The unintended consequences of the war on poverty.

Gallaway, Lowell E. ; Garrett, Daniel G.

"Conventional wisdom suggests that a rise in federal expenditures designed to help low income groups should produce some reduction in poverty and thus some reduction in measured income inequality." This passage is taken from Vedder, Callaway, and Sollars (1988). Often, this conventional wisdom's handmaiden is a negative critique of the economy's ability to produce an equitable distribution of income. For example, Galbraith (1958) and Harrington (1962) argued that economic growth no longer had a significant impact on tire incomes of those at the bottom of the income distribution.

Ideas of this sort took a prominent place in national public policy debates in the 1960s. In the 1962 Economic Report of the President, there is a reference on page 9 to people "whose poverty is barely touched by ... improvements in general economic conditions." The report adds, "To an increasing extent, the poorest families in America are those headed by people who are shortchanged even in times of prosperity." At the same time, concern about the growth in the volume of welfare-type income transfers was on the rise. The data in this respect are compelling. Between 1953 and 1964, real per capita public aid payments rose by 70 percent while real per capita disposable income increased by only 21 percent. Yet, toward the end of this period, tire official poverty rate appeared to have become stagnant. In the six years from 1956 to 1961, the poverty rate averaged 22.5 percent and declined by just one percentage point. Neither economic growth nor a surfeit of income transfers seemed capable of making significant inroads into the incidence of poverty.

Galbraith and Harrington appeared to be prophets. However, they were not simply prophets crying out in the wilderness. People listened to them--important people, who converted their ideas into the structural poverty thesis that became the dominant theme of anti-poverty public policy. The first anti-poverty legislation, The Economic Opportunity Act of 1964, was based on the notion of structural poverty. Perhaps this explains why, when he signed that legislation, President Lyndon B. Johnson proclaimed, "The days of the dole in America are numbered."

The reality of the future of what Johnson called "the dole" turned out quite differently from what he prophesied. As already noted, the real per capita cost in the United States of federal public aid rose 70 percent in tire 11 years between 1953--the first year the federal government reported an official poverty rate--and Johnson's 1964 remarks. In the 11 years that followed, however, that same real per capita cost increased by an astonishing 434 percent--that is, more than six times faster than in 1953-64. Far from disappearing, as the president's statement suggested, the data from the early years of the "War on Poverty" suggest that the dole was flourishing (Gallaway 1965).

The Impact of the Dole on Poverty

As to the effect these increases in public aid had on poverty, in 1953-64, every 10 percentage point increase in public aid was associated with a 1 percentage point drop in the official poverty rate. Compare that with the experience of the 11 years following the outbreak of hostilities in the War on Poverty. During that interval, every 1 percentage point fall in the poverty rate was accompanied by a 50 percentage point increase in real public aid.

What these observations suggest is that the relationship between public aid and the poverty rate is subject to the principle of diminishing returns. In a more formal fashion, it can be stated as follows:

(1) dP/dA = -(a - bA),

where dP represents the change in the official poverty rate and dA denotes the change in per capita real public aid expenditures (at 2009 prices). Public aid is defined as all government social benefits to persons, less OASDHI (Social Security) stipends, unemployment compensation payments, and veterans' benefits. The relationship shown in equation (1) is confirmed by an analysis reported to the United States Congress by Danziger and Plotnick (1985). In the academic world, it is suggested by Brehm and Saving (1964) and Kasper (1968). Murray's (1984) work is also germane.

Equation (1) implies the following relationship, which is derived by the process of integration:

(2) P = k - a A + (b/2) [A.sup.2].

This is a Laffer Curve type relationship, which is to say that while public aid initially decreases poverty, there eventually comes a point at which additional increases in public aid increase poverty. The statistical properties of this relationship were explored extensively in the mid-1980s by Callaway, Vedder, and Foster (1985), as well as Gallaway and Vedder (1985, 1986), using per capita levels as the measure of real public aid. At that time, it was noted that the findings implied the effectiveness of additional real public aid expenditures, as a policy instrument designed to reduce the poverty rate, had been exhausted by the mid-1970s. Indeed, any additional public aid beyond the mid-1970s levels would result in an increase, not a decrease, in the poverty rate.

