Vol. 16 No. 1

 WELFARE AND THE CULTURE OF POVERTY 

 William A. Niskanen

At the dawn of the American welfare state, in his 1935 State of the Union message, President Franklin D. Roosevelt (1938: 19-20) proposed social security, unemployment insurance, and (what was then called) aid to dependent children to help the deserving poor, but he added an ominous warning:

The lessons of history, confirmed by evidence immediately before me, show conclusively that continued dependence on relief induces a spiritual and moral disintegration fundamentally destructive to the national fiber. To dole out relief in this way is to administer a narcotic, a subtle destroyer of the human spirit. It is inimical to the dictates of sound policy. It is a violation of the traditions of America.

More than 60 years later, it should be clear that the narcotic of "continued dependence on relief" is less subtle and more destructive than Roosevelt feared.

 
Summary

Welfare is both a consequence and a cause of several conditions best described as social pathologies. These conditions include dependency, poverty, out-of-wedlock births, nonemployment, abortion, and violent crime. The basic hypothesis of this study is that welfare dependency and the other pathologies are jointly determined and are derivative of a common set of other conditions.

Differences in the levels of these conditions among the states provide a basis for estimating the specific effects of welfare benefits, the relations among the social pathologies, and the extent to which the pathologies are based on a common set of root causes.

Analysis of the state data for 1992 yields the following estimates of the effects of an increase in Aid to Families with Dependent Children (AFDC) benefits by 1 percent of the average personal income in the state: the number of AFDC recipients would increase by about 3 percent; the number of people in poverty would increase by about 0.8 percent; the number of births to single mothers would increase by about 2.1 percent; the number of adults who are not employed would increase by about 0.5 percent; the number of abortions would increase by about 1.2 percent; and the violent crime rate would increase by about 1.1 percent.

The social pathologies associated with the current welfare system no longer seem acceptable, not so much because of their fiscal costs but because of their malign effects. An important question addressed in this study is the extent to which these pathologies are dependent on conditions that could be changed by government policy.

Social Pathologies

For this study, the six conditions are described as social pathologies, not because they are necessarily illegal or immoral in an individual case but because the level of these conditions is broadly considered as undesirable. There is less consensus, however, about the relative undesirability of these conditions. An increase in welfare dependency, for example. may be considered desirable if it reduces one or more of the other conditions. As it turns out, however, an increase in AFDC benefits increases all of the six pathologies that are the focus of this study.

Welfare Dependency

In 1992, 5.4 percent of the national population were dependent on cash benefits from AFDC, with a range from 2.0 percent in Idaho to 10.8 percent in the District of Columbia. This program is jointly financed by the federal and state governments and is administered by the states subject to numerous federal guidelines. All AFDC recipients are also eligible for food stamps and medicaid, and many also receive benefits from special food programs, utility assistance, and housing assistance.

An additional 2.6 percent of the population receive cash benefits from other federal and state programs, for a total of 8 percent who are dependent on cash benefits. A broader 9.9 percent of the population receive food stamps, and 11.9 percent are covered by medicaid. A more complex study would be necessary to estimate the causes and consequences of the broader set of means-tested programs. For this study, the level of welfare dependency is defined as the percent of the population that receive cash benefits from the AFDC program.

Poverty

A total of 14.5 percent of the population have money income below the official poverty line, with a range from 7.6 percent in Delaware to 24.5 percent in Mississippi. The national poverty rate is now about the same as when the War on Poverty was instituted 30 years ago, despite the expenditure of over $5 trillion (at 1993 prices) for means-tested public assistance programs in the intervening years and a 75 percent increase in average real income. More means-tested benefits may or may not have contributed to the incidence of poverty but they have clearly not reduced it.

Any definition of poverty, of course, is somewhat arbitrary, depending on what types of income are included. The government estimates national poverty rates for 15 different aggregations of income, taxes, and transfers, with a range from 10.4 percent based on all after-tax income and transfers to a high of 22.6 percent based only on pre-tax money income. For this study, the level of poverty is defined as the percent of the population with pre-tax money income and cash transfers below the official poverty line, the only such data available by state.

Out-of-Wedlock Births

Out-of-Wedlock births are the most rapidly increasing social pathology. On a national basis (in 1991), 29.2 percent of births were to single mothers, with a range from 14.3 percent in Utah to 65.9 percent in the District of Columbia. Since 1960, the illegitimacy rate has increased from 2.3 percent to 22 percent for whites and from 21.4 percent to 68 percent for blacks. A substantial part of the current generation of inner city young people has grown up without a father, a contributor to the increase in violent crime and the decline in school performance as well as to some of the pathologies addressed in this study.

