Category Archives: Regional growth and development

Trends in Education and Income in Chicago

Harvard economist Edward Glaeser shows that education is one of the strongest predictors of urban economic growth.   This is particularly the case for older cities like Chicago.  One of the reasons for this is that a higher density of college-educated workers is associated with higher levels of worker productivity.

There is very good news for Chicago.  Recent data for 2016 from the United States Census Bureau’s American Community Survey shows that the city of Chicago now has the highest percentage of college graduates of the seven largest cities in the United States (Table 1).  Almost 2 out of 5 adults twenty-five and older in Chicago have at least a bachelor’s degree. Chicago beats New York City, Los Angeles, San Antonio, Houston, Phoenix, and Philadelphia.  Of the ten largest cities, only San Diego and San Jose have higher levels of educational attainment as measured by the percentage of adults with at least a bachelor’s degree.

If the sample is limited to non-Hispanic whites, Chicago even beats San Diego and San Jose, the home of Silicon Valley.  For this population, over 3 in 5 have a college degree in Chicago.  In some community areas in Chicago like Lincoln Park, Lakeview, and the Loop, about 4 out of 5 have a college degree.

One of the reasons for Chicago’s success in this arena is that the city of Chicago is an attractive place to live and work for college graduates, especially young grads.  Over half of the young college graduates in the Chicago metropolitan area live in the city of Chicago.  This is up from about forty percent in 1990.

Another reason is that migrants to Chicago are more likely to have a college degree.   Last year about 3 of 4 migrants to Chicago from other states and from abroad had a college degree.  Ten years ago only about 1 in 2 migrants to Chicago had a college degree.  It is particularly noteworthy that in 2016 seventy-three percent of foreign migrants to Chicago had a college degree.

If one goes back in time, Census data indicates that adults in the city of Chicago were significantly less educated than their suburban counterparts.  This is no longer the case.  The percentage with a college degree in Chicago is higher now than in the suburbs of the Chicago metropolitan area although some suburbs have higher levels of attainment (Table 2).  For example, 2 out of 3 residents of Evanston have at least a bachelor’s degree.

Although non-Hispanic whites who account for about one-third of the population of Chicago are doing well, the situation for African-Americans and Hispanics is more mixed.  Of the ten largest cities in the United States, African-Americans in Chicago rank seventh and Hispanics rank ninth in the percentage of adults with a college education.

The good news is that both the percentage and number of college-educated African-Americans and Hispanics in the city of Chicago has increased since 2010.  This is also the case for non-Hispanic whites and Asians. In Table 3, data are arrayed on the number and percentage of college graduates in the city of Chicago by race and ethnicity.  The data indicate that the largest gain in the number of college graduates was for non-Hispanic whites followed by Hispanics.  The largest relative gain was for Hispanics (over fifty percent) followed by Asians.  Overall, there was a 20% increase in the number of college graduates in the city of Chicago between 2010 and 2016.   Part of the increase is a result of growth in the number of non-Hispanic whites, Hispanics, and Asians in the city.  For the African-American population, growth in the number of college graduates cannot be attributed to population growth because the number of blacks in the city declined by ten percent during the 2010-2016 period.

In suburban areas, the story is different, as shown in Table 4.  For example, in suburban Cook County non-Hispanic whites are over twenty percentage points less likely to have a college degree than their counterparts in the city of Chicago.  Further, there has only been a one percent increase in the number of non-Hispanic whites with a college degree in the suburbs of Cook County. This is partly a result of a seven percent decline in the non-Hispanic white populations in suburban Cook County.  At the same time, there have been large increases in the number of African-American, Hispanic, and Asian college grads in suburban parts of Cook County.  For example, for the 2010-2016 period, the African-American population in suburban Cook County increased twelve percent.

Further, Chicago public schools that disproportionately serve African-American and Hispanic families have improved considerably. Over the past ten years, CPS high school graduation rates have increased from fifty-seven percent to seventy-four percent. Of high school graduates, a higher percentage are going to college. Test scores are up as well. Low-income students in Chicago outperform other low-income students in other districts in Illinois.  A Stanford study argues that CPS is the fastest-improving school district in the country.

It is worth noting that in the 1980s Secretary of Education Bennett called the Chicago public school system “the worst in the nation.”  Although it is not clear if that was ever the case, it certainly is not now.

Although there is good news on education, the evidence on income is more mixed.  This is partly a product of national trends over the past couple of decades.  For the city overall, median real household income increased three percent last year.  Since 2010, real income has increased close to eight percent  although there is substantial variation by race and ethnicity.  Non-Hispanic white income increased seven percent while Hispanic income increased almost ten percent.  Although household income in the African-American community increased this past year, it is almost five percent lower in real terms since 2010.  Over a longer period of time (since 1979) African-American income has declined even more (twenty-one percent) although non-Hispanic white income is up substantially (forty-three percent).

In the suburbs of Chicago, household income has also increased modestly (about three percent) this past year.  Over a longer period of time, non-Hispanic white income in the suburbs is about where it was at in 1979.  However, median household income for African-Americans in the suburbs has remained constant since about 1979.

Although education has increased in the city of Chicago relative to suburban Chicago, median household income in suburban areas is still significant higher than in the city of Chicago.  In 2016, household income in the city was eighty percent of median household income in the metropolitan area.  This is up two percentage points since 2010.

These changes have driven other important changes in Chicago.  On the positive side, the high concentration of talent in parts of the city is resulting in high skilled jobs following the talent, especially to areas in and around the central business district.  Further, more educated and affluent African-Americans and Hispanics have been able to move to suburban locations for better opportunities for their families.

