovery is independent of that of others in the population. Thus, when the population size is larger, the likelihood of a technological improvement which affects the entire economy is higher, so the growth rate of per capita income will then go up. The increase in per capita income, in turn, can support a further increase in population. His model implies that the growth rate of population is proportional to the size of the population when the latter is not too large because of the interaction between population and technology. When the income level is sufficiently high, however, he expects diminishing returns to population to set in because of the scarcity of natural resources, and therefore a lower rate of population growth.

6. Endogenous economic and population growth Our survey began with the classical model which pioneered a comprehensive treatment of population and income as endogenous variables. The neo-classical models, while offering a rigorous reformulation of the dynamics of the growth process, have treated both population and per capita income growth as essentially exogenous variables. The later literature covered in the preceding section has concentrated on the micro-foundations of a more complete population theory but has still treated income level or growth as exogenous variables. One of the main objectives of the more recent literature on 'endogenous growth' has been to go beyond these constraining assumptions by treating both population and income growth as endogenous variables (see Ehrlich, 1990). A major challenge facing such complex dynamic models has been to explain how developing economies have typically moved from a state of low and fairly stagnant per capita income, high mortality, and high fertility to a regime of persistent growth in which first mortality, and then fertility are continuously declining while per capita income exhibits persistent growth - a phenomenon known as the 'demographic transition' (see Section 3). The challenge has caused a shift of emphasis in population studies from concern for 'population explosion' in the less developed countries, to concern for 'population implosion' and the aging of the population in the more developed ones. A few attempts have been made to explain this phenomenon. For example, Notestein (1945) argues that fertility was high in pre-modern society because of religious beliefs, moral codes, and related factors. High fertility was also necessary to offset the high mortality rate. But the evolution of cities with large and mobile populations altered the situation by shifting the family-based life style to one centered on individualistic aspirations which bring fertility down. Sah (1991) shows that even in a discrete choice model where parents' expected utility depends on the (integer) number of children and their survival probability, a reduced child mortality rate will lower the fertility rate (also see Ben-Porath and Welch, 1972).

/. Ehrlich, F. Lui / Journal of Economie Dynamics and Control 21 (1997) 205-242

Becker (1992) also discusses the theoretical relationship between child mortality and fertility. In a high-mortality environment, the portfolio of assets that elderly parents may rely on to secure old-age consumption needs is much less diversified if parents concentrate their investment of human capital in only one or two children, as opposed to investing little human capital in many children. When child mortality falls to much lower levels, the portfolio diversification advantage of a large family is greatly reduced. These explanations are quite insightful, but incomplete. The phenomenon of demographic transition incorporates trends in mortality, fertility, intra-family transfers, economic growth, and their interactions during different phases of economic development. These trends have been observed with remarkable regularity in virtually all countries that have experienced significant economic development. The real challenge for contemporary population theory has been to offer a unified approach capable of explaining all the relevant interactions. In the following section we survey two papers and an extension of one of these that have attempted to deal with the issue. 6.1. A dynastic family model Becker, Murphy, and Tamura [BMT] (1990) have been the first to offer a comprehensive model of endogenous population and economic growth, based on explicit micro-foundations, that also provides an explanation for some of the features of the demographic transition. The model assumes that the dynasty head is maximizing a dynastic utility function exhibiting a purely altruistic motive for having children (see our discussion in Section 4.1). The objective function can be represented formally as

/. Ehrlich, F. Lui / Journal of Economie Dynamics and Control 21 (1997) 205-242

23
(1)

