CHAPTER 10 THE OVERALL EFFECT OF IMMIGRANTS UPON NATIVES' STANDARD OF LIVING This chapter combines estimates of the partial influences discussed in previous chapters in order to arrive at an overall assessment of the effect of immigrants upon the standard of living of natives. As we proceed, it should be remembered that this chapter refers to the aggregate of all natives, employing the concept of the arithmetic average. Differential effects upon separate segments of the population through wages and unemployment, and as measured by income distribution, are discussed in Chapters 11, 12 and 13. We shall first examine the effect upon the economy, and upon natives, of immigrants simply as additional people, rather than by way of any special characteristics that immigrants have. That is, we first inquire about the observed effect of population growth historically and cross-nationally upon economic growth and level. The second section utilizes a simulation model to combine the effects of immigrants qua immigrants, as well as their effects as more people, to estimate the overall effect. THE EMPIRICAL RELATIONSHIP OF POPULATION TO ECONOMIC GROWTH Empirical evidence bearing directly upon the overall relationship between the rates of immigration and of economic growth is hard to come by. Individual countries lack observation for time-series studies, and there are too few countries with comparable histories of substantial immigration to support systematic cross-national studies; nor has anyone yet produced a successful pooled analysis. The cases of massive immigration into West Germany, Japan, and Israel after World War II (See Chapter 3 for data) would seem to show, however, that taking in large groups of people--even refugees who are all-prepared for migration--is not a barrier to excellent national economic performance. For some illumination, therefore, we turn to analyses of the effects of population growth in general. The standard presumption is that additional people--children or immigrants--have a negative effect upon the incomes of the rest of the people. The usual reasoning is diminishing returns to fixed stocks of agricultural and industrial and social capital, together with the dependency burden of additional children and the consequent need for additional "demographic investment." The special characteristics of immigrants in this regard are discussed in Chapter 4). Because this Malthusian theory is so firmly fixed in conventional thought, let us first confront the theory with the data, which flatly contradict the conventional Malthusian theory. Many of these data refer to countries when they were, or are, poorer and less-industrialized than the United States is now. But we lack a sufficiently large body of more comparable experience. And the data on less-developed stages and countries are still relevant because many of the same fundamental processes occur in economic development at all stages. Furthermore, the data on the less-developed stages of development are even more surprising to most people, and therefore the picture that they show may have greater power to make the basic point. The concurrent explosion in Europe of _b_o_t_h population and economic development from l650 onwards is important case evidence. The failure of France to excel economically in the hundred years prior to World War II despite its low birth rate is an important vignette in this history. A fuller picture is given by the samples of countries for which long-run data on the growth of population and output per person are available. Figures 10-1 and 10-2 show that there is no strong relationship, which is confirmed by correlational analysis. --------------------- Figures 10-1 and 10-2 --------------------- Concerning the experience since World War II, the overall pattern is revealing: Contrary to common impression, per person income in less-developed countries (LDC's) has been growing faster than in the more-developed countries (MDC's) (Morawetz, l978) even though population growth in LDC's is much faster than in MDC's. No evidence here of a negative connection between population growth and economic growth. Many systematic cross-country comparisons of recent rates of population growth and economic growth have been done by now. They have recently been reviewed and summarized by Lee as follows: [D]ozens of studies, starting with Kuznets' (1967), have found no association between the population growth rate and per capita income growth rate, despite the obvious fact that at least since WWII, population growth rates have varied considerably. These studies control for other factors such as trade, aid and investment to varying degrees. Two recent studies add historical depth to this analysis; even within countries (and thus looking only at disequilibrium), over periods as long as a century or as short as 25 years, there is no significant association of [the population growth rate] and [the income growth rate], for either DCs or LDCs; put differently, one can't reject the hypothesis that the regression