Trickling down the rising tide: new estimates of the link between poverty and the macroeconomy.

• Author: Freeman, Donald G.

1. Introduction

   Do rising tides lift all boats? Do income increases at the top of the distribution trickle down to those on the bottom? The insufficiency of economic growth alone to eliminate poverty has been a central tenet of public policy since the "war on poverty" officially began in the 1960s. Concern that the poor would remain trapped in a cycle of poverty motivated the passage of federal antipoverty programs designed to help the unfortunate move up the economic ladder.

   Despite these programs and a doubling of real per capita incomes in the almost 40 years since the original War on Poverty legislation, poverty rates remain stubbornly high. In 2001, the official poverty rate stood at 11.7% of the population, about two and one-half percentage points below its level of 10 years ago but six-tenths of a percentage point above its record low in 1973 (U.S. Census Bureau 2002a, c). And though the definition of who is poor and who is not is the subject of some controversy (see, e.g., Uchitelle 2001), there is little doubt that a large segment of the population lives on the margins of society, lacking secure housing, transportation, and medical services.

   Current research, such as Blank (2000), Haveman and Schwabish (2000), and Freeman (2001), emphasizes a set of stylized facts that characterize the relationship between the macroeconomy and poverty on a more or less decade-by-decade basis. Strong growth in the macroeconomy during the 1960s, together with an expansion in federal aid programs, accompanied a decline in poverty rates from 22.2% of the population in 1960 to 12.1% in 1969. In the 1970s, however, poverty rates were almost unchanged despite a 25% increase in real gross domestic product (GDP) per capita. Oil shocks, high inflation, and a demographically based increase in the natural rate of unemployment during the 1970s were seen as altering the usual poverty-growth nexus.

   In the 1980s, an increase of 22% in real GDP per capita was accompanied by an increase of 1.1 percentage points in the poverty rate. This anomalous juxtaposition of a growing economy and a growing poverty rate led to a series of papers--Cutler and Katz (1991) and Blank (1993) are representative--attempting to explain why, in the words of Danziger and Gottschalk (1995, p. 8), "poverty rates ha[d] become detached from economic growth." This research pointed to growing inequality in wages and benefits, both across and within industries, and changing demographics--including increases in single-parent households and in minority groups whose incomes traditionally lagged the national average--offsetting increases in mean income. Retrenchments in government spending programs for the poor were also cited (Hanratty and Blank 1992).

   In the 1990s, a different picture emerged, as prospects for low-skilled and less educated workers improved dramatically. The longest economic expansion on record reduced unemployment rates and lifted inflation-adjusted median incomes to levels not seen since the early 1960s and 1970s, respectively. Poverty rates began to fall again at last, from a cyclical high of 15.1% in 1993 to a low of 11.3 % in 2000.

   This paper reexamines the link between poverty and the macroeconomy using regional data over the time period 1969-1999 with special attention to the "lost" decade of the 1980s. The analysis here replicates previous studies using national data with pooled time series of census region-level data. Using pooled data allows controls for regional heterogeneity and for demographic changes in the composition of regional populations over the past 30 years. (1) In addition, the efficiency gains from the increase in degrees of freedom and the reduced multicollinearity from the additional variability in the regressors can provide a much greater level of precision in the estimates of the effects of the variables of interest on the poverty rate. (2)

   The principal findings of the paper are that the effect of the business cycle on poverty rates was significant throughout the 1980s and 1990s and strengthened successively over time, holding constant demographic characteristics and regional effects. These results provide a somewhat different view of the evolution of the poverty/macroeconomy nexus as described in the previous research of Blank and Blinder (1986), Cutler and Katz (1991), Blank (1993, 2000), and Haveman and Schwabish (2000), who rely primarily on national-level data to conclude that poverty rates were mostly unresponsive to business cycle conditions in the 1980s, only to return to responsiveness in the 1990s. This paper also presents evidence that increases in mean income have had smaller effects on poverty reduction in recent decades, though the change in the estimated coefficient over time is relatively small. (3)

   Section 2 describes the evolution of regional poverty data over the past several decades. Section 3 provides the regression analyses of poverty rates against regional economic and demographicvariables, and section 4 concludes.

