Why Poverty Persists

"Changes in family structure - notably a doubling of the percent of families headed by a single woman - can account for a 3.7 percentage point increase in poverty rates, more than the entire rise in the poverty rate from 10.7 percent to 12.8 percent since 1980."

Over the past 45 years, the United States has experienced an ever-growing standard of living, with real GDP per capita more than doubling between 1959 and 2004. In contrast, living standards among some populations in the United States seem to have stagnated. Between 1970 and 2003 the non-elderly poverty rate rose from 10.7 to 12.8 percent. This is in spite of dramatic increases in female labor force participation and overall education levels, and an almost 50 percent increase in cash and in-kind welfare spending per capita. All of these factors should have put substantial downward pressure on poverty rates in the United States, yet they have remained relatively stable. In Poverty in America: Trends and Explanations (NBER Working Paper No. 11681), co-authors Hilary Hoynes, Marianne Page, and Ann Stevens seek to understand why this is the case. They examine post-war trends in American poverty, the work habits and family structures of the non-elderly poor, an d the likely effects of immigration, and they attempt to estimate the effects of the various government programs designed to alleviate poverty.

The authors first review some basic facts about the nature of poverty in the United States: according to the March Current Population Surveys, poverty rates are generally higher among children than among adults. In 2003, children were approximately 29 percent of the non-elderly population but they constituted 40 percent of the non-elderly poor; 17.6 percent of all children lived in households with incomes below the poverty line. Overall, only 7 percent of those living in households headed by a married individual were poor, whereas households with an unmarried head and children present -- 83 percent of which were headed by women -- had poverty rates of 40.3 percent. Likewise, the probability of being poor varies tremendously by race: blacks and Hispanics are much more likely to be poor than whites, even though most of the poor are white.

The persistence of poverty also depends strongly on individual and family characteristics. Among those beginning a spell of poverty, about 83 percent of white children living in two-parent households headed by someone with at least a high school education will escape long-term poverty. In contrast, only 10 percent of poor black children in a household headed by a single woman without a high school diploma will avoid it.

To explore the determinants of trends in poverty, the authors use data on state poverty rates over the period 1967-2003. Possible explanations for changes in poverty include: changes in labor market opportunities, female labor force participation, family structure, and government assistance for the poor, and immigration. Hoynes and her co-authors show that labor market opportunities are the major determinant of poverty.

Specifically, they find that the unemployment rate, median wages, and wage inequality in the lower half of the wage distribution all are significant determinants of poverty rates. Overall, increasing the unemployment rate by 1 percentage point increases the poverty rate by 0.4 to 0.7 percentage points. Increasing the median wage by 10 percent decreases the poverty rate by about 2 percentage points. Increasing the ratio of the median wage to the average weekly wage in the 20 percentile of the wage distribution (a measure of inequality) by 10 percent increases the poverty rate by roughly 2.5 percentage points.

The strength of the relationship between these business cycle and labor market indicators and the poverty rate has declined in the past two decades, though. After 1980, the effects of unemployment, median wages, and wage inequality were about half their pre-1980 magnitudes, the authors estimate. Predicted poverty rates based on coefficients estimated with data from the entire period (1967 through 2003) are significantly higher than the actual poverty rate.

In contrast, actual poverty rates are very close to the predictions for the post-1980 period. This close correspondence between the actual poverty rate after 1980 and the poverty rates predicted by unemployment, median wages, and wage inequality in part solves the mystery of why poverty rates have not declined by more. The "answer" is familiar to those acquainted with trends in inequality over this period: poverty has not fallen despite robust economic growth because this growth did not result in rising wages at the median and below.

Missing from this analysis of labor market opportunities and poverty, though, is the dramatic increase in female labor force participation over this period (a rise from 57 percent in 1970 to 76 percent in 2000). Once the authors incorporate female labor supply into their poverty rate models, the puzzle returns. Specifically, after 1980 actual poverty rates are substantially higher than predicted poverty rates.

The period after 1980 saw large changes in family structure -- notably a doubling of the percent of families headed by a single woman. Because poverty rates among female-headed families are typically 3 or 4 times the level in the overall population, such changes in the distribution of family types can have potentially large effects on poverty. The authors find that these changes in family structure can account for a 3.7 percentage point increase in poverty rates, more than the entire rise in the poverty rate, from 10.7 percent to 12.8 percent since 1980.

Using Census data for 1960-2000, the authors find that the increase in the U.S. immigrant population has had only a marginal effect on poverty. Even though recent immigrants are "poorer than their predecessors, their fraction of the population is simply too small to effect the overall poverty rate by much." These results do not, however, take account of the possibility that a rising immigrant population could directly affect the wage opportunities of natives.

Finally, the authors consider the effects of welfare spending on poverty, using four measures of welfare generosity. Overall, their results consistently show that increases in welfare spending have produced only modest reductions in poverty, and that their effect has become more modest over time. This result is partially driven by the nature of the official poverty definition, specifically the fact that increments to aftertax income (as resulting, for example, from the significant expansions in the Earned Income Tax Credit) or provision of in-kind benefits will not be reflected in poverty rates based on pretax cash income definition. Furthermore, the lack of an effect on official poverty does not mean that these programs have not significantly improved the well being of the poor.

Taken together, the results suggest that the lack of improvement in the poverty rate reflects a weakened relationship between poverty and the macroeconomy. The lack of progress despite rising living conditions is attributable to the stagnant growth in median wages and to increasing inequality. Holding all else equal, changes in female labor supply should have reduced poverty, but an increase in the rate of female-headed families may have worked in the opposite direction. Other factors often cited as having important effects on the poverty rate do not appear to play an important role -- these include changes in the number and composition of immigrants and changes in the generosity of anti-poverty programs. Future work should focus on understanding why the poverty rate's responsiveness to macroeconomic indicators has changed over time.

-- Linda Gorman