Immigration and poverty
Disappointing income growth in the 1990s not solely the result of growing immigrant population
by Jeff Chapman and Jared Bernstein
Recently released data from the 2000 Census show that the nation’s poverty rate fell less than one percentage point between 1989 and 1999, dropping from 13.1% to 12.4%.1 In some states, including California and New York, the poverty rate was higher in 1999 than in 1989. In addition, some areas of the country failed to see the increases in real median family income that were hoped for given the strong economy of the latter 1990s. For example, Census data reveal that median family income in New York grew only $113 (0.2%) in real terms over the decade. These results are disappointing for those who expected more from the strongest economy in decades.
Media coverage has downplayed these disappointing findings as being largely the result of a growing immigrant population.2 The reasoning behind this claim is that income failed to grow as much as expected due to the addition of more low-income families to the population through immigration. The implication is that the United States’ lack of economic progress in regard to poverty or median income growth should not be of great concern-it wasn’t the strong economy that failed to lift incomes, but rather the increase in low-income families from abroad that makes it appear like no progress has been made in these areas.
This intuition is by no means without merit. It is true that the immigrant share of the population increased over the decade, and that immigrants’ incomes are, on average, lower than natives’. It is also the case that the increase in the immigrant share of the population can and has put downward pressure on overall income growth over time, a phenomenon we refer to as the “share effect.”
But the existence of the share effect by no means justifies citing immigration as the sole, or even the most important factor behind the less-than-favorable Census results. Without much more evidence, it is a mistake to simply conclude that, since immigration expanded and immigrants have lower incomes, the lack of progress must be due to immigration.
The needed evidence is at least twofold. First, the magnitude of the share effect must be quantified, that is, how much did the increase in the share of the immigrant population lower real incomes or raise poverty rates? Second, the impact of the share effect can be offset by trends in immigrants’ own income and poverty status. The growth of immigrant incomes will offset the share effect, and analysts need to quantify this effect, as well (we refer to this as the “income effect”).
In a period like the 1990s, when both immigrants’ population share and their incomes rose, the question of immigration’s impact can be viewed as the outcome of a race between the share and income effects. That is, did immigrants’ income improve fast enough to offset the downward pressure exerted by their increased share in the population? Without quantifying these two countervailing effects, researchers have little useful authoritative information to bring to the discussion. In fact, as we show below, over the 1994-2000 period, immigrants’ rising incomes offset the negative impact of their rising shares.
At this point, the 2000 Census microdata have not yet been released, so we do not have all the data we need to fully explore this issue. The hope is that, given the data available, we can introduce a note of caution into any interpretation of the Census results that heavily depends on increased immigration.
In this analysis we look at the nation as a whole and also specifically at New York and California-two states where one may expect immigration to play a larger factor. In fact, over one-in-four New Yorkers and one-in-three Californians are immigrants (see the data appendix for details on our more comprehensive definition of immigrants). It is also important to consider these two states because their poverty rates were higher in 1999 than in 1989, according to Census 2000 data.
An analysis of the currently available data shows:
- Over the 1994-2000 period, poverty rates fell much more quickly for immigrants than for natives. For example, the national poverty rates of recent immigrants (those here for 10 years or less) fell about four times as fast as for natives (11.6 percentage points for immigrants versus 2.9 points for natives); the rate for all immigrants fell 2.7 times as fast as that of U.S. natives.
- Immigrant families also experienced greater increases than U.S. natives in real median family incomes from 1994 to 2000. After adjusting for inflation, the median family incomes of immigrants rose 26.3% from 1994 to 2000, while the median family incomes of native U.S. families grew half that fast. For recent immigrants, the growth in real median family income was even larger at 40.5%.
- These gains in immigrant income over the 1994-2000 period were substantial enough to offset the negative impact of the share effect.
- In a preliminary analysis of the full Census period between 1989 and 1999 for California and New York (two states with large immigrant populations) the increase in immigration added about a percentage point to the growth in poverty over the decade. But absent this effect, poverty would have been unchanged in California and would have risen slightly in New York (still a disappointing result for this period).
- Immigration’s role has been exaggerated and has crowded out other, more fundamentally economic factors, such as inequality and unemployment, from the discussion. These factors hurt the economic prospects of all low-wage workers, regardless of nativity.