On the negative side, the large and continued decline in income in the African-American community in many parts of the city of Chicago is cause for concern.  This is resulting in well over half of the children in community areas like Englewood and West Garfield Park growing up in poverty.  It is also resulting in large population declines in communities like Englewood and West Englewood.


Updated forecasts of Seventh District GSP growth

Updated forecasts of Seventh District GSP growth

Each year, the Chicago Fed provides estimates of annual gross state product (GSP) growth for the five states in the Seventh Federal Reserve District.[1] While presenting last year’s projections, we proposed a new forecasting model incorporating the U.S. Bureau of Economic Analysis’s (BEA) quarterly GSP data. In this post, we again use our quarterly model to generate GSP growth forecasts for 2016 (whose actual values will become available next month) and compare our estimates to those produced by our previous (annual) model described in Brave and Wang.[2]

GSP growth and the MEI

The BEA releases annual GSP data for the prior year each May. However, in an effort to provide a more frequent reading on regional economic growth, the BEA has also been releasing preliminary quarterly estimates of GSP since 2014. While the lag time in the production of these estimates is not as substantial as it is for annual GSP, it can still be rather long. For example, through April 2017, the 2016:Q4 GSP data have not yet been released, and will not be released until May 11, 2017.

The Chicago Fed’s Midwest Economy Index (MEI) provides an even higher frequency reading on economic growth for the five states of the Seventh District.[3] A weighted average of 129 state and regional indicators measuring growth in nonfarm business activity, the monthly MEI (like GSP) is a broad-based measure of economic conditions. After aggregating its monthly values to obtain a quarterly reading, the MEI also correlates quite well historically with Seventh District GSP growth.[4] Figure 1 illustrates this relationship, featuring a sample correlation coefficient of 0.57. [5] For this reason, the MEI forms the basis of our forecasting model for Seventh District GSP growth described in the next section.figure-1

2016:Q4 projections

To project the 2016:Q4 annualized GSP growth rate for each of the Seventh District states, we use the following linear regression model:


This model relates each state’s current quarter GSP growth rate to national (the current quarter’s annualized growth rate of national gross domestic product, or GDP), regional (the current quarter’s MEI and relative MEI, or RMEI, index values), and state factors (the current quarter’s annualized growth rate of real personal income, or PI, and the previous quarter’s annualized GSP growth rate).

National and regional economic conditions play an important role in capturing state-level growth for all five Seventh District states. However, some states respond differently to regional economic conditions depending on the health of the national economy. The relative MEI reflects Midwest economic conditions relative to those of the nation, such that its inclusion in our model is designed to capture the interaction of national and regional factors on state-level growth. The inclusion of lagged GSP and state PI is then intended to capture state-specific factors that our regional and national indicators fail to. This is particularly important for states such as Iowa, where a larger share of economic activity is in agriculture, as the MEI only considers nonfarm business activity.

The extent to which each of these factors contributes to explaining GSP growth in our forecasting model varies across the five states. National growth seems to be the most important factor for Iowa, Illinois, and Wisconsin; but state and regional factors dominate for Indiana and Michigan, respectively.


Figure 2 shows our projections for 2016:Q4 for the five Seventh District states and the Seventh District states combined.[6] All five states featured fairly strong growth readings for 2016:Q3. As a result, the model predicts some mean reversion in Q4 for each state. Compared with the other four states, Iowa has a somewhat weaker GSP growth projection of 1.0%. This is explained in part by Iowa’s negative PI growth in Q4, whereas the other Seventh District states featured positive PI growth. On the other end of the spectrum, Indiana has a somewhat stronger projection of 2.3%.

2016 projections

Based on the quarterly GSP data for the first three quarters of 2016 and our projection for the fourth quarter, we can also project 2016 annual GSP growth for the Seventh District states. These estimates are shown in the first column of table 1. Our projection of 1.6% for annual GSP growth for the Seventh District states combined is identical to the national GDP growth rate. There is variation, however, in our annual forecasts across individual states, which can be largely explained by differences in observed growth in the first three quarters of the year (see figure 2). Michigan’s annual forecast is consistent with its strong growth readings in Q2 and Q3. Conversely, Iowa’s large negative growth rate in Q1 led to a negative annual growth rate forecast. Finally, Illinois’s and Indiana’s similar growth rates throughout the year yield similar annual forecasts; and Wisconsin’s weak first quarter weighs down its annual forecast despite relatively strong growth in Q2 and Q3.


We also present in the second column of table 1 projections from the annual GSP growth forecasting model described here. These projections are quite similar across the Seventh District states; but as we saw last year, the annual model tends to forecast higher growth than predicted by our quarterly model (the only exception being Michigan). In general, we believe the added information provided by considering GSP quarterly data should make our forecasts from the quarterly model more accurate than those from the annual model. However, some caution should be exercised while reviewing even these projections from the quarterly model. For instance, the readings shown in figure 2 may in fact be revised. At this time last year, Iowa had a similarly poor Q1 growth rate that was later revised up to a slightly positive value.


We will release our GSP growth forecasts for 2017:Q1 in June when state personal income data for the first quarter becomes available. These projections can be found in the MEI background slides.

[1] The Seventh District comprises all of Iowa and most of Illinois, Indiana, Michigan, and Wisconsin. Further details are available at and

[2] Scott Brave and Norman Wang, 2011, “Predicting gross state product growth with the Chicago Fed’s Midwest Economy Index,” Chicago Fed Letter, Federal Reserve Bank of Chicago, No. 293, December,

[3] The entirety of the five states that are part of the Seventh District is considered for the MEI.

[4] The MEI is constructed as a three-month moving average. Hence, to obtain the quarterly average values shown in figure 1, we use the MEI readings corresponding to the last month of each quarter.

[5] We aggregate the five state values to one value for all Seventh District states using their nominal GSP shares. The horizontal solid black line in figure 1 corresponds to the average GSP growth rate over the sample period.