This article replicates and extends those earlier results through the year 2010. In line with earlier work in the 1980s, we introduced additional control variables in the statistical analysis to account for variations in overall economic conditions. The statistical results for the parameters a and (b/2) are reported in Table 1.1 They are consistent with the earlier analysis and are statistically significant at the 5 percent level. These parameters can be employed to calculate the impact of public aid expenditures on the incidence of poverty in the United States. The greatest poverty-reducing effect occurs at $1,291 of per capita expenditures on public aid, which produces a 6.07 percentage point reduction in the overall poverty rate. (2) However, as the level of real per capita public aid rises beyond $1,291, the poverty reducing effect is eroded. At the $1,500 per capita level, the reduction in the poverty rate falls to 5.81 percentage points; at $2,000 per capita, the poverty rate decline is only 3.74 percentage points; at $2,407 of per capita public aid, all of the initial reductions in the poverty rate have disappeared. This is summarized in Table 2.

This inflection point was passed during the first decade of the 21st century. By 2010, real per capita aid stood at $2,697--a level that produces a 2.52 percentage point increase in the poverty rate. Thus, the impact of per capita public aid in 2010 being $1,406 greater than the optimal, poverty-reducing level was to increase the poverty rate by 8.59 percentage points, according to our analysis. Since the official poverty rate in 2010 was 15.1 percent, this implies that in the absence of that extra $1,406 of per capita public aid, the official poverty rate in 2010 would have been 6.5 percent.

Counterfactual propositions, such as the 6.5 percent estimate, can be tricky. For example, this particular estimate is static, taking no account of the dynamic effects that would occur if per capita public aid was rolled back to $1,291. The decline in government spending accompanying a $1,406 reduction in real per capita public aid would exceed $400 billion. This would lead to a fall in federal spending as a percentage of gross domestic product (GDP), which was reported to be 22.0 percent in 2010. At the 22.0 percent level, federal spending has a significant negative impact on overall economic activity, as was demonstrated in a series of live monographs published in 1995 and 1996 for the Joint Economic Committee of Congress (Gallaway and Vedder 1995, 1996a, 1996b, 1996c, and 1996d). These reports identify 17.5 percent as the critical level of federal government spending as a share of GDP, beyond which any additional spending has the effect of reducing national output.

To be sure, it may be argued that a decline in public aid expenditures could generate a rise in the poverty rate because people have become dependent on public aid. If this is the case, it would suggest that the tradeoff between leisure and work-related income is not very responsive to decreases in income from public aid. However, based on the experience of 1995-2000 (the "Contract with America" years), we are inclined to think this is a minor consideration. During that period, the United States Congress slowed the relative growth in government social benefits to persons. In 1995, all such benefits constituted 13.02 percent of personal income, while public aid spending stood at 5.26 percent of personal income. In 2000, those numbers were lower at 12.36 and 4.95 percent, respectively. Over that same period, federal spending fell from 20.3 to 18.4 percent of GDP, the average real GDP growth rate was 4.11 percent, and the official poverty rate declined from 13.8 to 11.3 percent. The behavior of the family assistance subcategory of public aid over this period is particularly interesting. Real per capita family assistance spending fell by 28.7 percent in those five years. These figures do not indicate the existence of a meaningful dependency effect.

This being the case, our counterfactual estimate of the 2010 poverty rate may, in fact, be too high. Taking dynamic factors into consideration would probably lower the figure to less than 6 percent. This implies that the actual poverty rate in 2010 was more than two-and-one-half times higher than it could have been were it not for the excessive use of public aid income transfers as an instrument of policy. In other words, it may be argued that public aid overreach was responsible for approximately 30 million extra people living in poverty in 2010.