V, = [ ( . :ct)-l]/(l-<x) + anrwF(+1,

24

/ Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

where V, and V t + l denote the utilities of parents and each child, respectively, c, is parental consumption, n, is number of children, 1/a is the pure rate of discount for altruism time preference (when n 1), is the constant elasticity of parental altruism per child as their number increases, and a is the inverse value of the elasticity of substitution in consumption. The latter parameters must satisfy the following restrictions: 0 ^ e < 1 and 0 < a < 1 (see Fn. 8). This function is maximized subject to the available technology of producing human capital and the consumption good, c. A slightly simplified version of these production functions is given by Ht+1=A(H0 + Ht)h, c, = ( H 0 + H t ) ( T- n t { v + h,}), (2) (3)

where H , represents the parent's stock of human capital at t and H 0 the parent's natural (unskilled) production capacity, or raw labor, which we here treat as a perfect substitute for human capital; h, denotes the time spent by the parent on teaching each child at t, A is a technology parameter governing human capital accumulation, T is total productive time available to the parent per period, and v is the time cost of rearing each child. As indicated by Eq. (2), human capital is the engine of growth in this model (also see Lucas, 1988). Since time spent on investing in children's human capital, h, is a choice variable for parents, growth in per capita income and consumption becomes an endogenous variable. Similarly, since the number of children, n, is a choice variable, fertility is also endogenized. Given the model's assumptions, the rates of return to investment in both the quantity of children and their human capital are found to generally depend on both the level of human capital of the dynasty head and the chosen number of children. The optimal solution to this optimization problem is given in the form of a second-order simultaneous difference equation system. The rate of return to investment in human capital is shown to be a decreasing function of the optimal number of children, which is shown to be a decreasing function of the head's endowed level of human capital as long as the latter is low relative to the endowment of raw labor. Consequently, three steady-state equilibria can be shown to exist: a stable low-level Malthusian trap, an intermediate but unstable state of development, and a stable state of persistent and self-sustaining growth. The model shows that if the initial level of human capital is higher than a threshold level associated with the unstable development equilibrium, then investment in human capital becomes more worthwhile and the economy will keep on growing. If initial human capital is lower than that threshold level, the economy will converge towards the Malthusian trap with a high fertility rate, zero investment in children's human capital, and zero growth. A high fertility rate is thus shown to be associated with a low level of human capital, hence income, and a zero growth rate. Specifically, in a steady state of perpetual growth, the optimal fertility level is shown to be lower than under both the development and Malthusian equilibria. The above results provide an explanation for some, but not all aspects of the demographic transition. The role of mortality, which is an important part of the demographic transition, is not explicitly modeled. Also, while the model produces a clear-cut comparison of the fertility levels in the two stable steady states discussed above, the behavior of fertility in the transition from a Malthusian to a growth equilibrium is less clear. The authors illustrate the model's transitional dynamics by a phase diagram in which H t + 1 i s plotted against Ht. For this to be a general characterization of the model's transitional dynamics, the law of motion of human capital would have to be representable by a first-order difference equation. However, the model's first-order optimality conditions form a system of simultaneous equations of a higher order. The model cannot

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

26

therefore produce an explicit analytical solution for the transitional paths of either human capital formation or fertility. Another important implication of the multiple equilibria result deserves attention. The result means that under the same objective function and production technology, the economy may either sink into a Malthusian trap or be elevated to a regime of self-sustaining growth, depending solely on the initial human capital of the dynasty head. There is no problem of time inconsistency associated with this result. Under the model's assumptions, the dynasty head in any generation consistently maximizes the dynastic utility function given by Eq.(l). One may question, however, whether the equilibria derived through the dynastic utility maximization are also Pareto optimal. To see this, suppose that the initial level of human capital is just below the threshold level. Then the dynasty head is expected to choose a time path of investment that will lead the descending families to a Malthusian trap with no investment in human capital and a minimal level of income per capita. Suppose, however, that the head invests just a trifle more than what the 'private' optimization solution dictates. Then the descendants would be destined to experience perpetual economic growth. It is arguable that the existence of such a critical threshold is a reflection of some implicit market failure. There are indeed potential externalities within the dynastic system. Suppose that the altruistic dynasty head, who currently has only one son, figures that the optimal solution requires having another child. The son can rationally expect that if a sibling is born, the family will invest less in his own human capital, thus lowering his future income. If the son does not have the same altruistic sentiments toward his unborn sibling as do his parents, the latter's decision to have another child would impose a negative externality on the son's welfare. The optimal investment decisions of the dynasty head may not be socially optimal in this case. 6.2. A family model incorporating optimal intergenerational transfers Ehrlich and Lui [EL] (1991,1994) have attempted to resolve this problem by allowing for implicit contracts between parents and children. Young parents are generally motivated by a desire to secure old-age material support, as well as psychic rewards from interacting with their offspring, the latter identified as 'companionship'. In the context of this three-period overlapping generations model, young parents decide on the number of children, the amount of investment in the human capital of each child, and the amount of material compensa tion they expect from their children at old age that would be optimal from the point of view of both parents and children. There is, of course, a problem regarding the enforceability of such implicit contracts. Under some specific conditions, the implicit commitments by children