[6] We arrive at the Seventh District forecasted value for 2016:Q4 by aggregating the state forecasts using nominal GSP shares from 2016:Q3.

Detroit and Chicago: Real Property Value Comparisons

Both Chicago and Detroit have become poster children for city government financial stress in recent years. Chicago’s city and school district alike have been running structural deficits, meaning that the government has been covering its normal operating expenditures by issuing or over-extending debt and running down its assets. Both Chicago’s municipal government and school district face large shortfalls in required contributions for the future pensions of current and retired employees. Both have raised local property taxes as partial steps toward balancing their budgets. In the case of Detroit, the city has only just emerged from Chapter 9 bankruptcy while its school district teeters on insolvency under state-mandated “emergency manager” operation.

Are these places comparable in terms of their outlooks and situation? The value of real property in each place offers one fascinating indicator of the resources available to their local governments, as well as a look into how private homeowners and commercial property owners perceive the general prospects of Detroit and Chicago.

Why examine real property values? In some sense, real estate and improvements are long-lived assets that are largely fixed in place. In the market for these properties, buyers and sellers must assess and incorporate the government fiscal liabilities and service benefits – present and future – attached to these properties in the prices at which they buy and sell. High (and rising) property values may indicate that home owners and commercial property owners expect that the prospects for value in these locations are good and that they will continue to improve. And from the local government’s perspective, high values indicate that there may be room for further imposition of local taxes to fund government services, if need be.

Nuts and bolts

The estimation of the value of real property—land and improvements– in a locality is far from an exact science. The actual sales prices of property can be thought of as one reflection of an individual parcel’s value. However, as we saw during the financial crisis last decade, sales transaction prices can be very volatile, and sometimes speculative. More practically, parcels of property do not turn over frequently, so that transactions prices of all property parcels are not observed in any one year. In practice, then, local public officials often rely on various estimation methods in assessing the value of property for taxation purposes, most of which involve using a sample of information from similar properties that were sold during a year. From the recorded sales price and a property’s particular features such as size, location, age, and configuration, the property assessment office infers the value of each property and, ultimately, the total value of property against which taxes are levied.1 These taxable values are often termed assessed values or sometimes equalized assessed values and often represent some fixed percentage of market or true value.2

In the charts, we draw on such data from the local and state governments of Detroit, Chicago, and Illinois to estimate the market or true value of real property in both Detroit and in Chicago.3 For Detroit, figures on total assessed property value are published in the city’s Comprehensive Annual Financial Report (CAFR). By law, assessed value must amount to one-half of true value in Michigan. And so, to arrive at our estimates of full value, we simply double the assessed value. We believe that this yields a far upper bound on the value of residential property in Detroit because there is much evidence that, following the steep plunge in Detroit’s property market over the last decade, assessments were not reduced to accord with actual market value in a timely fashion.4

Drawing on prices of homes that recorded sales, the following chart shows that average home prices began to fall in 2004, while assessed value did not begin to fall until 2009.Detroit_Assessed_HomePricesFor Chicago, a prominent local government “watchdog” and research foundation has long been using state and city data on recorded property sales by class of property to estimate full market values.5 The accuracy of these data is believed to be reasonable, though far from exact.6

What do we see?

Due to lags in data availability, the estimates below are representative through calendar year 2014. The charts display total property value across all types, as well as the largest category in both Chicago and Detroit, that being residential property.

As shown here, Chicago real property value rose dramatically during the early part of last decade before dropping off just as dramatically. Detroit’s property values remained flat. However, Detroit’s apparent stability belies the fact that assessed values of residential properties have not been allowed to fall off in tandem with actual market transactions price there. As measured by volume of sales, the market for residential real estate in the city of Detroit became almost nonexistent during this period.7 Few homes were sold using conventional financing; almost all of them sold for cash. By some estimates, prices of homes sold fell by many multiples during this time, though they have since been heading back up in parts of the city.

Even using generous measures, and with rising home prices in some neighborhoods, residential property value overall in Detroit have continued to drift lower in recent years. In contrast, following the steep decline, Chicago property values have begun to recover for both residential and commercial (not shown) property.

Most telling, at over $80,000 per resident, the value of overall taxable real property in Chicago remained markedly higher than that of Detroit as of 2014. By even our generous measure, Detroit’s property values were only about $25,000 per resident.Detroit_Chicago_PropValuesDiscussion

It would appear that, as measured by real estate values, Chicago’s economy and prospects remain much stronger than Detroit’s. And from a local government perspective, Chicago’s taxable resources with which to pay down liabilities and fund public services appear to be much larger. Of course, there are myriad political and institutional factors at play that render such a simple assessment of wealth inadequate to characterize the fiscal capacity of these cities. In both places, for example, property wealth is concentrated in a subset of places such as near the downtown areas and along the waterfronts. Accordingly, it may be difficult to tap available property wealth selectively because existing statute largely requires that tax rates be applied uniformly. And voters and their representatives may be reluctant to allow tax hikes at all. Similarly, there may be different levels of sensitivity to taxation in these two places and among different constituencies. For example, the imposition of new and higher taxation may cause economic activity and investment to decline more sharply in one place as opposed to another.