This might appear to be a devastating conclusion, but one significant qualification must be noted. The mechanism through which per capita aid income transfers operate to shift people from above the poverty line to below it involves their substituting leisure (nonwork activity) for work-related income. In the process, therefore, there may be some increase in individual satisfaction, since most people prefer leisure to labor. It follows that those who are forced back above the poverty line by reduced public aid might not consider it an improvement in their life condition. Yet this qualification must itself be qualified. Some people, most notably children in low income families, are not shifted across the poverty line willingly--that is, according to choices that they themselves make. Rather, they are at the mercy of their parents' economic decisions.

In 2010, 21.1 percent (15.75 million) of America's 70 million children were classified as living in poverty. The extent to which this can be attributed to adult decisions made in response to the availability of public aid can be estimated based on our counterfactual estimate of what the overall poverty rate would be if public aid expenditures were reduced to their optimal poverty-minimizing level. Recall that our static estimate of that rate is 6.5 percent; at that level, 55.6 percent of the actual poverty rate of 15.1 percent is induced by excessive public aid payments. If that same fraction (55.6 percent) is applied to tire total child poverty population, then 8.75 million children live in poverty. This suggests that one in every eight American children is living below tire poverty line because public aid payments exceed the level that would minimize the poverty rate. It can hardly be argued that the adult decisions that have taken those children's families into official poverty have redounded favorably on the conditions of their life.

Of course, this is a lower bound estimate of the impact of excessive public aid expenditures on child poverty, since no account is taken of dynamic effects. If those effects were to reduce the counterfactual poverty rate by another full percentage point, to 5.5 percent, the estimated number of children living in poverty due to excessive income transfers would rise to 10 million children, or one child out of every seven.

Poverty and Income Inequality

Beyond the poverty question, there is the closely related issue of inequality in the distribution of income. Very recently, this matter has resurfaced in a major fashion in a variety of arenas--academic, political, and popular--largely due to the publication of Thomas Piketty's (2014) book Capital in the Twenty-First Century, which makes an argument for the forceful use of economic policy to reduce inequality.

For our purposes, we will begin by establishing some stylized facts about income inequality in the United States. There are multiple ways to measure income inequality, such as Gini coefficients, Paglin-Gini coefficients, the ratio of the income share at the top to the share at the bottom of the distribution, or reference to movement between income quintiles over time. There are also several income-receiving units on which to base measurement of income inequality--you can use individuals, households, or families. Whatever measurement paradigm is employed, they often, but not always, tell a similar story. In this discussion, we will focus on the ratio of the share of income received by the top 5 percent of families to the share of income that accrues to the bottom quintile. The data needed to calculate such a ratio are reported for families by tire Census Bureau on an annual basis beginning with the year 1947. Such information is also available on a household basis, but only starting in 1962. We have chosen to analyze the family data because important changes in the ratio took place in the years 1947-61.

The results of calculating the necessary ratios are reported in Table 3. The data are presented in the form of five-year averages, beginning with the period 1947-51. The averaging technique is employed to smooth the data. For the earliest period, the ratio averaged 3.59. In the ensuing years, it consistently fell in each successive interval until, over the period 1972-76, it reached a value of 2.73. During this particular period, in 1974, the share of income accruing to families in the bottom quintile reached 5.7 percent, which was the high for all the years 1947 through 2011. In that same year, the share of income for the top 5 percent was under 15 percent. Between 1947-51 and 1972-76, the top 5 percent to bottom quintile ratio declined by 24 percent. This suggests that income inequality in America declined consistently for three decades.

All this changed dramatically in the years that followed. In the period 1977-81, the top to bottom ratio rose slightly to 2.76. This marked the beginning of a continuous rise in our chosen inequality statistic. What began as a very small increase rapidly accelerated. Over the next 15 years, from 1981 to 1996, it increased by 73 percent, rising to an average of 4.78 in the period 1992-96. After that, die inequality measure continued to grow, but more slowly. By 2007-11, it averaged 5.24. In about a third of a century, in other words, our measure of income inequality had nearly doubled.