27

/ Ehriich, F. Lux / Journal of Economic Dynamics and Control 21 (1997) 205-242

can be shown to be fully credible and enforceable, especially when parents make special investments to ensure that their children live up to these commitments. In the benchmark version of the EL framework, based on an extended family compact in which the overlapping generations are motivated strictly by material incentives, the existence of such intergenerational trade is shown to eliminate both any intergenerational externalities and the multiple equilibria outcome of the BMT model. Given any set of parameters, there can be only one equilibrium at any particular point of time, although the latter may involve either stagnation or growth, depending on the productivity of the roundabout methods of investment in children. The rate of return to parents is determined by the life expectancy of children and parents, their time preference, the technology of producing human capital, and the optimal rate at which children compensate their parents for the latter's investment in them. Whatever the attained equilibrium, the level or growth rate of per capita income across generations is always maximized in this model, given that intergenerational transfers are optimal. This result is a manifestation of the Coase Theorem in the context of endogenous growth. This benchmark case of the EL model is not sufficient to explain the demographic transition since the optimal strategy of investment in children in this case implies the dominance of quality over quantity, i.e., it dictates a corner solution in fertility at a minimal number of children per family. This anomaly is remedied by recognizing 'companionship' (or altruism of parents toward their immediate offspring) as a complementary motive for having children. The generalized model also allows for private savings as a complementary means of assuring old-age needs. Formally, the EL (1991) model can be summarized by the following utility-maximizing problem (for simplicity, savings opportunities are suppressed in the following specification): max I c ^ t ) 1 ' " - 1]/(1 - a )
+ ^i{[c2(i

+ l)1"-l] +

1 -a

1]}/(1 - o ) ,

(4)

[c3(i + l
subject to the same law of motion of human capital represented by Eq. (2) above, and the following constraints on young-age consumption (cj), old-age consumption ( c 2 ) , and companionship (c3):
Cl

(i) = ( H t + H 0 ) ( l - n (v + h,)) - Tt w ,H,


t 2

(5) (6) (7)

c2(i + 1) = re^^ + jH,


c 3 { t + 1) = (*,/#?.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 28

In these equations, S is the discount rate, 7t, is the probability of survival from childhood to young adulthood, n2 is the probability of survival from young to old adulthood, and w,H, is the amount of support that a surviving young adult provides to the retired parent when the latter is alive. The parameters must satisfy the following conditions: 0 < a < 1, 0 < a < < 1, /} < 1/(1 a ) and a < jS. The 7t's in this model captue the effects of longevity and w is determined optimally by the implicit contract between the parent and each child. The magnitude of a is generally assumed to be less than one, implying diminishing returns to the 8 psychic benefits from companionship. An explicit solution can be obtained in this model for the transition path of an economy moving from a stagnant to a growth equilibrium. Suppose the economy is initially in a stagnant equilibrium with zero growth in per capita income. An exogenous increase in either young-age or old-age longevity (especially a reduction in child mortality as was the case in many European countries in the late 18th century), or advances in educational technology which increase the productivity of investment in children, will raise the rates of return to parental investment in both quantity and quality of children. Both the level of income and the fertility rate may go up as a result. And if the one-time longevity improvement is sufficiently large, it will trigger a take-off toward a regime of perpetual growth. Moreover, once human capital accumulation reaches a sufficient level, the rate of return on investment in human capital is likely to exceed that on the quantity of children. This is because a higher level of human capital increases the marginal cost of rearing children relative to that of educating them, and it also lowers the benefits from companionship relative to material support if the former is subject to diminishing returns from human capital (i.e., if a is less than one). The rate of fertility is then shown to fall continuously after an initial increase, and converge on a minimal level of fertility in a growth equilibrium. This theoretical path is consistent with all the basic features of the demographic transition. In Ehrlich and Lui (1994), the model is further generalized to the case in which implicit contracts are subject to uncertainty concerning both the survival and good behavior of adult children. Such uncertainty would generally bias the optimal investment decisions of riskaverse parents towards a larger number of