Poverty rates and median family income, by nativity and entry
It is true that immigrants are much more likely to live in poverty than are native U.S. citizens (Table 1). This is especially true of recent immigrants, whose poverty rate is over twice that of U.S. natives. Because of this, at any point in time, the poverty rate would, indeed, be lower in the absence of immigration. Also, increasing the immigrant share will raise the poverty rate. However, as noted, we need to quantify both this share effect as well as the offsetting income effect that occurred over this period (the impact of faster income growth among immigrants).
As shown in Figure 1 and Table 1, the national poverty rates of recent immigrants fell about four times as fast as it did for U.S. natives; the rate for all immigrants fell 2.7 times as fast as that of U.S. natives. Poverty rates of immigrants living in New York and California also fell further during the boom than did the poverty rates of U.S. natives (Table 2). The poverty rates of recent immigrants fell 13.3 percentage points from 1994 to 2000 in New York and 12.5 percentage points in California, while those of U.S. natives fell 2.4 points in New York and 3.0 points in California.
“Figure 1″ border=”0″ height=”361″ width=”600” />
Immigrant families also experienced greater increases in real median family incomes during the same period (Table 3). After adjusting for inflation, the median family incomes of immigrants rose 26.3% from 1994 to 2000, while the median family incomes of native U.S. families grew half that fast. For recent immigrants, the growth in real median family income was even larger: 40.5%, an increase of over $10,000.
Since immigrants’ income growth outpaced that of U.S. natives, we need to measure the extent to which this income effect offsets the share effect before we discount the disappointing Census results.
The impact of the share and income effects
The share effect will largely be driven by the magnitude of the increase in the immigrant share of the population. Nationally, this share grew by 2.6 percentage points between 1994 and 2000. The share of the population that is recent immigrants grew less than one percentage point during the same period.3 In New York, the immigrant share of the population grew by 3.6 percentage points, while in California it only grew by 1.2 points.
We are able to use a simple shift-share technique to decompose the change in the overall poverty rate, assigning separable contributions to the impact of changes in the population shares of immigrants and natives (holding the poverty rate constant) and changes in their poverty rates (holding the population shares constant).4 The first row, second column of Table 4 (the share effect) shows that, as expected, the increase in the share of immigrants raised poverty in each case, though in no case by as much as a percentage point. For the case of recent immigrants-the focus of much discussion around the Census results-the increase in poverty due to their larger national share was only 0.2 percentage points.
But the decline in immigrant poverty rates (the income effect), as shown in Figure 1, more than offset this factor, and on net, immigration lowered poverty for each group. Take, for example, the case of California. While the share effect added 0.3 percentage points to the poverty rate, the income effect-the fall in immigrant poverty in California-contributed 3.2 percentage points to poverty’s decline. On net, the impact of immigration on California poverty was to lower the state’s rate by 2.9 percentage points. For New York, the result is less dramatic because, while the immigrant rate fell steeply (see Table 1), the share grew more quickly than in California and thus added just under a point to the change in poverty between 1994 and 2000. Here too, however, the poverty-reducing impact of the income effect more than offsets the share effect.
Median incomes do not allow the same type of decomposition as poverty rates. So, in order to gauge the share and income effects, we apply a technique that is similar in spirit to the poverty shift-share analysis (see Table 5).5 If the national immigrant population had remained at its 1994 population share in 2000, then real median family income would have been only 0.6% higher. In both New York and California, the share effect lowered income growth by 1.6%. While we cannot isolate the income effect here, as we could with the poverty rates, the large growth in immigrant income (Table 1) likely offset share effects of this magnitude.
The 1989-99 period: preliminary analysis
As noted, the Census data needed to perform an analysis on the full 1990s business cycle are not yet available. To gain some preliminary insight into what these results are likely to show, we examine the poverty rates and population share of immigrants and U.S. natives in two states, New York and California, in 1989 using the 1990 Census data, and in 1999 using the March CPS.
By crossing data sets in this manner, we are surely introducing some error into the analysis. For example, the 1999 CPS poverty rates for New York and California are 14.1% and 13.8%, while the published Census rates are 14.6% and 14.2%, respectively. However, we suspect that these errors are of a relatively small order of magnitude; while the numbers would surely be a bit different were we able to use Census microdata, the substance of the results would likely be unchanged. However, Census and CPS estimates of median family incomes are quite different, so we focus solely on analysis of poverty rates.