Nonemployment

The percent of the adult population that is not employed has been declining for many years and, compared to many other nations, is unusually low. For lack of a better word, I will define this condition with the inelegant word nonemployment, because the word unemployment has usually been used to describe those not working but seeking work. Formally, the nonemployment rate is the percent of the civilian noninstitutional population age 16 and over that is not employed. On a national basis the nonemployment rate is 38.6 percent with a range from 31.1 percent in Nebraska to 51.7 percent in West Virginia.

Abortion

Abortion is the most contentious issue in contemporary American politics, primarily because the polar positions have dominated the debate. The "pro-choice" advocates consider any restriction on abortion as a violation of a woman's rights. The "pro-life" advocates consider any abortion as murder. No study of the causes and consequences of abortion would reconcile these positions. Most Americans, however, appear to favor legal and social rules that would make abortion legal, safe, and rare. For the most part, abortion is legal and safe but it is not rare. On a national basis there are 379 abortions per 1,000 live births (somewhat lower than the prior peak of 436 in 1983), with a range from 74 in Wyoming to 1,104 in the District of Columbia. For the broad group of Americans who regard most abortions as undesirable but not a crime, abortion is appropriately described as a pathology and an understanding of the conditions that explain the large variation in abortion rates can be valuable to aid their judgment on this contentious issue.

Violent Crime

Violent crime may be the most serious social pathology in the United States. The reported violent crime rate has increased substantially for several decades and is much higher than in other nations. And the number of violent crimes estimated from victimization surveys is much higher than the number reported to and by the police. On a national basis, there were 758 reported violent crimes per 100,000 residents, with a range from 83 in North Dakota to 2,833 in the District of Columbia.

Root Causes of the Culture of Poverty

One objective of this study is to identify the extent to which the six social pathologies are the consequence of a common set of root causes. The conditions examined included the level of welfare benefits in each state, a measure of general economic conditions, the racial and ethnic composition, and several social and cultural indicators.

 Welfare Benefits

The one variable that best reflects the welfare policy in a state is the level of AFDC benefits per recipient household. State governments also set eligibility standards subject to federal guidelines, but differences in these standards are not easily measurable. For this study, welfare benefits are defined as the annual AFDC benefits per recipient household as a percent of the pretax personal income per capita in the state. On a national basis such benefits are 23 percent of personal income per capita, with a range from 10.4 percent in Mississippi to 40.8 percent in Alaska.

Total welfare benefits per recipient household, one should recognize, are much higher than the direct cash benefits from AFDC. The total benefits for those households from AFDC, food stamps, and medicaid only range from 50 to 100 percent of personal income per capita. And the total benefits for those AFDC recipients who also receive housing assistance, utility assistance and specialized food programs range from 90 to 125 percent of personal income per capita. These estimates of total benefits provide a better sense of why welfare has become a trap for so many women. For reasons that are not clear, however, the several measures of social pathology that are the focus of this study are more closely related to the narrow cash benefits from AFDC than to the broader measures of total welfare benefits, maybe because many AFDC recipients value cash benefits more than noncash benefits or are not eligible for the broader set of benefits.

Economic Conditions

General economic conditions in a state are represented by the pretax personal income per capita. On a national basis average personal income is $20,105, with a range from $14,082 in Mississippi to $27,909 in the District of Columbia.

Race and Ethnicity

The racial and ethnic composition in a state is represented by the percent black and the percent Hispanic. Blacks are 12.4 percent of the national population, with a range from 0.3 percent in Montana to 65.0 percent in the District of Columbia. Hispanics are 9.5 percent of the national population, with a range from 0.4 percent in West Virginia to 38.2 percent in New Mexico. These two groups are combined in a percent minority measure when preliminary tests indicate that their effects are not statistically different.

Social and Cultural Indicators

Social and cultural conditions in a state are represented by four indicators: church membership, educational level, percent metropolitan, and average temperature. The distinctive attribute of each of these indicators is that they reflect individual choices of whether to join a church, continue education, and where to live.

Church membership is measured as the sum of the percent of the resident population who are Christian adherents (1990) plus the percent who are Jews. By this measure, 55 percent of the national population are church members, with a range from 32.6 percent in Alaska to 80 percent in Utah.

The educational level in a state is measured as the percent of the resident population age 25 and over with a high school or higher education (1990). On a national basis 75.2 percent are educated at this level, with a range from 64.3 in Mississippi to 86.6 percent in Alaska.