8These restrictions are needed to satisfy the second-order conditions for optimal consumption flows (as in BMT,
1990), as well as for the optimal rate of compensating old parents, w. Note that a in EL (1991) or BMT (1990), restricted to range between 0 and 1, is not directly comparable to its value in discrete-time (annualized) equivalents of Lucas (1988) and conventional macro models, where it may assume values higher than 1, because in the generational settings a denotes the inverse value of the elasticity of substitution in consumption across consecutive generations, rather than years. Thus, if a generation spans 25 calendar years, a value of a less than 1 in El or BMT could be equivalent to a value substantially above 1 in these other models.

29

/ Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

children and a smaller investment in each, as anticipated in Becker (1992). Even in this model, however, a high enough one-time increase in the probability of children's survival or compliance with the implicit family contract triggers the same path of the demographic transition summarized above. The main limitation of the EL model, however, is that it treats longevity itself as an exogenous variable. A fuller analysis would consider longevity itself, at least in part, to be an endogenous variable which is responsive to human capital accumulation and growth (see Ehrlich and Chuma, 1990). 6.3. A comparison of the two approaches and related empirical evidence Despite some formal differences between the BMT and EL models, both share key features. In both approaches human capital is the engine of growth and fertility decisions are strongly influenced by the quantity-quality tradeoff. Furthermore, both imply that the relationship between population and growth is not the result of a specific causal relationship between the two: population does not affect the level or growth rate of income anymore than the latter affect population. Movements in both reflect the dynamic impact of some exogenous or predetermined factors that jointly determine the phase or process of economic development. While the major application of this basic proposition concerns the covariation of population and growth over time, as reflected by the demographic transition, these dynamic effects are also relevant for understanding the association between income, education, fertility, and longevity across different economies or geographical units at a point in time. Although the BMT model is silent on the role of longevity, it is possible to incorporate the latter into their model without affecting its basic structure. Such an approach is pursued in Meltzer (1992). In this study, the objective function involves dynastic utility maximization and pure altruism, as in BMT, but the model recognizes three basic periods for each generation: a childhood-youth period of dependency, adulthood, and old age, as is the case in the EL model. Moreover, since EL's objective function incorporates both an altruistic and a material incentive into the objective function of the decision-making young parent, the models can be formally linked. One way to establish this link is to show that the BMT-Meltzer specification of the problem can be presented operationally as a special case of EL's. Since BMT-Meltzer do not recognize the possibility of intergenerational trade between overlapping generations, one may assign a zero value to the parameter w - the optimal rate of compensation of old parents by children - in EL's model. Once this is done, however, the rate of investment in children's human capital, hence the growth rate, becomes independent of longevity, which is precisely the result generated by the BMT-Meltzer model. Moreover, if the BMT-Meltzer infinite horizon is reduced to a two-period case, in a long-run growth equilibrium their altruism function becomes operationally equivalent to the compan