More broadly, we might ask whether property wealth is a good indicator of potential resources that local governments may draw on to fund services. A look at more U.S. cities may be helpful. The next chart undertakes the same exercise for the most populous cities. Here we see that Chicago continues to look fairly robust by this measure, though less so than the cities of San Diego, Los Angeles, Austin, and New York City.MajorCities_ProvValues

  1. Assessed value of property for taxation purposes is often a fixed percentage of market value across all property parcels, or else it is a fixed percentage across all parcels of a certain type or class such as residential, commercial or industrial. In turn, estimates of full value of all property can be made by taking sample average ratios of “assessed value/sales price” and applying these ratios to all parcels’ assessed values.
  2. Equalized assessed values refers to the practice of further adjusting the totalities of assessed value of property across jurisdictions so that each locale’s assessed valuation represents the same or “equalized” value in relation to (percent)  sales price or true value.
  3. The city of Detroit also taxes tangible personal property of commercial enterprises such as computing equipment and furniture; Chicago does not.
  4. Using sales price and assessed values for a sample of 8,650 residential parcels in 2010, Hodge et al find an average assessment to sales price ratio of 11.47, which suggests an average over assessment or property values many times over. See Timothy R. Hodge, Daniel P. McMillen, Gary Sands, and Mark Skidmore, 2016, “Assessment Inequity in a Declining Housing Market: The Case of Detroit,” Real Estate Economics.
  5. See
  6. Discrepancies arise because only sample values of real estate transactions are available in any one year. In addition, full value projections derive from the median value of property in each class. However, the median property value may not represent the entire distribution of property values.
  7. See

Updated Forecasts of Seventh District GSP Growth

Several years ago, the Chicago Fed began providing estimates of annual gross state product (GSP) growth for each of the five states in the Seventh Federal Reserve District.1 The U.S. Bureau of Economic Analysis (BEA) releases annual GSP data for the prior year each June. This post discusses GSP projections for 2015 and presents an alternative forecasting model using quarterly GSP data from the BEA.2

The 2015 Growth Picture

To provide context for our projections, we first take a brief look at the main indicators of our model. Figure 1 shows annual U.S. gross domestic product (GDP) growth and GSP growth aggregated across the five states in the Seventh District from 2005 through 2014. Actual GSP data for 2015 will not be released for another month. However, we can get a sense of what this data is likely to show by comparing the recent histories in the figure. While growth in the Seventh District has lagged behind the nation in recent years, it has tended to follow a similar trend over longer periods. The U.S. maintained an annual growth rate of around 2.4% from 2014 to 2015, providing a reasonable starting point for our estimate of District growth in 2015.


The Chicago Fed’s Midwest Economy Index (MEI) then provides a useful link between national and regional growth that can give us a sense of the likely persistence of the recent shortfall between District and national growth. As figure 2 shows, the MEI indicated that the Midwest economy experienced growth that was somewhat above trend in the first half of 2015 and slightly below trend in the second half. Additionally, the relative MEI dipped below zero in the third quarter, suggesting that Midwest economic growth was further below its trend than national growth was in the second half of the year. Though the MEI does not explicitly pertain to GSP growth, historically a zero value for the index has been roughly consistent with 1.5% annual GSP growth for the District. In light of this, both the MEI and relative MEI suggest that District GSP growth in 2015 likely rebounded from its 2014 rate of 1.1% to somewhat above its trend rate of 1.5%, but still below the national growth rate of 2.4%.


Finally, we turn to annualized quarterly growth of state real personal income over 2015which provides an indication of how state-specific factors may have affected District GSP growth in 2015. Figure 3 shows that both Illinois and Michigan experienced strong income growth in the first quarter of 2015. Though the District states generally experienced weak income growth in the second quarter in comparison with the national average, growth rates in the remaining quarters of 2015 were of similar magnitudes to the national rate. Taken together, these data suggest some likely variation in GSP growth rates across the District states, but for the most part, they are consistent with the MEI and U.S. GDP growth data in figures 1 and 2.


Forecasts for 2015

Our forecasts for 2015 combine the information in the indicators discussed in the previous section to arrive at an estimate of annual GSP growth for the District states and the District as a whole. Since 2011, the Chicago Fed has used the following statistical model to estimate annual GSP growth:


This model explains the annual GSP growth rate of each Seventh District state as a function of national GDP growth, regional economic conditions as captured by the monthly MEI and relative MEI, and state-specific conditions (specifically, quarterly real personal income growth and annual GSP growth in the previous year).3 We aggregate state projections into a District-wide forecast using each state’s respective share of nominal District GSP.

Figure 4 shows for each District state and the entire District their respective historical GSP growth (blue bars), in-sample fits (orange lines) of GSP growth obtained from our statistical model, and 2015 out-of-sample projections (green lines). With the exception of Iowa, the model predicts an increase in the GSP growth rate for the Seventh District states, as well as the District as a whole. Interestingly, this seems out of line with the national GDP data and the MEI and relative MEI data discussed earlier. It is of note, however, that for 2014 the model also estimated higher GSP growth than what was realized for each state.


Motivated by our model’s recent shortcomings, we developed a similar model estimated using the experimental quarterly GSP data recently published by the BEA. Figure 4 also contains projections of annual GSP growth (red bars) obtained from this new model. At the time of writing, these data were available through the third quarter of 2015. To obtain a GSP growth projection for all of 2015 with these data, we need only estimate the fourth quarter’s value. Using the new statistical model to obtain this estimate and combining it with the data from the first three quarters, we arrive at an alternative forecast for 2015.

Figure 4 clearly shows a large difference between our two forecasting models. As table 1 further demonstrates, the projections from our new quarterly model (Q4 forecasted column) are below those of our original annual model in every instance, and often by quite large magnitudes. For the District as a whole, our quarterly model predicts more modest GSP growth of 1.6%, compared with 2.5% as forecasted by the annual model. This estimate from our quarterly model is also in line with the previous evidence suggesting growth was slightly above the historical trend of 1.5% but below the national growth rate of 2.4% for 2015.