Table 6 provides poverty rates in the two periods, along with a shift-share analysis like the one in the previous table. In our data, California poverty goes up 1.4 percentage points, or from 12.4% to 13.8%, between 1989 and 1999. Poverty rates are essentially unchanged for immigrants in California from 1989 to 1999 and are slightly higher for U.S. natives (1.1 percentage points). But the immigrant share (not shown) rose by 6.2 percentage points, so the question is again, how quantitatively meaningful are these shifts in determining California poverty rates over the period?6
The shift share shows that, holding poverty rates constant, the increase in the immigrant share of the population added 1.3 percentage points to California poverty over this period. In other words, the strong economy of the 1990s (the impact of which was concentrated in the second half of the decade), failed to reduce California poverty, even after we extract the impact of a larger immigration share of the population.
The New York data tell a similar story. Poverty rose 1.3 percentage points, with native poverty up 1.5 points and immigrant poverty down slightly. The immigrant share grew by 4.5 percentage points, which, holding poverty rates constant, added 0.9 points to the growth in poverty (the decline in immigrant poverty reduced the overall growth slightly, by 0.2 points). Thus, even in the absence of a larger New York immigrant share, poverty rates in that state would have increased from 1989 to 1999.
Given that we are shifting between the two data sets, the 1989-99 results are less reliable than the 1994-2000 CPS results, but they do have the advantage of covering the full business cycle. The 1989-99 analysis shows that the conventional wisdom regarding immigrants’ contribution to poverty has some merit in that the increased share of immigrants did place upward pressure on poverty rates in these two states.
But the results also show that immigration is by no means the whole story in understanding poverty trends over the 1990s, as some news stories and commentators have implied. Based on a simple shift-share analysis, once we extract the impact of the growth of immigration, poverty would have been unchanged over the decade in California and slightly higher in New York. Considering that the 1990s were widely hailed as one of the greatest economic periods in decades, this is an unsettling result. It implies that, as the Census data are released and scrutinized, researchers cannot simply cite the increase in immigration and leave it at that. We need to understand what other factors were responsible for the lack of progress in the fight against poverty over the 1990s.
While the boom of the latter 1990s did lift low incomes, Census data reveal that economic progr
ess was not as great as might have been expected, particularly in certain states. With very little analysis, however, numerous commentators have been misinterpreting these results as the effect of increased immigration. While no analysis could completely account for the effects of immigration (both positive and negative), our analysis reveals that poverty rates would have been only slightly lower and median income only slightly higher between 1994 and 2000 if immigration rates had remained constant.
Our preliminary analysis of the 1989-99 period yields a similar conclusion. Though data limitations suggest we must view these results with caution, we still find that, had immigration not increased between 1989 and 1999, poverty rates in California would not have fallen and in New York would have still increased slightly.
None of this should be taken to imply that immigration plays no role in the economic trends of the 1990s. But, thus far, immigration’s role has been exaggerated and has crowded out other, more fundamentally economic factors from the discussion. Both New York and California, for example, saw larger than average increases in inequality over the decade, and the incomes of the wealthy pulled far ahead of those at the middle and the bottom of the income scale.7 In many states, the increase in inequality meant that the growth that did occur went disproportionately to those at the top of the income scale, leaving those at the lower end, regardless of their nativity, more vulnerable to poverty.
Similarly, the 1990s economic boom arrived later in both New York and California. For example, unemployment in New York City was 8% in 1998, compared to 4.5% for the nation. The fact that unemployment remained high in New York City meant that all less-advantaged workers, not just immigrants, faced a slack labor market. Any defensible accounting of the trends in income and poverty over the 1990s needs to include at least these explanations, and probably others as well.
As noted, most of this analysis runs from 1994 to 2000, since these are the years for which data exist to examine changes in native and immigrant income trends, and their population shares. The widely cited Census data, however, provide comparisons between 1989 and 1999. Since the Census 2000 microdata are not yet available, we cannot fully analyze these years, although we do offer some analysis of them, comparing Census 1990 data (with poverty data for 1989) to March 2000 Current Population Survey (CPS) data for the year 1999.
The eventual release of the Census microdata will allow us to analyze these trends from one business cycle peak (1989) to the next (1999, although 2000 was the actual peak). The Census microdata are also consistent over the two years and have large sample sizes. The Census-to-CPS comparison used in this analysis, while meeting the “peak-to-peak” criterion, introduces some inconsistencies (discussed below), because we are analyzing data from two different data sets.