/. Ehrlich, F. Lui / Journal of Economie Dynamics and Control 21 (1997) 205-242 30

ionship function in EL (1991): when a is set equal to unity and /? is set to exceed unity, the EL model yields exactly the same steady-state growth rate as that of BMT-Meltzer. When w is positive, however, the EL specification always yields a higher level or growth rate of income relative to the (BMT) specification based on pure altruism. One difference in the behavioral implications of BMT-Meltzer relative to the EL specification (with w > 0) concerns the effect of longevity on fertility. In BMT-Meltzer, a higher longevity always increases the optimal number of children as long as the constant elasticity specification of the objective function is pursued. Thus, these specifications cannot reproduce a basic feature of the demographic transition. In EL's specification, in contrast, a higher longevity may trigger an increase in fertility only in the first stage of the demographic transition; the trend in fertility becomes decidedly negative in later parts of the demographic 9 transition as the economy joins a regime of self-sustaining growth. Another difference between the two approaches concerns the relative importance of youngage, as opposed to old-age longevity in enhancing the level or long-term rate of growth of per capita income over time. In the EL formulation, an increase in young-age longevity, as measured by 71!, increases the growth rate by a larger percentage than an equal percentage increase in old-age longevity, 7t2, because an increase in the latter also initially increases the expected cost to the younger generation of supporting the old. No such prediction follows unambiguously from a specification recognizing pure altruism as the only motive for having children. Some empirical evidence in Ehrlich and Lui (1991) appears to support the prediction that young-age longevity enhances the economy's growth rate by a larger percentage than oldage longevity. Their analysis indicates that aging of the population strictly as a result of reduced fertility (with no change in 7ti or n 2 ) may therefore lower the economy's long-term growth rate.

7. Related policy issues As the previous sections indicate, a central focus of the literature on the process of economic development as it relates to population has been the demographic transition. Since the transition involves an increase in longevity (at both young and old age) and a decrease in fertility, the unambiguous

9Meltzer(1992) argues that this shortcoming in his model can be remedied by further complicating the functional
form of the dynastic utility function. In particular, if the elasticities of the altruism function with respect to its arguments (e and a in Eq. (1)) are allowed to vary, growth might go up or down when longevity improves.

31

/ Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242

consequence is the aging of the population. This trend has been shared by all the developed and fast-developing countries in recent decades. Two major policy issues have been recognized in the literature dealing with these developments. The first concerns a rather old theme - the merits of population control policies as an inducement to growth. The second concerns the interaction between growth, aging of the population, and public policies concerning retirement income protection plans. In the following section we review each of these subjects separately. 7.1. Population control The pessimistic implications of the Malthusian theory of population about the prospects of growth and prosperity have led many of its followers, beginning with John Stuart Mill, to advocate government interference in the form of population control policies. Such policies 10 have also been advocated in recent years by some demographers and policy makers. Government interference has been justified on the normative argument that the exercise of free choice by parents does not necessarily lead to socially optimal outcomes, as well as on the positive argument that since per capita income is nothing but the ratio of output to population, q = Q / N , an effective way of permanently uplifting the latter above a miserable level of subsistence would be to lower the denominator. Both ideas have been challenged 11 repeatedly in the modern economic literature. Kuznetz (1960) may have been the first to suggest that population size, far from being detrimental to growth, may actually enhance it. The reason is that in a larger population there is a higher probability of useful innovations which promote productivity growth. Leibenstein (1969, 1971) cautions against possible adverse effects of population control. He believes that birth control policies influence mainly middle-class families whose members are relatively more productive than other families. Moreover, under a high population growth regime, older workers are replaced by younger ones who are more productive. As a result, the quality of the labor force improves. Simon (1992, Ch. 8) constructs a model designed to capture many of the characteristics of less developed countries (LDCs), such as economies of scale in

10See our references to Coale and Hoover (1958) and Enke (1960, 1976) in Section 3. An extreme example of
contemporary population control policies is China's policy of one child per family, in practice since 1980 (see Coale, 1981).

11

1 'Malthus's own basic objections are discussed in Seection 2. He was even opposed to the Old Poor Law in England, granting child allowances as a form of relief to poor workers, on the grounds that this would undermine the preventive check to population growth. Boyer (1989) finds evidence supporting this Malthusian argument.