The metropolitan population is measured as the percent of the population resident in metropolitan areas. On a national basis 79.7 percent of the population are residents in a metropolitan area, with a range from 24 percent in Montana to 100 percent in the District of Columbia.

The average daily low temperature measure used in this study is roughly proportional to the distance from North Dakota. This variable, which is significant only in the welfare dependency regression, reflects some combination of the tighter welfare eligibility standards in the southern states and the social and cultural differences among regions that are correlated with temperature. Whatever the balance of these effects, including this measure is important to increase the precision by which the effects of other conditions is estimated.

Methodology

The patterns of pathology are estimated by least-squares regression techniques. For each of the six pathologies, two types of relations are estimated: The first relation includes one or more other jointly determined pathologies and a subset of the root causes; this relation is estimated by a weighted two-stage (TS) regression where the whole set of root causes is used as instrumental variables. The first relation provides estimates of the relation among the several pathologies and the partial effects of specific root causes given the level of the other included pathologies. The second relation includes only a set of root causes; the relation is estimated by a weighted least squares (LS) regression. The second relation provides estimates of the total effect of specific root causes that operate both directly on the specific pathology and indirectly through their effect on other related pathologies. All variables in both relations are weighted by the resident population of each state. This increases the relative effects of conditions in the largest states and makes the estimates correspond more closely to the effects of national conditions. Only those variables that are statistically significant at a 95 percent level or more are included in either of the relations; as it turns out, most of the included variable are significant at a much higher level.

Patterns of Social Pathology

The patterns of social pathology are summarized in Tables 1, 2, and 3. First, some general advice on reading these tables. The top tier of coefficients in each table are estimates of the percent change in the focus pathology from a 1 percent increase in some other condition. For example, in Table 1, a 1 percent increase in the poverty population in a state increases welfare dependency (the number of AFDC recipients) by about 0.6 percent. The lower tier of coefficients (except for temperature) are estimates of the percent change in the focus pathology from a 1 percentage point increase in some other condition. For example, in Table 1, a 1 percentage point increase in the population living in a metropolitan area increases the AFDC population by about 1.3 percent. The coefficients on the temperature variable are estimates of the effect of a 1 degree (Fahrenheit) increase in the average daily low temperature. The numbers in parentheses are the standard errors. If the estimate of the coefficient is unbiased, there is a 95 percent probability that the true (unknown) level of the coefficient is within two standard errors of the estimate. The R² is the percent of the weighted and unweighted variance of the focus pathology among the states that is explained by each relation.

Welfare Dependency

The patterns of welfare dependency summarized by the first two columns of Table 1 support the following conclusions:

1. Welfare dependency is strongly related to poverty. Specifically, a 1 percent increase in the poverty population in a state increases the population of AFDC recipients by about 0.6 percent.

2. An increase in AFDC benefits by 1 percent of personal income, about $17 a month in 1992, would increase the dependent population in a state by about 2.2 percent given the number of the poor and by about 3.0 percent including the effect on poverty. Some part of this increase may be induced immigration from other states, so the proportionate effects of a uniform national increase may not be as high.

3. Economic conditions and the minority population affect welfare dependency only through their effects on the poverty rate. A 1 percent increase in average personal income reduces dependency by about 1.0 percent. A 1 percentage point increase in the minority population increases the dependent population by about 0.7 percent.

4. Given the number of the poor, welfare dependency declines with an increase in church membership. A 1 percentage point increase in church membership reduces the dependent population by about 0.7 percent. A 1 percentage point increase in the population completing high school reduces the dependent population by about 4 percent. And a 1 degree (Fahrenheit) increase in average temperature is associated with a 2.1 percent decrease in the dependent population. This temperature effect probably reflects the tighter welfare eligibility standards in the southern states.

5. Urbanization increases dependency. A 1 percentage point increase in the population residing in metropolitan areas increases the dependent population by about 1.3 percent.

Most of these findings are expected, in direction if not in magnitude. Welfare dependency is primarily determined by the level of welfare benefits and the conditions that affect the poverty rate. One important finding is that an increase in the minority population does not increase dependency except to the extent that it increases the poverty rate. In other words, poor minorities are no more likely to be dependent on welfare than are poor whites. One puzzling finding is the positive effect of urbanization on dependency, given that urbanization (as also shown on Table 1) has a negative effect on the poverty rate; the urban poor are apparently more likely to be dependent on welfare than are the rural poor.

Poverty

The patterns of poverty summarized by the last two columns of Table 1 support the following conclusions:

1. Poverty is also related to dependency. A 1 percent increase in the dependent population increases the poor population in a state by about 0.3 percent. An increase in AFDC benefits by 1 percent of average personal income increases the number of poor residents of a state by nearly 0.8 percent.