Still, there are numerous advantages to the CPS data. Most importantly, the CPS allows us to calculate income and poverty status for natives and immigrants, 1994-2000. While these years do not cover the entire cycle, they do cover the boom years. If the share effect truly dampened progress against poverty or lowered income growth, these data should reveal it as effectively as the Census data. Also, since the main objective is to compare immigrants to U.S. natives (and to measure the extent to which increased immigration kept poverty from falling further), we are somewhat less concerned with going peak-to-peak as we are with comparing the two groups over the same years. Presumably, both groups were affected by the growing macro-economy over this period, which provides some control for the cycle.
In this analysis we look at the nation as a whole and also specifically at New York and California-two states where one may expect immigration to play a larger factor. In fact, over one-in-four New Yorkers and one-in-three Californians are immigrants (as defined here). It is also important to consider these two states because their poverty rates were higher in 1999 than in 1989, according to Census 2000 data.
The Bureau of Labor Statistics and the Census Bureau define the “foreign-born” population as those persons born abroad to parents who aren’t U.S. citizens. To this group, we add persons born in Puerto Rico and other U.S. territories since they share many of the economic characteristics of the foreign-born. Children born within the United States are U.S. citizens and are not included in the Census statistics on the foreign-born. However, since the income level and poverty status of children depend on their parents, we define children living with only immigrant parents as immigrants. Since both Puerto Ricans and the citizen children of immigrants have higher poverty rates than the Census foreign-born, adding them to our definition should increase our estimates of the impact of immigration. For the purposes of our analysis, “recent immigrants” are those who entered the United States within the last 10 years.
The authors thank the Foundation for Child Development, The Joyce Foundation, the John D. and Catherine T. MacArthur Foundation, the Charles Stewart Mott Foundation, and the Rockefeller Foundation for support of our living standards work. We also benefitted from comments by Steve Camarota, Deborah Reed, Cordelia Reimers, and Larry Mishel.
1. Since the poverty rate tends to rise during periods of recession and fall during periods of expansion, it is desirable to compare poverty rates at similar points in the business cycle. Fortunately, the Census results are from one peak (1989) and a near-peak (1999-the 1990s recovery went through 2000). The official source for year-to-year estimates of poverty and income is the March Current Population Survey, our main data source in this paper. According to the CPS, the U.S. poverty rate grew from 12.8% in 1989 to 15.1% in 1993, then fell to 11.8% by 1999.
2. For example, see Scott, Janny. 2002. “Census Finds Immigrants Lower City’s Income.” New York Times, August 6. Scott, Janny. 2002. “Census Finds Rising Tides, Many Who Missed Boat.” New York Times, June 17. Washington Post. “’90s Boom Had Broad Impact; 2000 Census Cities Income Growth Among Poor, Upper Middle Class.” June 5.
3. Data from Census 2000 support these findings. According to Census data, the increase in the share of the national population that were foreign-born increased 3.2 points from 1990-2000 and the share of the population that were recent entrants increased 1.2 points. This does not include persons born in U.S. territories or the citizen children of immigrants.
4. The first component mentioned is the change in population shares for each group times the average poverty rate across the two periods (1994 and 2000). The second component is the change in the poverty rates times the average population share. The sum of these components equals the change in the overall poverty rate. Note that this technique only measures the share and income effects as described in the text. There is a large literature evaluating the impact of the presence of immigrants on native incomes, employment, and wages which goes well beyond this simple shift share analysis.
5. Our approach is to adjust the sample weights in the final year so that the immigrant share of the population is the same as it was in the base year, and to recalculate median income in the final year using these adjusted weights. Because of the share effect, this will result in a higher value of me
dian income than the actual level. The difference between the simulated and actual median represents the impact of the increased share of immigrants on income growth between the base and final year.
6. The 2000 March CPS weights will be adjusted to reflect data collected from Census 2000. However, comparing Census 2000 counts of the foreign-born population with 2000 March CPS counts suggests that the CPS undercounted naturalized citizens and overcounted non-citizens. Because naturalized citizens have a lower poverty rate than non-citizens, this adjustment should actually lower the immigrant poverty rate, decreasing estimates of the impact of immigration on poverty and income.
7. Bernstein, Jared, Heather Boushey, Elizabeth McNichol, and Robert Zahradnik. 2002. Pulling Apart: A State-by-State Analysis of Income Trends. Washington, D.C.: Economic Policy Institute and Center on Budget and Policy Priorities.