32

/ Ehrlich, F. Lui j Journal of Economic Dynamics and Control 21 (1997) 205-242

production, intersectoral shifts in employment, and multiplier effects of investment. Based on alternative parameter estimates used to calibrate the model, his results indicate that moderate population growth produces considerably better economic performance in the long run than a slower population growth, although in the short run the latter performs slightly better than the former. Moreover, he argues that a declining population has the worst performance in the long run. The underlying reason is assumed economies of scale in LDCs. Kremer's (1993) paper, which we surveyed in Section 3, echoes both Simon (1992) and Kuznets (1960) in suggesting that a larger population scale increases the rate of technological change, which brings about a higher rate of economic growth. Razin and Ben-Zion [R&B] (1975) investigate the implications of endogenous fertility in the context of a neo-classical growth model. They pursue a dynastic utility maximization problem, based on pure altruism, in which the number of children is an argument and a choice variable, and government subsidies, financed by income taxes, provide parents an opportunity to increase either investment per child, the number of children, or both. The subsidy they consider is proportional to the amount of parental investment per child, and the tax is proportional to wealth. Their analysis suggests that the combined subsidy-cum-tax effect is to increase investment in children but lower their optimal number. Since longevity is constant in their analysis, the net effect is a reduction in the rate of population growth. This does not mean, however, that such government intervention is welfare improving. A similar framework was subsequently used by Nerlove, Razin, and Sadka (1986) to demonstrate that the competitive market solution for population is efficient. Child allowances and tax exemptions, (which are financed by lump-sum taxes) designed to alleviate poverty and reduce inequality, will create distortions by artificially lowering the cost of children to the parent. In fact, welfare is reduced because parents can achieve the post-subsidy allocation under laissez faire. Eckstein, Stern, and Wolpin [ESW] (1988) construct a model partly similar to Nerlove et al., except that they rely upon an overlapping generations structure rather than a dynastic approach. In the ESW model, the parent's utility is defined over young-age consumption, oldage consumption, and the number of children, and production is a function of labor, capital, and a fixed supply of land, subject to a constant-returns-to-scale technology. The value of land in this economy depends on the future path of land per capita. The model shows that, as long as fertility is endogenous, population growth will not be excessive and a decentralized economy will attain a steady-state per capita consumption level above subsistence, because selfish individuals (who also care about the value of their land) will exercise fertility control voluntarily. The resulting equilibrium is also shown to be efficient. A Malthusian trap can occur only if the rate of fertility is exogenous and high.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 33

The preceding models, however, do not consider growth itself to be endogenous. In Becker et al.'s (1990) endogenous growth model (see Section 6.1) based on pure altruism, one of the stable equilibria positions remains a Malthusian trap. Moreover, because of the existence of an unstable development equilibrium as a critical divide between growth and stagnation, population control, or other 'pro-growth' policies by the government, may be effective means of nudging the economy toward the growth outcome. The Ehrlich and Lui (1991) model, which also allows for intergenerational transfers between parents and children, has an implicit implication concerning population control policies. Since these generally imply an increase in the cost of children's quantity, this can bring about a reduction in the desired quantity of children in an early phase of an economy's transition from a stagnant to a growth equilibrium. Since in later stages fertility may reach a minimal level, however, a higher cost of children will then bring about a decline in the steadystate level and rate of growth of the income path. This model also suggests that under an efficient family system of intergenerational transfers population control policies would not be Pareto improving. 7.2. Social security, population, and growth In Section 5, we have already discussed an aspect of the literature on social security and population as it relates to tests of the relative importance of the old-age security motive for having children. In this subsection we focus on the effects of social security on the determinants of an economy's growth path, as well as on related welfare implications in studies viewing population, growth, or both as endogenous variables. We also discuss some papers dealing with the effects of fertility and longevity on the viability of social security programs. One may interpret the Ricardian Equivalence proposition to imply that not only private savings can be immune to social security, but fertility as well. In Section 5.1 we have already surveyed a number of studies showing that an increase in social security taxes can lower the desired fertility rate either temporarily or permanently even if parents are altruistic towards their children. One reason for the permanent effect on fertility is that social security, by increasing the net cost of children, also raises the economy's capital-to-labor ratio and lowers the steady-state rate of interest in a closed economy. In these studies, growth is exogenous to the economy. Ehrlich and Lui (1990) and Becker (1992) point to a more basic reason that applies more generally, regardless of the effect of social security taxes on the capital market: a conventional (unfunded) social security system creates a discrepancy between private and social gains from having children. Parents do not take into account the benefits their children confer on other families through the taxes their children pay to support other people's elderly parents. This argument has been used earlier by Willis