coefficient of [the income growth rate] on [the population growth rate] is unity. I know of just two exceptions to this general picture. . . both dealing with cross-sections; both find negative effects of population growth of magnitude roughly equal to the share of non-labor inputs in production, as many theories would predict. However, data problems render these results suspect. (Lee, 1983 ? notation and references omitted. See also my earlier review, 1977, Chapter 3.) That is, population growth does not have a negative effect upon economic growth. Elsewhere (Simon, 1986) I discuss in detail why these studies constitute solid evidence of the absence of causal influence of population growth upon economic development.0 ***** These overlapping empirical studies do not prove that fast population growth in the more-developed world as a whole increases per person income. They do imply that population growth does not decrease economic growth. In the past, I was not prepared to say more than that. Now I believe that we can proceed beyond this, however: Recall that the studies mentioned above do not refer to the very long run; rather, they cover only a quarter of a century or at most a century. The main negative effects of population growth occur during perhaps the first quarter or half of a century so that, if the negative effects are important, these studies will reveal them. These shorter-run effects upon the standard of living include the public costs of raising children -- schools and hospitals are the main examples - - and the costs of providing additional production capital for the additional persons in the work force.1 ***** The observations that immigrants a) do not have a detrimental effect upon the supply of natural resources and a clean environment (Chapter 9), and b) have a net positive rather than negative effect upon the public coffers, far exceeding their negative effect through the use of public capital (Chapters 5 and 7), combined with the absence of an observed negative effect on economic growth in the intermediate-run statistical measures, are enough to suggest that in the very long run more people have a positive net effect, because the most important positive effects of additional people -- improvement of productivity through the contribution of new ideas as well as the learning-by-doing that accompanies increased production volume -- occur in the long run, and are cumulative. To put it differently, the statistical measurements of the relationship of population growth to economic growth are biased in favor of showing the shorter-run negative effects. If such negative effects do not appear, one may assume that an unbiased measure of the total effect would reveal a positive effect of population growth upon economic growth. And of course the positive effect may be expected to occur much faster with immigrants than with additional births, and births rather than immigrants are the main constituent of the empirical studies under discussion. The empirical studies mentioned above focus on the process of population growth. Examination of the relationship between the attained level of population -- that is, the population density as measured by the number of persons per areal unit -- shows an even more impressive effect. Studies of MDC's are lacking. But for LDC's, Hagen (l975) and Kindleberger (l965) show graphically, and Simon and Gobin (l979) show in multivariate regressions, that higher population density is associated with higher rates of economic growth (see Figure 10-3). The effect seems strongest at low densities, but there is no evidence that the effect reverses at high densities. Strycker shows a similar effect for agricultural productivity (l976). These data showing a positive effect of density upon economic growth also constitute indirect proof of a positive long-run effect of population growth upon economic growth, because density changes occur very slowly, and therefore, density picks up the very-long-run effects as well as the short-run effects. ----------- Figure 10-3 ----------- Greater density can have a cost in congestion, as is seen in the news story about Lower Manhattan in Chapter 9. It may at first seem unbelievable that greater populatin density leads to better economic results. As noted in Chapter 9, this is the equivalent of saying that if all Americans moved east of the Mississippi, our incomes would rise. But upon reflection, this proposition is seen not to be as implausible as it sounds at first hearing. AN INTEGRATED SIMULATION MODEL Now it is time to weave together the various influences of immigrants upon the standard of living of natives. An economist worth her/his keep must take into account the size and importance of the various effects and calculate the net effect. One can only obtain a satisfactory overall assessment of the effect of immigrants on the standard of living of citizens by constructing an integrated model of the economy and then comparing the incomes produced by the economy under various conditions of immigration and population