2. Poverty declines with an increase in average income and education. A 1 percent increase in average personal income reduces the poor population in a state by about 0.8 percent. A 1 percentage point increase in the population with high school or higher education reduces the poor population by about 1.8 percent.

3. The size of the poor population in a state is strongly related to the size of the black and Hispanic population. A 1 percentage point increase in the percent minority increases the poor population by about 1.1 percent.

4. For reasons that are not obvious, a 1 percentage point increase in church membership appears to increase poverty by about 0.4 percent.

Again, most of these findings are expected. Poverty is primarily determined by the level of AFDC benefits, general economic conditions, education, and the percent minority. The major puzzle is the positive effect of church membership on poverty, compared to its negative effect on welfare dependency.

Out-of-Wedlock Births

The patterns of out-of-wedlock births summarized by the first two columns on Table 2 support the following conclusions:

1. Out-of-wedlock births are strongly related to welfare dependency. A 1 percent increase in the welfare dependent population in a state increases the number of births to single mothers by about 0.5 percent.

2. Illegitimacy is also related to nonemployment. A 1 percent increase in the nonemployed population increases the births to single mothers by about 0.9 percent.

3. The level of welfare benefits, in turn, indirectly increases illegitimacy through the effects on the size of the dependent population and on the number of the nonemployed. An increase in AFDC benefits by 1 percent of average income increases the number of births to single mothers by about 2.1 percent.

4. Out-of-Wedlock births decline with an increase in church membership. A 1 percentage point increase in church membership reduces the number of illegitimate births by about 0.4 percent.

 5. In this case, the effects of the two large minority groups are very different. A 1 percentage point in crease in the black population increases the number of illegitimate births by about 2.3 percent. In contrast, a 1 percentage point increase in the Hispanic population reduces the number of illegitimate births by about 0.5 percent.

Nonemployment

The patterns of nonemployment summarized by the last two columns of Table 2 support the following conclusions:

1. Welfare dependency reduces employment. A 1 percent increase in the dependent population increases the number who are not employed by about 0.1 percent.

2. An increase in welfare benefits reduces employment by increasing the number of welfare dependents. An increase in AFDC benefits by 1 percent of average income increases the number who are not employed by about 0.5.

3. Education has a strong effect on employment. A 1 percentage point increase in the population with high school or higher education reduces the number who are not employed by about 2 percent.

4. Employment is also related to the relative size of the metropolitan and Hispanic populations. A 1 percentage point increase in the metropolitan population increases the number who are not employed by about 0.5 percent. A 1 percentage point increase in the Hispanic population reduces the number who are not employed by about 0.4 percent.

Abortion

The patterns of abortion summarized by the first two columns of Table 3 support the following conclusions:

1. Abortion is strongly related to nonemployment. A 1 percent increase in the adult population not working increases the number of abortions by about 1.7 percent.

2. An increase in AFDC benefits by 1 percent of average income would indirectly increase the number of abortions by about 1.2 percent by increasing the nonworking population.

3. A 1 percent increase in average income increases the number of abortions by about 1 percent.

4. Education, like income, contributes to abortion. A 1 percentage point increase in the population with educa- tion at the high school level or higher, for a given number of nonemployed, increases the number of abortions by about 3.6 percent.

5. The effects of the two large minority groups are somewhat different. A 1 percentage point increase in the black population increases the number of abortions by about 1.8 percent, whereas a 1 percentage point increase in the Hispanic population increases abortions by about 0.8 percent.

The pattern of abortions is not consistent with the usual patterns of the culture of poverty. The number of abortions increases with nonemployment and the percent minority, but it also increases with education and income. Maybe the most surprising finding is another blank space: the number of abortions appears to be independent of church membership.

Violent Crime

The patterns of violent crime summarized by the last two columns of Table 3 support the following conclusions:

1. The level of violent crime is strongly related to welfare dependency. A 1 percent increase in the welfare dependent population increases the violent crime rate by about 0.6 percent.

2. An increase in welfare benefits indirectly increases the violent crime rate by increasing the number of welfare dependents. An increase in AFDC benefits by 1 percent of average income increases the violent crime rate by about 1.1 percent.

3. The level of violent crime is also related to the composition of the population, reflecting both a direct effect and an indirect effect operating through the level of the welfare dependent popula- tion. A 1 percentage point increase in the metro- politan population increases the violent crime rate by about 1.3 percent. A 1 percentage point increase in the black population increases the violent crime rate by about 3.8 percent, and a 1 percentage point increase in the Hispanic population increases the violent crime rate by about 1.9 percent.