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 34

(1986) and Wildasin (1990) to show that the Ricardian Equivalence proposition does not hold for fertility choices. Nishimura and Zhang [N&Z] (1992) make a related point. They comment on a line of thought in the literature that sees social security as enforcing an optimal allocation of gifts to the old by the young in a steady state of sustainable allocations (also see Veall, 1986; Verbon, 1988). N&Z point out that even if the government sets a pension scheme at the optimal gift level, this would result in an optimal allocation only if savings were zero and fertility were exogenous. Allowing for both to be endogenous, they show that a mandated social plan would result in a stationary allocation different from either an optimal or a Nash equilibrium outcome. In their analysis, however, the steady-state level of per capita income is still constant over time. Ehrlich and Lui (1990,1994) have explored this line of research further in the context of a model of endogenous growth in which both quantity and quality of children are choice variables, parents are either selfish, altruistic, or both, savings opportunities are either present or absent, and a competing family insurance system, based on implicit intergenerational contracts, is either fully viable, inoperative, or subject to risks of default because of nonsurvival or misbehavior of adult children. All models produce the same basic result concerning the real effects of social security: at least one of the determinants of the economy's growth path - fertility, savings, or human capital accumulation, hence growth, must be adversely affected, although not necessarily all three. The basic reason is an externality or 'moral hazard' effect due to the structure of payments and benefits. Given an unfunded social insurance scheme, such as the pay-as-you-go (PAYG) systems in the US and most other Western countries, there is little linkage between parental investment in either the quantity or quality of their children and the uniform mandated old-age benefits they obtain from the social security administration. Parents cannot realize the benefits from a marginal increase in their investments in the number or productivity of their children (through investment in education), because any increase in the latter's contribution to the social security fund will benefit essentially other elderly persons. An increase in social security taxes therefore increases the marginal rate of substitution in consumption across old and young age relative to the rate of return young parents derive from their investments in children, and this leads to a reduction in the incentive to invest in either quantity or quality of children. The existence of savings opportunities doesn't change this conclusion. Since in equilibrium the rates of return on savings must be equalized to those on investment in the number or the human capital of children, at least one of these variables must fall. Under some additional assumptions, the model also suggests how these variables are likely to be affected at different phases of economic development. In a less developed country, where fertility is high, the incentive effects of higher social security taxes are likely to produce a reduction in fertility. In this case,

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 35

investment in human capital and savings may either increase or decrease, essentially because lower fertility also reduces the marginal cost of investment in human capital. In a developed country experiencing self-sustaining growth, in contrast, fertility falls to a low, and perhaps a minimal level. The moral hazard effect of social security may not lower fertility further. Under an actuarially fair family insurance system the steady-state savings rate will be unaffected, but investment in human capital, and hence growth, will unambiguously decline. These implications appear to be consistent with a regression analysis of some available cross12 country data on fertility and growth. The effect of social security on the economy's growth path is not sufficient to determine the welfare implications of social security. Samuelson's 1958 consumption-loans paper has popularized the idea that the competitive equilibrium in an overlapping generations model may not be Pareto optimal (see also Samuelson, 1975b). For example, if population is increasing over time, and if the rate of return to investing (or storing) part of the earnings (endowments) of the young generation is lower than the population growth rate, n, then a competitive equilibrium may not produce an efficient resource allocation. The latter may be improved by introducing a costless mechanism that forces transfers of goods from the young to the overlapping old generation through a pay-as-you-go scheme. The argument is subject, of course, to some obvious caveats. If the total fertility rate is below the population replacement level, then the rate of return on pay-as-you-go transfers will never dominate the one under a competitive system. Even if the size of the younger generation always exceeds that of the old, the Ponzi-like bonus from increasing population would mean that the higher the fertility rate, the greater would be the improvement in social welfare. But the incentive to keep increasing the fertility rate would ultimately lead to undesirable outcomes (see Samuelson, 1975a). For a few generations welfare may improve. But as population kept rising, the law of diminishing returns would set in, leaving all subsequent generations worse off. Studies favoring a PAYG social security system have also generally treated population and productivity growth as exogenous, thus ignoring the adverse incentive effects of a PAYG system on parental incentives to invest in the number or education of children summarized in the preceding paragraphs, which would lower the system's internal rate of return and contribute to its own demise. Moreover, the argument concerning the inefficiency of the competitive system is valid only to the extent that the latter fails to produce any voluntary

