The 2013 American Community Survey shows that the construction industry continues to rely heavily on immigrant labor with foreign born workers accounting for 23 percent of the U.S. construction labor force, and their share exceeding a quarter of the construction work force in some states.
The article examines ten years of the American Community Survey data, constructs and presents new estimates of the immigrant labor flow into the construction industry over the housing boom and bust years. The results show that the construction immigrant flow is highly variable exceeding 135,000 at the height of the housing boom in 2005 and plummeting to a low of 23,000 in 2011. The estimates highlight the role of new immigrant workers as an extremely flexible labor pool that adjusts quickly to the changing home building environment, and moves in sync with the single-family construction business cycle.
Data and Methodology
The immigrant labor flow research in this article is based on the American Community Survey (ACS) data. The ACS replaced the decennial Census long form and provides the same detailed data, including information about the country of origin, age, year of entry, industry and employment status of immigrants. As such, the data allow estimating the number of construction foreign born workers by a year of their arrival into the United States. The ACS, however, does not gather information on the legal visa status of immigrants and only differentiates between naturalized citizens and not citizens of the United States.
Looking at the latest 2013 ACS is not the most accurate way of capturing construction immigrant flow in earlier years, as foreign born workers who came to work in the construction industry during the housing boom years might have left the country or switched to a different industry by 2013.
A much better way of measuring the construction immigrant labor flow is by counting new foreign born workers as close to their date of entry to the United States as possible. For that reason, the ACS data from earlier years are expected to provide more accurate estimates of the immigrant flow in prior years. The ACS data from any given year, however, cannot deliver the comprehensive inflow estimates for the same year as the ACS interviews take place on an ongoing basis, January through December, rather than at the end of the year and, therefore, cannot capture the immigrant labor flow over the entire calendar year. This suggests that any particular ACS edition is best used to estimate the immigrant flow one year prior to the interviews. For example, the 2006 ACS should provide the timeliest estimate of the number of immigrant construction workers who arrived in 2005, and the 2005 ACS should provide the closest estimate of the 2004 immigrant labor inflow. Nevertheless, the estimates may not capture foreign born workers who spend less than a year working in the United States and, thus, may still undercount immigrant workers arriving to work in construction in any given year.
For this research project, NAHB Economics uses the 2004–2013 ACS Public Use Microdata Samples (PUMS) and estimates the flow of new immigrants into construction from 2003 to 2012. The presented estimates are gross inflows as they are not adjusted for any immigrant workers who might be retiring, leaving the country or construction industry. Rather, the estimated flows only capture new additional immigrants arriving to work in construction.
Annual Flow of New Immigrants into the Construction Labor Force
Figure 1 displays the estimated flow of new foreign born workers entering the US construction industry. The flow is highly variable and highlights immigrant workers’ role as an extremely flexible pool of labor. During the housing boom years, more than 100,000 foreign born workers were entering the United States to work in the construction industry on annual basis. In 2005, that number exceeded 135,000.
The immigrant labor flow responded remarkably quickly to the housing downturn, already in 2006 the number of new immigrants entering the US to work in the construction industry dropped below 100,000. The flow of incoming construction workers continued to shrink and reached a low of 23,000 in 2011. The trend reversed itself only recently, and 2012 became the first year since 2005 to register the rising immigrant construction worker inflow.
Immigration Relative to Home Building Activity
In addition to showing that the number of immigrants into the U.S. construction industry is highly variable, the annual pattern—peaking in 2005, plummeting to a low in 2010-2011, and recovering to a modest degree by 2012— suggests a relationship with economic activity in those years. As Figure 2 shows, single-family housing starts also went through a similar boom-bust-modest recovery cycle at about the same time.
A relatively simple way to study this more formally and see which series matches up most closely with construction immigration is to look at the correlations. Figure 3 shows the correlations between the annual flow of immigrants into the construction labor force and a number of economic time series. The first part of the table shows the correlations of construction immigrants with general economic indicators like GDP, mostly for the sake of comparison. The correlations in this section are either negative or not very large, suggesting that the construction immigrant flow was not moving in sync with the overall economy during the recent years.
Housing market statistics fare better. All are positively correlated with the annual flow of construction immigrants, and, with exception of Residential Fixed Investment (RFI) measured as percentage change, the contemporaneous correlations are all larger than .8. RFI is the new residential construction component of GDP (see Housing's Contribution to Gross Domestic Product). Following common practice for GDP and its components, RFI is measured in real (adjusted for inflation) terms. In addition to RFI, the second, shaded section of the table shows components of manufactured housingand housing starts, except for those specific to single-family construction.