growth.2 *** For simplicity and clarity, the model deals with a single cohort of immigrants; a continuous analysis yields similar results, however. Also for simplicity, I sometimes talk of a representative family instead of the cohort as a whole. The question is whether the resident population--that is, the people living in the United States before the immigrant family arrives--is better off or worse off economically if the immigrant family comes or does not. In more precise terms, we wish to know if the lifetime income of the (average member of the) resident population is higher or lower if immigrants come.3 Therefore, we must estimate the natives' yearly gross incomes and taxes if there are, and if there are not, immigrants. The Structure of the Simulation Model We start with the effect of the immigrant on natives' incomes through the two major lines of influence: the capital- dilution effect, and the economies-of-scale-and-productivity effect. The combined effect of these two forces will be estimated in a simple macro-economic simulation. The main conventional element is a Cobb-Douglas function, whose labor and capital coefficients add to unity and where saving is a fixed steady-growth proportion of the prior year's output. A less conventional element is the effect of output and labor force on the technological-level coefficient. In other recent work, I have explored a wide variety of technological progress functions and have found that, in a policy context such as this one, the result is rather insensitive to the choice of function. I have chosen Phelps' well-known and elegant function (though used by him for a different purpose), which is "conservative" in this sense: Phelps' function indicates that technical progress should have been progressively lower as population growth has declined in the twentieth century in the United States and in Western world generally. In fact, technical progress has apparently been higher in the more recent decades than in the early decades than in the early decades of this century as discussed earlier. This implies that Phelps's function understates the contribution of population size and growth to the advance of economic welfare.4 In place of the size of the research force in Phelps' function, for simplicity I have used the size of the labor force. The function was written in Cobb- Douglas form to make its meaning obvious: gamma delta (10-1) At - At-1 = bAt-1 Lt-1, with gamma, delta = 0.5 The exponents fit Phelps' requirement that the function be homogenous of degree one, and his assumption that "if the technology level should double we would require exactly twice the amount of research to double the absolute time rate of increase of the technology." The assumption of the steady-state savings rate is also conservative in the sense that it is less advantageous to a larger population (and hence to immigrants) than would be a higher savings rate. This is reasonably clear upon inspection and is verified in other work by this writer. The coefficient b is that complement of the initial values chosen for A and L that starts the simulation smoothly into motion and that corresponds to the steady-state rate of change of A in the nonimmigrant case, which is equal to the rate of growth of the labor force. This, too, is a conservative assumption. An iterative program is used to make investment a function approximate of current-period income rather than prior-period income, so that the computer model would approximate the steady-state analytic model. The results are much the same with and without this refinement, however. The other equations and parameters of the model are as follows: alpha beta (10-2) Y = A L (10-3) K , K = sY + K t t t t t t-1 t-1 and (10-4) Lt = Lt-1 + 0.02 Lt-1. The initial values are At + 1.0, Kt = 1000, Yt = 500, alpha = 0.67, beta = 0.33, gamma = 0.5, and delta = 0.5; b is chosen so that the initial rate of change of A equals 0.02 yearly. The initial L equals 1,000 for the without- immigration case, and 1,020 for the with-immigration case.5 For the income-effect calculations, the increment of immigrant workers in period t=1 must be large enough so that the effects are not obscured by rounding error. It was therefore set equal to the 2 percent increase in native labor force in year t=1 (10 percent in some runs to show that the size of the increment matters little). Then the difference in citizens' incomes in future years between the situations if the immigrants do come in t=1 and if they do not come are calculated. The final calculation is in terms of the effect of one additional immigrant. A key issue is the "returns" to the capital that the immigrant works with, and the payment for the additional demographic capital required to take care of the publicly- supplied services such as schooling that the immigrant family needs, as discussed in Chapter 7. Properly, each of these matters should be handled separately and in its full detail. The simulation, however, deals with capital in a much cruder fashion, by simply choosing