In a broad sense, this is a familiar pattern, mimicking tire behavior of the poverty rate. In the case of the poverty measure, up to the mid-1970s, government cash income transfers (public aid) were increasing tire incomes of those in the bottom quintile of the income distribution by more than work-disincentive effects were reducing them. The result was a reduction in the official poverty rate. It is not a coincidence that the poverty rate reached a low of 11.1 percent in 1973, the year before the share of income garnered by the bottom quintile reached its high of 5.7 percent. However, as the volume of public aid payments continued to increase, die work-disincentive effect more than offset the income enhancements generated by the dow of public aid. As diis happened, the poverty rate began to drift upward and the percentage share of all income received by diose in the bottom quintile of the income distribution began what would turn out to be a long and steady decline.

The impact of excessive public aid payments on the share of money income received by those in the bottom quintile is illustrated by the decline in that share from 5.7 percent of total income in 1974 to 3.8 percent in 2010 (and 2011 and 2012). Such a relationship was verified in a more formal analysis presented in Vedder, Gallaway, and Sollars (1988) and extended in Gallaway and Vedder (1989). These findings reflect the Laffer Curve type effects of increasing public aid payments, and if those effects are ignored by economic policymakers, they will tend to result in outcomes that seem, at first, to be puzzling. Such results are often dismissed as unintended consequences. But unintended or not, they are real consequences. The upshot of this is that for 40 years, a policy agenda has been pursued in the name of reducing income inequality that has, in effect, produced increasing inequality.

Conclusion

This article has updated an analysis first conducted nearly 30 years ago by Richard Vedder (Vedder, Gallaway, and Sollars 1988). When the conclusions of that work were presented to the United States Congress, either in the form of appearances before congressional committees (Gallaway, Vedder, and Foster 1985) or as a major monograph published by the United States Government Printing Office (Gallaway and Vedder 1986), they were greeted with a combination of disbelief, disdain, and even, at times, ridicule on the part of the governing congressional majorities of the time. The findings of Vedder and others were ignored by those majorities as they continued along the same path they had pursued since tire inception of tire War on Poverty. And that, roughly speaking, brings us to where we stand today.

The status of the American economy at the present juncture is illustrated by the information displayed in Table 4. Six separate, but related statistics are presented for three different years, 1947, 1974, and 2010. The first of the six statistics featured is the percentage of personal income that takes the form of government transfer payments. In 1947, this number stood at 6.18 percent. At that point, one out of every 16 dollars of personal income took the form of a government transfer. By 1974, this percentage had reached 9.79 and 36 years later, in 2010, it had ballooned to 18. The transfers referred to include all government programs, such as Social Security, unemployment compensation, veteran's benefits, and the public aid payments used in our poverty rate analysis. Those public aid programs became a more significant portion of total transfers over time. In 1947, at 1.47 percent of personal income, they totaled 24 percent of the income transfer package. By 1974, public aid as a percentage of personal income had risen to 2.86 and accounted for 29 percent of all transfers. The relevant numbers for 2010 were 6.78 percent of personal income and 38 percent of all transfers.

The second of our six statistical information groups is total federal spending as a percentage of GDP. Here, the pattern is one of persistent growth, rising from 1947's 14.8 percent to 18.7 percent in 1974, and then to 22.0 percent in 2010. The significance of the overall level of federal spending lies in the finding that federal spending in excess of about 17.5 percent of GDP has a negative effect on the level of national output by slowing economic growth. By 1974, all of the output gains possible from expanding the size of the federal government had been captured and we had moved very modestly into the range where additional spending is counterproductive. Today we are very significantly into that range.