Table 1. Annual GSP Growth Forecasts for 20154


To illustrate the sources of the discrepancy between our model forecasts, we plot in figure 5 the annualized quarterly GSP growth data from the BEA (blue bars) for 2015, including our fourth quarter estimates (red bars) and fitted values from the model (orange lines). It is important to note that both our annual and quarterly models use the same 2015 data for the variables that they share in common, with the exception of the three quarters of GSP data that we have for 2015. These data show sharp contractions in GSP for every District state (with the exception of Illinois) in the first quarter. More than anything else, this feature of the quarterly data is the dominant source of the discrepancies between our annual and quarterly model projections. The model fits in figure 5 make this clear, as they demonstrate very large negative residuals in the first quarter and only small misses in the other quarters, reflecting the fact that the declines in GSP in the first quarter are not consistent with the indicators in our statistical model.


It is possible that the quarterly GSP data for 2015 will be revised upon the release of the annual figure next month. As noted previously, our model predicts weak first quarters for the District states, but not nearly as dramatically as what has been released by the BEA thus far. Considering the apparent inconsistency between the quarterly data and the model, we also generate a forecast for 2015 that uses the fitted values for the first three quarters of 2015 GSP growth instead of the quarterly data. These projections, presented in the Q1–Q4 forecasted column of table 1, are larger than the quarterly model’s estimates using the quarterly values for 2015, but are still below those of the annual model. Moreover, these projections suggest that at 2.0%, District GSP growth improved in 2015 and was closer to the national average, but still below it.


We will continue to monitor the performance of both our annual and quarterly forecasting models. However, based on the results presented here, we intend to report annual GSP growth rates for each District state from our new quarterly model combined with the available quarterly data for the 2015 forecast (as presented in the Q4 forecasted column in table 1). From now on, we will continue to report estimates from this model as long as the quarterly data from the BEA make it possible to do so.

  1. The Seventh District comprises parts of Illinois, Indiana, Michigan and Wisconsin, as well as all of Iowa.
  2. The quarterly GSP data provided by the BEA is still in an experimental phase. For more information, visit
  3. The model is explained in more detail in Brave and Wang (2012).
  4. To allow for “like-for-like” comparisons among District forecasts, we aggregate state-specific annual forecasts to the District level using annual nominal GSP shares. The 2015 projections were aggregated using the 2014 shares.

Introducing the Chicago Fed Survey of Business Conditions (CFSBC)

On January 13, 2016, the Chicago Fed published the inaugural release of its Survey of Business Conditions (CFSBC). This new data product offers a series of diffusion indexes that track a broad set of topics related to the business conditions of firms in the Seventh District: overall activity (or product/service demand), outlook for the U.S. economy, current and planned hiring, current and planned capital spending, and wage and nonwage cost pressures.

The indexes are derived from questions included in a survey of Seventh District business leaders (started in March 2013) that was originally designed to support the Chicago Fed’s contribution to the Beige Book. Table 1 shows that CFSBC respondents come from a variety of industries, with the largest representation coming from the nonfinancial services sector and the manufacturing sector. The survey has averaged about 75 respondents over its history, though more recent surveys have averaged about 100 respondents.

CFSBC - 1 - Tab1

While the Beige Book is explicitly an anecdotal (or qualitative) account of current economic conditions, we decided to also include quantitative questions in our survey so that we can calculate indexes. These questions follow a seven-point scale. For example, here is our question about product/service demand:[1]

In the past four to six weeks, demand for my firm’s product or service has

  • increased substantially.
  • increased moderately.
  • increased slightly.
  • not changed.
  • decreased slightly.
  • decreased moderately.
  • decreased substantially.

The diffusion indexes we calculate from questions such as the one above are designed to be leading indicators, capturing changes in the prevailing direction of economic activity. The formula for the indexes is:

CFSBC - 4 - Formula 2

In many ways, this is a very traditional formula for a diffusion index, but we make an important adjustment that we would argue improves it: we measure individuals’ responses relative to their respective average responses. To calculate a respondent’s average response to a question, we assign numerical values ranging from +3 to –3 along the seven-point scale, and take the average across all responses. We then count a response as positive if it is above a respondent’s average response and negative if it is below a respondent’s average response. For example, if a respondent’s average response is +1.5, substantial and moderate increases are counted as positive responses and all other answers are counted as negative responses. Given our formula, the index ranges from +100 to –100 and will be +100 if every respondent in a given survey has an above-average response to a question and –100 if every respondent has a below-average response.

Calculating the indexes using a survey participant’s average response as a baseline—also known as detrending—allows us to correct for two types of potential biases. First, individuals may interpret phrases such as “substantially increased” differently, so that our numerical scores have different meanings for different people. Second, the industries and firms represented in our data may have different growth trends than the overall economy, which could bias the indexes because we do not have a random sample of respondents. For example, because manufacturers represent a sizable share of our respondents, our index could overrepresent trends in the manufacturing sector. The share of manufacturing output and employment in the U.S. economy has been declining for decades, so that in general, trends in the manufacturing sector are becoming less and less representative of trends in the overall economy.

Respondents’ respective average answers to a question can be interpreted as representing their historical trends or long-run averages. Thus, when we summarize respondents’ detrended responses by calculating diffusion indexes, we interpret zero index values to indicate that, on balance, the activity indicators are growing at their trend rates (or that outlooks are neutral). Likewise, positive index values indicate above-average growth (or optimistic outlooks) on balance, and negative values indicate below-average growth (or pessimistic outlooks) on balance.

For a concrete example of how we interpret the diffusion indexes, consider figure 1, which shows the CFSBC Activity Index (which is based on the question about product/service demand). It shows that 2013 was a bumpy year, with growth in activity about at trend on average. According to the index, 2014 started slowly (as you may recall, there was a lot of bad weather in the winter of 2014–15), but growth was consistently above trend for the rest of the year. Activity then slowed throughout 2015 to the point that growth was below trend by year’s end.