The last column of the table shows the correlation of construction immigrants with each of the variables measured the previous year. Potential immigrants could respond to economic conditions in the U.S. with a lag, so that a pick-up in construction activity one year would attract immigrants to the industry primarily in the next. However, this seems not to be the case. Except for the change in RFI (for which the contemporaneous correlation is very small), construction immigration is more highly correlated with construction activity in the same year.
The lowest correlations in the second section of the table are the ones associated with multifamily construction. Correlations with the more general measures that include single-family construction are all well above 0.9 and seem to be driven strongly by the single-family component. For this reason, single-family statistics are collected and shown separately in the third section of the table.
The correlations between single-family activity and the number of workers immigrating into the construction industry in the same year are all remarkably high. The lowest is +0.97, and this is for sales of existing single- family homes. For new single-family permits, sales or starts, the correlations are all above +0.98. Although still relatively high, the correlations of construction immigrants with the prior year’s single-family activity are all below 0.90.
In short, the annual flow of immigrant workers into the construction sector is highly correlated with measures of new home construction, especially new single-family home construction. And the response of immigration is quite rapid, occurring in the same year as a change in single-family construction activity.
Immigrants per Single-family Start
Given high correlations, it is also interesting to explore long-run ratios relating a number of new incoming construction immigrants to a specific measure of construction activity. Since the immigrant labor inflow is most highly correlated with single-family home building, this section explores the relationship between the construction immigrant labor flow and single-family construction.
As an example, single-family starts are used in this analysis. Starts is the statistic commonly emphasized by the press when reporting on the New Residential Construction release from HUD and the Census Bureau, and the primary measure of residential construction typically forecast by NAHB and other industry experts.
Although the flow of construction immigrants is highly correlated with single-family starts, the ratio is not perfectly stable over the 2003-2012 time horizon captured by the ACS (Figure 4). As the figure shows, there were 68 construction immigrants per 1,000 single-family starts in 2003, but the ratio fluctuated from a high of 82 per thousand in 2004 to a low of 53 in 2011.
The ratio has been well below 68 per 1,000 single-family starts for the past three years pulling the long-run average down. But three most recent years could reflect atypical conditions—either in the market for housing or the environment for immigration—so NAHB prefers the longer term average of 68, which incorporates the most recent experience but averages in earlier data as well.
For forecasting purposes, relying on the long run average is bound to produce an inferior forecast, underestimating the immigrant flow during the housing expansion years and overestimating the number of immigrant arriving to work in construction during the past three years. Typically, regression analysis serves as a more reliable and accurate forecasting tool. The regression analysis shows that over the ten years covered by the ACS data, the construction industry was absorbing about 81 new immigrants for each additional 1,000 single-family starts during the housing expansion.  In the same way, during the housing contraction, the immigrant flow was reduced by 81 immigrants for every 1,000 lost single-family starts.
Figure 5 illustrates how closely the forecasts based on the long-run average and regression track the actual flow of construction immigrant workers. As expected, the immigrant flow forecast based on the regression analysis is more accurate but still underestimates the actual flow of immigrant workers during the housing boom years.
In summary, the analysis of ten years of the ACS data highlights the role of new immigrants as an extremely flexible pool of construction labor that adjusts quickly to the changing economic environment, especially the business cycle of single-family home building.
 A correlation is a statistic that ranges from +1.0 to -1.0, where +1.0 means two series are perfectly correlated—i.e., they move together step-by-step in the same direction at the same time. For example, the correlation of the outside temperature measured in °C and °F would be +1.0. The correlation between the temperature measured at two points a mile apart is likely to be positive and high, but a little less than 1.0 and so not quite perfect. The temperature outside and length of time your furnace is running are likely to be negatively correlated (correlation between zero and -1.0).
Manufactured housing is assembled entirely in a factory and shipped to a site where it is then attached to a permanent foundation. It is regulated by HUD, excluded from the new residential construction statistics on conventional single-family and multifamily housing released by HUD and the Census Bureau, and distinct from modular housing. Modular housing is shipped to a site in sections, is included in the standard HUD/Census residential construction statistics; and, like conventional single-family and multifamily construction, is regulated by locally adopted and enforced building codes.
The estimated regression equation is Immigrant Flow = -7,827 + 81.34 x SF Starts(000s).