2. Evolution of Regional Poverty Rates

   The origins of the War on Poverty have been laid to John F. Kennedy's campaign experiences in rural West Virginia (Bauer 1982; Danziger and Gottschaik 1995), culminating in the official declaration by Lyndon Johnson in his State of the Union Address in 1964. It was fitting that a president from a southern state should make the declaration, however, because historically the South has had the highest poverty rates of any region.

   The South's official poverty rate still leads the nation, although the gap has narrowed considerably since the 1960s, and when geographic cost of living differences are taken into account, the West may now have the highest poverty rate. (4) Table 1 presents summary statistics of the annual national and regional data on poverty and related variables used in the empirical analysis in section 3. Levels for 2000 are given for each variable for each region and for the national average, and series averages and standard deviations over the 1969-2000 time period are also provided. Regional maxima for each series are in bold type; regional minima are underscored. (5)

   The poverty rate as originally reported is the ratio of the number of individuals in the region with incomes below the national poverty threshold to the total regional population. The national poverty threshold was first established by Mollie Orshansky of the Social Security Administration in the 1960s as roughly three times the cost of an economy food plan, as food was estimated to be one-third of the after-tax expenditures of a poor family (Short 2001a). Since that time, the poverty level has been adjusted by changes in the consumer price index, so that it is always reported in current dollars. For example, a family of four with two children would be in poverty in 2001 if the family income fell below $17,900 (U.S. Census Bureau 2002b).

   However, Ruggles (1990), among others, points out that the use of the national poverty threshold to calculate regional poverty rates ignores cost-of-living differentials across regions and introduces a source of measurement error in the data. To address this and other concerns regarding poverty measurement, the Panel on Poverty and Family Assistance of the National Academy of Sciences (NAS) developed experimental poverty measures for the individual states, including those adjusted for geographic differentials in housing costs. These are reported in Citro and Michael (1995) and updated to 1999 by Short (2001a, pp. 12-13). (6) To calculate the time series of geographically adjusted poverty rates used in this paper, Short's indexes have been aggregated to the regional level, and extended to earlier years using the regional consumer price indexes maintained by the Bureau of Labor Statistics. The adjusted rates are also reported in Table 1, and both measures will be tested in the empirical section that follows. (7)

   The official U.S. poverty rate of 11.3% in 2000 comprises a range from 9.5% in the Midwest to 12.5% in the South. The geographic adjustment tends to raise poverty rates in the East and West, with their higher cost of housing, and lower rates in the middle of the country. Each region's poverty rate exhibited greater variation over time than the national rate, with relatively low correlation of poverty cycles across regions. For example, the bivariate correlation of the South and West poverty rates over the sample is 0.05.

   Of the independent variables, the unemployment rate (total and male) controls for the business cycle, the ratio of the official poverty level of income (the "poverty line") to the average income (both in current dollars) controls for trend income growth, and the inflation rate controls for instability and erosion of fixed incomes. These variables are common to previous research on poverty (Blank and Blinder 1986; Blank 1993, 2000; Haveman and Schwabish 2000), as are transfer payments.

   Transfer payments clearly have a direct role in poverty reduction: Scholz and Levine (2001) estimate that poverty rates would have been roughly twice the reported level were it not for the effects of the tax and transfer system. For example, the sharp decline in elderly poverty, from one-third of the over-65-year-old population in 1959 to one-tenth in 2000, is due largely to more generous Social Security payments (Burtless and Smeeding 2001). Other analysts, however, such as Ellwood and Summers (1986), note that some programs, like the former Aid to Families with Dependent Children (AFDC), may have created perverse incentives that perpetuated the cycle of poverty through behavioral effects. (8) Because the fixed effects regression utilized here is a within-region estimator, the estimated coefficient will report on how changes in transfer payments within a region have affected poverty rates over time.

   The Gini coefficient measures income inequality, with zero representing complete equality and 100 representing concentration...