CFSBC - 2 - BBAct

We’ve found that the CFSBC Activity Index seems to do a good job of tracking U.S. real gross domestic product (GDP) growth, so that the index may be a good early indicator of the current state of the economy. Figure 2 shows this relationship, where I’ve aligned a zero value of the CFSBC Activity Index with the average of real GDP growth over the comparison date range. In general, the CFSBC Activity Index is positive when real GDP growth is above average and negative when real GDP growth is below average.

CFSBC - 3 - ActGDP

Because we use the results of the CFSBC to write our contribution to the Beige Book report, we will be releasing the CFSBC index data in conjunction with the Beige Book schedule. The Beige Book is released at 1:00 p.m. ET on the Wednesday two weeks before FOMC meetings, which take place eight times per year. The CFSBC index data are released two hours later.

For a more detailed description of the diffusion indexes and their properties, see the Chicago Fed Economic Perspectives article titled “The Chicago Fed Survey of Business Conditions: Quantifying the Seventh District’s Beige Book Report.”

[1] The full set of questions used for the CFSBC indexes is available here.

2015 Civic Research Forum: Finding New Approaches to Analyzing Urban Data

In partnership with World Business Chicago, the Federal Reserve Bank of Chicago hosted the Civic Research Forum on March 17, 2015. The forum was attended by researchers from a wide range of organizations and agencies throughout Chicago. It offered attendees an opportunity to discuss their current research and their positive and negative experiences with collecting and using data. Rob Paral of Rob Paral and Associates addressed the gathering, discussing his research on demographic trends in Chicago. Moreover, he described his research challenges given the  lack of some key historical data series, as well as the structure of available data sets and surveys. He also encouraged the audience to brainstorm innovative ways of using the accessible data.

Paral explained that his firm helps strengthen relationships between organizations and the broader communities they serve by providing data on the city’s social and economic conditions. He also shared that his firm gathers information on residents’ activities and attitudes. In his presentation, Paral focused on the income data he uses to study demographic trends in Chicago. While socioeconomic and demographic U.S. Census data are available for the 77 Chicago community areas (see map below) dating back to 1930, the data necessary to calculate median household income only go back to 1970. The limitations of these historical data hinder the potential to analyze income developments in Chicago over time (both at the neighborhood level and across demographics).


According to Paral, constructing income data for Chicago became even more difficult when the U.S. Census Bureau’s geographic grid, which includes the boundaries of blocks, tracts, and Public Use Microdata Areas (PUMAs) changed for the 2010 U.S. Census. The boundaries for these geographic units were redesigned in such a way that researchers could no longer aggregate PUMAs to match Chicago’s geography. Furthermore, beginning with the 2009–13 American Community Survey, it was no longer possible to select the Chicago-portions of census tracts for the few tracts that cover areas both inside and outside of the city limits, like it had been in previous versions. This change made it impossible to construct precise data for some individual community areas by combining data on the component census tracts.

Paral went on to discuss some of the questions related to income trends in Chicago that he is currently probing. He said his research focuses largely on the period between 1990 and 2010. Over this span, Illinois’s household income increased and then fell rapidly—a trend that was seen nationwide, though not to the degree it was within the state. According to Paral, the average household in Chicago earned 10% less income in 2010 than in 2000. Moreover, while nine out of every ten community areas had less income in 2010 than they did in 2000, some lost much more income than others. For instance, average household income declined by as much as 45% in some community areas, while other areas have seen an increase in income.

Looking at income trends of individual neighborhoods over time reveals interesting patterns about how wealth and poverty shift and consolidate geographically, Paral explained. In 1990, the wealthiest community areas were located in the far Northwest Side, the far Southwest Side, and the downtown and near-north areas. So, Chicago’s wealthiest neighborhoods were fairly dispersed back then. However, wealth became more geographically concentrated over time. In 2000, Chicago’s wealthiest areas were near the North Side and along the lakefront. And in 2008–12, this remained the case. In 1990, the poorest areas were largely consolidated in an area just west and south of the Loop (Chicago’s central business district). But over time, the poor moved farther away from the city center. In 2000, Chicago’s areas of greatest poverty were on the West and South Sides. And in 2008–12, this was still the case. The greatest income losses between 2000 and 2008–12 occurred on the far South Side, while the greatest income gains over that period happened in the “inner ring” areas near downtown (primarily just west and south of the Loop).

Paral said that while median income is the principal indicator he uses to analyze many trends, he also takes advantage of other measures of wealth provided by the U.S. Census Bureau—such as average and total household income—to get more robust views of wealth patterns and distribution across Chicago. Paral explained that by taking the ratio of average income to median income, he is able to assess the extent to which income distribution is skewed in an area—that is, how great the disparities are between the wealthiest and poorest residents in a given neighborhood. For example, a very high ratio of average income to median income suggests that there are some very wealthy residents pulling up the average (even if a majority of that neighborhood’s residents are poor).

As Paral shared, using innovative ways of combining available data, such as this method of studying the ratio of average to median income, has allowed him to examine potential weaknesses in current policies that are based on more-conventional income data and analysis. For example, he questioned the use of census tracts to determine how children are placed into public Selective Enrollment High Schools within Chicago. The current policy assigns each census tract to a socioeconomic tier, where 1 is the poorest and 4 is the wealthiest. The policy is designed to give children of poorer tiers a better opportunity of enrolling in a strong public school. However, some very poor children live in the same tract as very wealthy children. This means the average income is skewed up by these wealthy families, decreasing the chances the poor children have of being accepted into a strong public high school. Paral mentioned several census tracts where this pattern is especially problematic, such as those where there are large public housing buildings nearby very wealthy homes. In such tracts, the gap in median income between one (poor) subsection and another (rich) one can be over $100,000. Yet, because both poor and wealthy households are located in the same census tract, their children have an equal likelihood of entering the city’s best public schools.