These developments are reflected in the third of our statistical categories, the average annual growth rate in GDP. Between 1947 and 1974, a simple average of the annual growth rate in the United States' GDP is 3.88 percent. Over the following 36 years, that average declines to 2.73 percent a year. What is the significance of a difference of this magnitude? To answer that question, consider a simple thought experiment. Imagine a person born in 1953, who enters the labor market in 1974 (at age 21), and then lives another 60 years until 2034. What will happen to national output during those 60 years? Assuming a continuation of the post-1974 real growth rates and setting 1974 equal to 100, real GDP in 2034 would be equal to 495. However, if beyond 1974, real output had grown at a rate of 3.88 percent a year, the index of real output would be at 930 in 2034. If typical lifestyles reflected exactly levels of real output, our imaginary individual would be enjoying a lifestyle barely half as good as it might have been.

The last three of our statistics show changes in the poverty rate and movements in the pattern of income inequality. These have been discussed in detail previously. Their behavior fleshes out the picture of an economy suffering through the early stages of what might best be called creeping stagnation. The full scenario is a straightforward one. Attempts to ameliorate economic inequalities through the War on Poverty involved escalating the volume of public aid transfers. Individual behavioral responses to this additional flow of income produced dynamic effects that led to unintended consequences. In a more informed world, these consequences would not be passed off as unintended. They would have been anticipated. Further, they would have been recognized at an early stage of their appearance. Alas, that was not the case. Instead, policymakers continued to expand the public aid expenditures that have increased the relative size of the federal government to an overall level that reduces economic growth. All these outcomes--slower economic growth, higher poverty rates, and greater income inequality--are die predictable unintended consequences of the War on Poverty.

References

Brehm, C. T., and Saving, T. R. (1964) "The Demand for General Assistance Payments." American Economic Review 54 (6): 1002-18.

Danziger, S., and Plotnick, R. (1985) Children in Povertij. House Committee on Ways and Means, 99th Congress, 1st Session. Washington: U.S. Government Printing Office.

Economic Report of the President (1962) Washington: U.S. Government Printing Office.

Galbraith, j. K. (1958) The Affluent Society. Boston: Houghton-Mifflin.

Gallaway, L. (1965) "The Foundations of the War on Poverty." American Economic Review 55 (1): 122-31.

Gallaway, L., and Vedder, R. (1985) "Suffer die Little Children: The True Casualties of the War on Poverty." In War on Poverty--Victory or Defeat? Hearing before the Subcommittee on Monetary and Fiscal Policy, Joint Economic Committee, 99th Congress, 1st Session. Washington: U.S. Government Printing Office.

--(1986) Poverty, Income Distribution, the Family and Public Policy. Joint Economic Committee, 99th Congress, 2nd Session. Washington: U. S. Government Printing Office.

--(1989) "The Tullock-Bastiat Hypothesis and Rawlsian Distribution Strategies." Public Choice 61 (2): 177-81.

--(1995) "The Impact of the Welfare State on the American Economy." Washington: Joint Economic Committee, United States Congress.

--(1996a) "The Impact of the Welfare State on Workers." Washington: Joint Economic Committee, United States Congress.

--(1996b) "The Impact of the Welfare State on America's Children." Washington: Joint Economic Committee, United States Congress.

--(1996c) "The Impact of the Welfare State on Small Business and the American Entrepreneur." Washington: Joint Economic Committee, United States Congress.

--(1996d) "The Impact of the Welfare State on the American Family." Washington: Joint Economic Committee, United States Congress.

Callaway, L.; Vedder, R.; and Foster, T. (1985) "The New Structural Poverty: A Quantitative Analysis." In War on Poverty--Victory or Defeat? Hearing before the Subcommittee on Monetary and Fiscal Policy, Joint Economic Committee, 99th Congress, 1st Session. Washington: U.S. Government Printing Office.

Harrington, M. (1962) The Other America. New York: Macmillan.

Kasper, H. (1968) "Welfare Payments and Work Incentive: Some Determinants of the Rates of General Assistance Payments." Journal of Human Resources 3(1): 86-110.

Murray, C. (1984) Losing Ground. New York: Basic Books.

Piketty, T. (2014) Capital in the Twenty-First Century. Cambridge, Mass.: Harvard University Press.