two alternative levels of "returns to capital" and letting the level stand as the proxy for all the capital effects. One level chosen for experimentation is 20%; it seems at least as high as any plausible calculation would have it, on the following reasoning: Assume that a) 8% of immigrants (those who work for government) get the full returns from the capital they work with, and the other 92% get none of the returns (see chapter 7); b) no allowance need be made for private production capital, (reasons spelled out in chapter 7), and no allowance need be made for public capital such as roads and defense equipment which are essentially public goods; and c) demographic capital, including schools and hospitals, is 24% of all capital,6 and immigrants get a full share of that while paying for only (1-21/38=) 45% of it through bonds and taxes (see Chapter 7). The effects in (b) and (c) are not nearly substantial enough to raise the 8% in (a) to 20%, the figure in the main computation. Additionally, computations are shown for a level of returns to capital of 35%, far higher than is conceivable; the reason for including this level is to show that the conclusions drawn from the simulation are not sensitive to the level of capture of returns to capital that is assumed in a given run. Results Table 10-1 shows calculations with the two capture-of- capital-returns assumptions. Considering first the results without the effects of welfare transfers and taxes, and working with the 20% level, the pretax effects on citizens' incomes amount to the percentages of the immigrants' net income shown in column 1. Those figures may be interpreted as follows. In year 1, aside from taxes, citizens' incomes are (in the aggregate) lower by 7 percent of the income of the average immigrant, (though the effect on individual natives is small because of the small proportion of immigrants relative to natives). By year 7, citizens' net incomes are higher than they would otherwise be, because of the immigrants. By year 13, citizens' incomes are higher by an amount equal to 10 percent of the income of each immigrant who arrives in year zero. -------------- TABLE 10-1 -------------- Next we take into account the immigrants' savings-and- transfer effect, as discussed earlier. Social Security is the main issue. Immigrants collect no Social Security, both because of age distribution and because thy have no claims to benefits until they have worked for years. The immigrant family's contributions (will be roughly?) are assumed to be 10% of income (actually 12.8% of personal income in 1983; Social Security Bulletin, Annual Statistical Supplement, 1985. p.66). This makes the account slightly positive in year five and thereafter, as seen in column 3. Overall, the stream of negative and positive effects may be evaluated just as any other investment, with negative outgo's at the beginning and positive incomes later on. On a capital- returns assumption of 20 percent, the rate of return on the investment decision to bring in an immigrant is 18.4 percent per annum without the Social Security effect and 28.4 percent with it, an excellent investment by any standard. The results of a variety of other specifications of the basic model with respect to savings rate, initial rate of technical progress, proportion of returns to capital captured by immigrants, and exponents of the technical progress function are shown in Table 10-2. -------------- TABLE 10-2 -------------- DISCUSSION The model does not include a variety of other effects discussed elsewhere in the book. Hence a brief discussion of how they fit here seems appropriate. 1. Transfer welfare payments (excluding net retirement benefits) and taxes are likely to net out zero, as discussed in Chapter 5. 2. There is no reason to think that illegal immigrants are different than legal residents with respect to the issues discussed in this chapter. With respect to welfare and taxes, they bestow a special boon upon natives, as discussed in Chapter 15. 3. Case-by-case cross-national-comparisons of the effect of total immigration upon aggregate economic performance can provide extreme cases in either direction. For example, Japan has shown extraordinary economic performance in the total absence of immigration (except for the movement of Japanese from overseas immediately after World War II). On the other hand, Germany showed extraordinary economic performance after World War II in the presence of very large numbers of immigrants from outside West Germany; in the 1970's, when immigration was low or negative, economic performance was less impressive. The worst economic performance of the U. S. was during the 1930's, when immigration was negative. And so on. There are two reasons why one should not at present draw any conclusions from such examples: First, the number of examples is too small. And second, the causal relationship between economic performance and immigration runs in both directions, as in the U. S. in the 1930's. 4. The model implies that a given immigrant increases the