The forum ended with the attendees sharing the research they are working on, the data they use, and any challenges they are facing. This portion of the program gave researchers the opportunity to learn of new data sources and approaches to analysis and to meet others in the Chicago research community. The forum sponsors hoped that innovation and collaboration among the attendees would eventually yield more-productive research in the coming years.

Infrastructure and Economic Growth — A Conference Preview for November 3

A common trait among economists is that they rarely agree on anything. However, the latest survey of economic experts by the Initiative on Global Markets of the University of Chicago’s Booth School of Business found unanimity on the value of infrastructure to the economy. When the 44 participants were presented with the proposition, “Because the U.S. has underspent on new projects, maintenance, or both, the federal government has an opportunity to increase average incomes by spending more on roads, railways, bridges and airports,” exactly zero disagreed. When further asked whether the U.S. has underspent on infrastructure, 36 agreed, 3 were uncertain and 5 did not respond. While such overwhelming agreement among economists might scare some people, it does suggest that the best economic researchers clearly have identified a relationship between infrastructure investments and economic growth. However, this same group of economists was skeptical about the efficiency of infrastructure programs. Nearly half agreed that past experience suggests that many infrastructure projects would have low or negative returns. As Austin Goolsbee put it, “hard to argue with the reality that some money will end up in powerful [Congressional] districts without much need for it.”

Given the perceived value, why has the U.S. apparently fallen behind in the infrastructure race? One theory is that fiscal pressure at all levels of government during and after the Great Recession caused governments to put off infrastructure investments in order to balance operating spending. Evidence for this shows up in data on the average age of government fixed assets, which have risen from 21.6 years to 22.4 years since 2007 (see figure).


Further complicating this has been a cloudy picture for funding sources directly related to infrastructure spending. The most prominent federal source, The Highway Trust Fund, has seen growth in gas excise tax revenue steadily erode as the 18.4 cents per gallon rate has been unchanged since 1993 while vehicle travel has declined. This year, Congress acted to prevent a deficit in the trust fund through a short-term fund transfer and by allowing companies to smooth pension returns, which would boost tax revenues in the short run. This clearly will not be a long-term fix. At the same time, state and local governments have faced very difficult fiscal conditions emerging from the Great Recession. For most states, revenues are only now returning to pre-recession levels. States that rely on excise taxes on fuel to fund infrastructure have seen the same erosion in revenues as the federal government, while states with sales taxes have suffered from declining gas purchases. In Illinois, gas tax receipts fell from $1.59 billion in 2007 to $1.21 billion in 2013 adjusted for inflation. Similarly, vehicle miles traveled per capita in the state have fallen by 6.5% since 2004. This inability of fuel tax revenues to keep pace with inflation has left dedicated infrastructure spending squeezed.

On November 3, the Chicago Fed will host a half-day program looking at key issues related to infrastructure. The first panel will start with a presentation from Therese McGuire of Northwestern University’s Kellogg School, who chaired a Transportation Research Board study that examined the role of the infrastructure component of the American Recovery and Reinvestment Act of 2009 on economic outcomes. Joining McGuire will be Dan Wilson from the San Francisco Fed, who has written extensively on infrastructure and will present his work on measuring the economic impacts of highway infrastructure. (For an example, see Rounding out the panel will be Tracy Gordon, who recently joined the Urban Institute after a stint at the Council of Economic Advisors and will discuss how to incentivize state and local governments to do more to fund infrastructure.

The second panel will examine methods for paying for infrastructure. Ben Husch from the National Conference of State Legislatures will discuss the current status of federal infrastructure funding, including prospects for the Highway Trust Fund. Michigan state budget director, John Roberts, will discuss a comprehensive infrastructure funding proposal that was introduced by Governor Rick Snyder this year. This proposal would have boosted infrastructure funding for the state and allowed for future fuel taxes to be indexed to inflation. Joining us from Oregon will be James Whitty, who has been responsible for administering the state’s pilot effort for a vehicle mile tax. States are considering this type of tax as a possible replacement for traditional fuel taxes. Finally, public–private partnerships are frequently seen as key to expanding infrastructure funding. Stephen Beitler, CEO of the Chicago Infrastructure Trust, will discuss how this new approach is working.

The program will be held at the Chicago Fed on November 3, 2014, starting at 8:30 a.m. There is no charge to attend. To register, follow this link.

Michigan’s Automotive R&D Part II

By Thomas Klier, Bill Testa, and Thomas Walstrum

The automotive industry is somewhat synonymous with Michigan. This relationship was born of an explosion of technological innovation in Southeast Michigan, including the assembly line and key developments in the internal combustion engine and transmission system. Looking at innovative activity today, a hundred years later, it is not far-fetched to state that the geography of automotive innovation in North America resembles that of yesteryear, with Michigan retaining its dominant role. The state has been highly successful to date in sustaining its leading automotive R&D concentration. Yet, for good reason, policy initiatives in the state are aimed at retaining and building on its strength.

The research and development (R&D) activity of private industry is increasingly being recognized as an important part of the innovation that spurs economic growth and competitiveness. Companies undertake R&D both to improve their production processes for cost and quality and to create wholly new products and services.