Vedder, R.; Gallaway, L.; and Sollars, D. (1988) "The Tullock-Bastiat Hypothesis: The 'Natural' Distribution of Income." Public Choice 56 (3): 285-94.

(1) The full regression model is:

P = 29.02 - 0.00987 A + 0.0000037 [A.sup.2] - 0.00028 Y + 0.3719 U, (12.28) (1.74) (2.98) (1.25) (2.38)

[R.sup.2] = 0.928, ARMA = (0, 3), N = 59,

where P denotes the official poverty rate, A represents real per capita public aid (2009 prices), Y is real per capita national income (2009 prices), and U is the annual average unemployment rate. The constant term (29.02) can be thought of as the exogenous baseline poverty rate for this set of time series observations. The value in parentheses beneath each regression coefficient is its t-statistic.

(2) This value is estimated by setting the dP/dA term in (1) equal to zero and solving (1) for A. This produces A = a/b, where h = twice the value of the regression coefficient for the quadratic term in the regression model. The second order condition for (1) indicates that this value of A generates a minimum value for P.

Lowell E. Gallaway is Distinguished Professor of Economics Emeritus at Ohio University. Daniel G. Garrett is a graduate student in economics at Duke University.

TABLE 1
REGRESSION PARAMETERS FOR ANALYSIS OF POVERTY
RATE-PUBLIC AID RELATIONSHIP

Variable             Regression Coefficient   t-Statistic

Public Aid           -0.00987                      -1.74
Public [Aid.sup.2]    0.000004                      2.98

SOURCES: Audiors' calculations. Poverty rate data from U.S.
Department of Commerce, Census Bureau. Public aid data from U.S.
Department Commerce, Bureau of Economic Analysis.

TABLE 2
CHANGES IN U.S. POVERTY RATE ASSOCIATED WITH
VARIOUS LEVELS OF PER CAPITA PUBLIC AID

Level of Per Capita Public Aid    Change in the Poverty Rate

$0                                            0
$500                                        -3.94
$1,000                                      -5.87
$1,291                                      -6.07
$1,500                                      -5.81
$2,000                                      -3.74
$2,407                                        0
$2,500                                       0.32
$2,697                                       2.52
$3,000                                       6.28

SOURCE: Authors' calculations.

TABLE 3
FIVE-YEAR AVERAGES, RATIO OF SHARE OF INCOME OF
TOP 5 PERCENT TO INCOME SHARE OF BOTTOM
QUINTILE OF INCOME DISTRIBUTION

Time Period           Average Ratio

1947-1951                 3.59
1952-1956                 3.39
1957-1961                 3.23
1962-1966                 3.04
1967-1971                 2.82
1972-1976                 2.73
1977-1981                 2.76
1982-1986                 3.25
1987-1991                 3.85
1992-1996                 4.78
1997-2001                 5.00
2002-2006                 5.18
2007-2011                 5.24

SOURCE: U.S. Department of Commerce, Bureau of the Census, Current
Population Survey, Social and Economic Supplements.

TABLE 4
Changes in the U.S. Economy, 1947-2010

Statistic                       1947       1974    2010

Government Transfers             6.18       9.79   18.0
  as a Percentage of
  Personal Income
Federal Spending as a           14.8       18.7    22.0
  Percentage of GDP
  Average GDP Growth            --          3.88    2.73
  Rate since Previous Date
Percentage Share of              5.0        5.7     3.8
  Income of Bottom Quintile
Ratio Income Share of            3.50       2.56    5.26
  Top 5 Percent to
  Share of Bottom Quintile
U.S. Poverty Rate as a          31.7 (a)   11.2    15.1
  Percentage of Population

(a) This is not an official poverty rate. The earliest year in which
the  federal government provides an official poverty estimate is
1953. The value shown here is taken from Gallaway (1965) and
represents the percentage of families with less than $3,000
annual income, measured at 1963 prices.

SOURCES: U.S. Department of Commerce, Bureau of the Census and
Bureau of Economic Analysis.