standard of living of natives. But nothing has been said about whether the immigrant being evaluated is number 1 or number 100,000 or number 1,000,000 or number 10,000,000 in a given year. The question then arises as to whether the impact varies with the number of immigrants. In principle, immigrants at the rate of 1 per 100 natives should be much easier to absorb than at the rate of 1 to 10, or 1 to 1, or even 10 to 1. At the latter rate, it is the natives who would be absorbed into the immigrant culture if the immigrants are homogeneous. The question may be schematized with Figure 10-4. We wish for data to help us know which line in the diagram portraying the absorption process is not appropriate. We are interested both in the general shapes of the curves, and their absolute heights at each point. --------------- Figure 10-4 --------------- Satisfactory data surely will be hard to find. We may try to consult history (the U.S. at the turn of the twentieth century; the U.S. other periods; Germany and Japan after World War II; Israel in the 1950's versus other years). We may also look at a cross-section of U.S. cities with varying amounts of immigration. Please notice that the entire issue may well be one of time, however. That is, it may be that in a generation or so with immigrants of equal education, adjustment has taken place to the extent that income is no lower than if no immigrants have come (industry, capital effects to be discussed later). The second issue is education. If many immigrants come with no skills, it is reasonable that the present high-level productive capacity of the U.S. would be diluted so that the U.S. income level would be like the poor countries that the immigrants came from, as discussed in Chapter 8. From a practical point of view, in my judgment it can safely be said that at present levels of immigration, there is no difference in absorption between the first and the marginal immigrant. And judging by experience at the turn of the century, when immigration was relatively much higher, and judging also by experience in countries that have had much higher levels of immigration (see Chapter 3), this could also be said at much higher levels of immigration than at present. SUMMARY AND CONCLUSIONS Almost two decades of well-done statistical analyses show that additional persons qua persons rather than just as immigrants (as measured by the rate of population growth) do not imply a slower rate of economic growth. These analyses compare the sample of nations for which data are available back into the nineteenth century, and also the larger sample for which data are available after World War II. The analyses have been conducted in a variety of ways. Yet almost every study agrees with the others that population growth does not have a negative effect. These aggregate statistical studies are biased toward showing negative shorter-run effects if they exist; their time- span is too short to show fully the longer-run positive effects of added persons upon the rate of growth of technology and productivity, the most important effect of added people. Therefore, given the lack of negative effect in the short and intermediate run and the positive effect after decades, it is fair to conclude that the longer-run effect in the standard of living of natives of adding people--especially immigrants--is positive, based on the empirical evidence. The chapter describes a simulated macro-economic model which compares the incomes of natives with and without a hypothetical cohort of immigrants. This first model confirms the conclusion of Chapter 7 that the possible gain to natives through increased returns to native capital, if the immigrants receive only their marginal product, is small relative to the loss to the natives if immigrants receive a realistic proportion of the returns to capital. And the life-cycle saving-and-transfer process works in a positive direction for natives, and is larger than the capital- dilution effect, according to chapters 5 and 7. The simulation shows that adding the saving-and-transfer process indeed more than offsets the capital-dilution effect. The crucial effect of the immigrant upon productivity also is taken into account in the model, though this phenomenon has been omitted from previous work on the subject. This effect is the sum of learning by doing, creation of new knowledge, and economies of scale of various sorts. Within a few years, the productivity effect comes to dominate the results and dwarfs the capital-dilution and saving-and-transfer effects, yielding a high rate of return to natives on investment in immigrants, on any reasonable parameters. 