12

1 Simultaneous empirical tests of the effects of social security on economic growth as well as fertility and savings are still unavailable. Gordon (1988) presents some rudimentary statistical evidence showing that social security may have a negative effect on the growth rate in the more developed countries.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 36

intergenerational transfers within families. As Ehrlich and Lui (1994) demonstrate, under an ideal family insurance system, where families optimize on the rate of intergenerational transfers through actuarially fair implicit contracts that also affect the optimal rate of parental investments in the quantity and quality of offspring, the private system would achieve a firstbest population and growth path. Only a social security system that simulated such a voluntary system of intergenerational transfers could in principle be welfare-improving because it would eliminate any risks of default inherent in a nuclear family system (because of nonsurvival or noncompliance of children) without subjecting fertility and productivity growth to any moral hazard. But this would require the replacement of the defined-benefits PAYG system with one that provided social security benefits to old parents in proportion to their investments in the quantity and quality of their children.

8. Conclusion Despite the significant differences in the formal treatment of the problem of population and growth in the literature surveyed in this paper, the basic themes have remained largely the same. They include the role of scarcity of natural resources, capital market imperfections, or incomplete intergenerational links in creating the specter of a Malthusian trap, and the role of incentives and market conditions governing intra-family choices, technological innovations, and human capital accumulation, as means of lifting this specter. The same can be said about the relevance and actual consequences of public intervention through population control policies or mandatory social security schemes. Perhaps the major trend in the literature has been the progressive develop ment of a comprehensive dynamic paradigm which treats population, growth, and development as endogenously and simultaneously determined, rather than as disparate outcomes of different economic systems, market conditions, or phases of economic development. Such a comprehensive dynamic model attempts to explain the process of development and its associated demographic transition, or the direction of changes in long-term rates of population and economic growth as consequences of shifts in initial (historical) conditions or basic parameters of the model, rather than as a result of changes in contemporaneous economic or demographic variables. It also treats economic development and economic growth within a unifying analytical framework capable of explaining both growth and stagnation as potential consequences of basic conditions. Most of the work on endogenous population and economic growth to date has been theoretical. It is yet to be determined whether it can also enrich our understanding of the empirical evidence concerning the problem of development as defined in the introduction to this survey.

/. Ehrlich, F. Lui / Journal of Economic Dynamics and Control 21 (1997) 205-242 37

Moreover, the ongoing demographic transition in countries that have embarked on a regime of persistent growth creates new theoretical and empirical challenges. Whereas in the classical literature the threat to growth is believed to come from the tendency for population to explode, this concern is currently likely to be replaced with a similar concern for the tendency of population to implode because of falling birth rates and the persistent aging of the population. Some existing models of endogenous economic and population growth have recognized the role of longevity in economic growth and have reached a generally optimistic assessment of the relationship between population aging and economic growth. But these papers are still treating longevity as an exogenous variable. More complete insights into this relationship could be obtained by treating longevity itself as an aspect of economic growth and development. Another important challenge for the contemporary literature is the development of a model of endogenous population and economic growth based on heterogeneous rather than homogeneous agents. The rationale for this development is the existence of diversity in the pattern of fertility, longevity, and income growth not just across different economies, but also within the same economy at a given point in time. The progression of the literature along these lines may also provide new insights concerning the relationship between income inequality and economic growth.

Das könnte Ihnen auch gefallen