Patterns Across Pathologies

A comparison of the effects of the eight root causes across the six pathologies is useful to identify the conditions that most consistently contribute to or defend against these pathologies.

The level of AFDC benefits relative to the average personal income in each state is the one condition that increases each of the six focus pathologies, with effects ranging from a weak effect on nonemployment to disturbingly strong effects on welfare dependency and illegitimacy.

The next most consistent correlate of these conditions is the percent of the population that is black, a root cause for five of the pathologies other than nonemployment. The percent Hispanic contributes to welfare dependency, poverty, abortion, and violent crime but reduces illegitimacy and nonemployment.

The most consistent defenses against these pathologies are education and higher average income. A higher percent of the population with high school or higher education reduces dependency, poverty, and nonemployment. Higher average income reduces dependency, poverty, and violent crime. Both higher education and income, however, increase abortion.

The effects of the other conditions examined are more mixed. Church membership reduces welfare dependency and illegitimacy but appears to increase poverty. A higher percent of the population of a state that is resident in metropolitan areas increases dependency, nonemployment, and violent crime but reduces poverty. The contrary effects of church membership and the metropolitan population on welfare dependency and poverty are probably the most puzzling results of this study.

The Good Samaritan's Dilemma

For the most part, the political support for welfare reflects a generous motive to help those who are poor, single, and with children. Welfare would provoke little controversy and benefits would probably be higher if these conditions were substantially accidental or temporary--the result, for example, of the death, disability, or temporary unemployment of the major contributor to a family's income. That is why welfare was first promoted as a widow's allowance. That is why President Clinton supports welfare as a safety net but not as a way of life. The moral dilemma, of course, is that welfare, like the most forms of social insurance, increases the number of people with the insured condition. This study, for example, estimates that an increase in AFDC benefits per household by 1 percent of average income would increase the number of welfare dependents by about 3.0 percent and the number of births to single mothers by about 2.1 percent.

There is no obvious resolution of this age-old dilemma, and I claim no special moral insight. The patterns of pathology associated with the current welfare system, however, no longer seem acceptable, not so much because of their fiscal cost but because of their malign effects. The welfare legislation that Congress recently approved will give the state governments a greater incentive and opportunity to experiment with different approaches to welfare. The effects of this major welfare reform are difficult to predict, because state governments will have more flexibility to set benefit rates and eligibility conditions and there are many types of exemptions from the remaining federal mandates. The most important change is probably the substitution of lump-sum payments to the states for the current system of matching grants; this will increase the marginal cost to state taxpayers from the current 20 to 50 percent of AFDC benefits to 100 percent.

This study suggests that the state governments may be best advised to focus welfare on the innocent--widows, the genetically or accidently disabled, and children--and to set firm time limits on the welfare eligibility of others. Education and a strong general economic climate appear to be the most effective policy-responsive conditions to reduce the remaining pathologies. A blind compassion may be admirable but a knowledgeable compassion is twice blest.

References

Card, J. J. (1981) "Long-Term Consequences for Children of Teenage Parents." Demography 18 (137-56).

Hill, A. M., and O'Neill, J. (1994) "Family Endowments and the Achievement of Young People with Special Reference to the Underclass." Journal of Human Resources (Fall): 1090-91.

Niskanen, W. A. (1994) "Crime, Police, and Root Causes." Cato Policy Analysis No. 218. Washington, D.C.: Cato Institute.

Roosevelt, F.D. (1938) The Public Papers and Addresses of Franklin D. Roosevelt, Vol. 4, The Court Disapproves: 1935. New York: Random House.

Tanner, M.; Moore, S.; and Hartman, D. (1995) "The Work vs. Welfare Trade-Off." Cato Policy Analysis No. 240. Washington, D.C.: Cato Institute.

William A. Niskanen is Chairman of the Cato Institute and a former member of the Council of Economic Advisers.

 Cato Journal, Vol. 16, No. 1 (Spring/Summer 1996). Copyright © Cato Institute. All rights reserved.

The Cato Journal is published in the spring/summer, fall, and winter by the Cato Institute, 1000 Massachusetts Ave., NW, Washington, D.C. 20001-5403. The Views expressed by the authors of the articles are their own and are not attributable to the editor, editorial board, or the Cato Institute. Printed copies of the Cato Journal may be ordered by calling 1-800-767-1241. Back issues are also available on the Cato Institute Web site: http://www.cato.org. Email comments or suggestions to subscriptions@cato.org.