86-85 Overall1 1/16/87 FOOTNOTES 0Here are the points in brief: First, two-variable correlation studies certainly do not indicate the forces that influence economic development. But a two-variable zero correlation can be very strong evidence, especially when buttressed by multivariate studies with a variety of specifications, that one variable (population growth) does not cause the other (economic development). Second, many of the studies of population and economic development have indeed gone beyond simple two-variable correlations. Third, not only does a correlation not "prove" causation, as the popular saw has it, but no other scientific procedure--not even a lengthy series of experiments--can "prove" causation, either. Fourth, simple correlations of the rate of economic growth and the rate of population growth are biased toward a more negative (less positive) correlation, because the appropriate measure of economic development is the rate of change of output per worker while the usual variable is per person; substituting the former for the latter pushes the correlation coefficient in a positive direction. 1This point is a bit complex. Only the additional capital needed for the public sector would reduce native wealth and income. The additional capital investment in the private sector would reduce native consumption but it would not reduce their wealth and income, because the expenditures on the new capital simply mean a substitution of investment for consumption. 2After finishing this work, I discovered an interesting model by Ekberg (1977) that also makes technical progress endogenous in a migration context. Ekberg uses a Kaldor-like function, where the increment to technical progress depends upon the percentage change in the stock of capital. Elsewhere (Simon, 1986), I argue, however, that Kaldor's function is not appropriate for a study of this sort. My 1976 article on Russian immigration into Israel is the only other study of the sort that I know of. 3That is, we want to know whether Zm >Z or Zn >Z . Lifetime incomes with and without the immigration are, for our purposes here, functions of gross income less taxes v v v v v Z = (G - T ) + d(G - T ) + ... and m m,t=1 m,t=1 m,t=2 m,t=2 v v v v v Z = (G - T ) + d(G - T ) + .... n n,t=1 n,t=1 n,t=2 n,t=2 4Phelps' function can be made more realistic with the addition of arguments representing the effect of educational level and national income on the production of technology, with the function still retaining its convenient mathematical properties. See Steinmann and Simon (1970) or Simon (1986). 5There would appear to be no danger here that the choice of production function forces the outcome, as in the case in some studies of distributive shares. The cohort of immigrants whose effect is analyzed is small relative to the native population, and hence its effect upon the overall distribution between capital and labor is small. Also, Ekberg (1977) experimented with a CES function and obtained the same results as with a Cobb-Douglas model. 6Simon and Heins (1985) estimate that in 1975 the value of state and local capital was $781.5 billion, whereas the value of non-residential business capital in 1975 was $2470.6 billion (Statistical Abstract, 1981, p. 545). 86-85 Overall1 1/16/87 TABLE 10-1 THE EFEECT OF AN IMMIGRANT ON THE INCOMES OF NATIVES AT VARIOUS ASSUMPTIONS ABOUT THE PROPORTION OF CAPITAL RETURNS THAT GO TO IMMIGRANTS - (EXPRESSED AS A PERCENTAGE OF THE IMMIGRANT'S EARNINGS) Capture of Returns to Capital _________________________________________________________________________ 20% Capital Return 35% Capital Return ________________________ __________________________ Income Income effect Social effect Social Year (%) security Total (%) security Total Column (1) (2) (3) (4) (5) (6) _________________________________________________________________________ 1 - 7 + 10 + 3 - 14 + 10 - 4 2 - 7 + 10 + 3 - 12 + 10 - 2 3 - 5 + 10 + 5 - 11 + 10 - 1 4 - 4 + 10 + 6 - 10 + 10 0 5 - 2 + 10 + 8 - 8 + 10 + 2 6 - 1 + 10 + 9 - 7 + 10 + 3 7 1 + 10 + 11 - 5 + 10 + 5 8 2 + 10 + 12 - 4 + 10 + 6 9 4 + 10 + 14 - 3 + 10 + 7 10 5 + 10 + 15 - 1 + 10 + 9 11 7 + 10 + 17 0 + 10 + 10 12 8 + 10 + 18 + 2 + 10 + 12 13 10 + 10 + 20 + 3 + 10 + 13 14 11 + 10 + 21 + 4 + 10 + 14 15 13 + 10 + 23 + 6 + 10 + 16 16 14 + 10 + 24 + 7 + 10 + 17 17 16 + 10 + 26 + 8 + 10 + 18 18 17 + 10 + 27 + 10 + 10 + 20 19 18 + 10 + 28 + 11 + 10 + 21 20 20 + 10 + 30 + 12 + 10 + 22 21 21 + 10 + 31 + 14 + 10 + 24 22 23 + 10 + 33 + 15 + 10 + 25 23 24 + 10 + 34 + 16 + 10 + 26 24 25 + 10 + 35 + 18 + 10 + 28 25 27 + 10 + 37 + 19 + 10 + 29 26 28 + 10 + 38 + 20 + 10 + 30 27 30 + 10 + 40 + 21 + 10 + 31 28 31 + 10 + 41 + 23 + 10 + 33 29 32 + 10 + 42 + 24 + 10 + 34 30 34 + 10 + 44 + 25 + 10 + 35 _________________________________________________________________________ TABLE 10-2 RATES OF RETURN ON INVESTMENT IN IMMIGRANTS FOR A VARIETY OF MODELS (Increment of immigrants equal to 2 percent of labor force in t=1) __________________________________________________________________________ Rates of Return per Annum in Percent Capital ______________________________ Capture Without With social b s , (%) social security security __________________________________________________________________________ 0.02 0.04 0.5 0.2 18.4 28.4 0.02 0.04 0.5 0.35 9.3 19.3 0.02 0.07 0.5 0.35 12.2 22.2 0.02 0.10 0.5 0.35 14.8 24.8 _________________________________________________________________________