Among mainstay U.S. industries, automotive remains one of the most innovative in this regard. R&D that was both financed and performed by U.S. domiciled automotive companies amounted to $11.7 billion in 2011, representing 5.2 percent of total R&D spending. The R&D intensity of automotive manufacturing (as a share of the industry’s value added) is 15.3 percent, compared with 9.2 for all manufacturing, and 1.7 percent for all private industry.[1]

The importance of innovation to automotive companies remains paramount. A recent report by the Boston Consulting Group cited nine automotive companies among the world’s most innovative companies in 2013. The report names several factors behind the innovative burst among automotive companies, including the quickly tightening fuel-efficiency and environmental standards, which have spurred interest in electric and hybrid vehicle technologies. At the same time, auto companies continue to strive to meet ongoing demands for safety, comfort, and performance. Today’s vehicles increasingly comprise advanced electronic and IT components, which are developed both by automotive companies and purchased from technology companies in other industry sectors. By one estimate, “Electronics make up nearly 40% of the content of today’s average new automobile, and their share will continue to grow.” R&D initiatives to enhance the performance and to lower the cost of batteries that may power many of tomorrow’s autos are one example of an important and emerging R&D direction; automatic guidance systems for tomorrow’s (driver-less) cars is another.

Today, Michigan remains the epicenter of automotive R&D in the U.S. The state has maintained its leading place even while production has dispersed throughout the nation. According to data from the National Science Foundation that has been assembled for recent years only, R&D that is both funded and performed by auto companies in Michigan held fast at between 70 and 80 percent of the nation’s total from 1998 to 2011, amounting to $8.87 billion in 2011.

Automotive R&D has propelled Michigan to a leadership position among Midwest states. The table shows Michigan leading the region with $13.7 billion in total business-performed R&D by all industries in 2011, closely followed by Illinois ($12.0 billion), but far ahead of Ohio, Indiana, and Minnesota.[2]

Michigan is also a leader in employment of auto engineers to support long-term R&D and innovation. Drawing on data from the Census and the more recent American Community Survey, we can see how large Michigan’s share of the nation’s automotive engineers is relative to its share of the nation’s work force. Michigan employs over one-half of the nation’s automotive engineers, but its work force overall represents just 3 percent of the nation’s. Granted, Michigan’s share of the nation’s automotive engineers has fallen by ten percentage points since 1980; nonetheless, the state has added 18,000 (two thirds) of its engineers since 1980.

The remarkable importance of automotive technology in Michigan (as represented by engineering workers) can also be understood by comparing it with Michigan’s eroding share of automotive production. By overlaying Michigan’s automotive production workers as a share of the nation on the chart above, the strong role of automotive technology becomes clearer. Since 1950, Michigan’s share of production workers has fallen from 54 to 19 percent, a loss of approximately 255,000 jobs.

And while there are many technologically advanced industries in Michigan—including bio-pharma, medical equipment, industrial chemicals, and office furniture—automotive engineering has come to dominate further in recent decades. As the chart shows, automotive engineers once comprised 30 percent of engineers in all industries in Michigan. By 2012, their share had risen to 51 percent.

What is the future of automotive R&D in Michigan; will the region’s extreme concentration in the activity continue? There are no hard and fast answers, yet there are identifiable features that will come into play. On the one hand, there are many historical instances of geographically concentrated centers being very cohesive and long-lived. Once established, such “clusters” tend to grow and feed on themselves. Technology activity is drawn to technology activity. Skilled workers are drawn to activity-rich cities, and companies are, in turn, drawn toward pools of skilled workers. For example, global financial centers such as New York and London have held their dominant positions for many decades, even centuries. The San Francisco Bay area has enjoyed a long run of dominance in the areas of IT and biotech. In a similar way, Michigan’s established leadership in automotive R&D may persist.

On the other hand, company reorganizations and the geographical shifting of activities that tend to be interdependent with technological activities represent risks to Michigan’s position. It may be beneficial for some industries to locate technology and production in close proximity to facilitate (and reduce the cost of) communication and transportation between the two activities. Thus, the fact that Michigan has lost automotive production in recent decades may have negative implications for automotive R&D in the state.

Similarly, co-dependence between R&D and company headquarters activities such as marketing and strategic planning has also been seen as important in some industry sectors. Thus, any major shift in corporate headquarters activity away from Michigan would raise the concern that it might be accompanied by a shift in R&D activity.

Finally, mature industries such as automotive are often severely disrupted by the emergence of wholly new and sometimes unexpected technologies that greatly shake up their organization and geography. For example, the development of aerospace technologies for military uses shifted the locus of related U.S. production from the Northeast to the Southwest and West during the course of the twentieth century. So far, this has not yet taken place as Michigan continues as the U.S. leader in automotive innovation and R&D activity.


[1]For 2011, National Science Foundation, National Center for Engineering and Scientific Statistics, Business R&D and Innovation Survey, and U.S. Department of Commerce, Bureau of Economic Analysis. (Return to text)

[2]Latest data from the National Science Foundation, available here. (Return to text)

Manufacturing as Midwest Destiny

By Bill Testa and Norman Wang

In the Midwest, the terms “industrial” and “cities” are almost synonymous. Though agriculture has been important to growth and development, the region’s economy was built on manufacturing, and the sector continues to be prominent—for both small towns and large metropolis alike.

However, labor and income generated from the region’s factories began to wane 40 to 50 years ago. In response, the region’s cities have undertaken deliberate development strategies to maintain their economic vibrancy. Some strategies have focused on the historic mainstay—manufacturing—while some have focused on diversification into service sectors ranging from tourism to business services and finance. These efforts have met with mixed success, and the industry mix of most Midwest cities continues to be steeped in manufacturing. Accordingly, “Industrial cities” of the Midwest continue to address the same fundamental challenge—that is, how to sustain their communities as manufacturing’s ability to generate jobs and income continues to decline.

The chart below looks at manufacturing’s share of jobs going back to the year 1969. In both the Great Lakes region and in the U.S., the share of jobs to be found in manufacturing has declined by one half or more. A much greater share of workers now find employment outside of the manufacturing sector than are employed by the sector.[1] Importantly, though, the Great Lakes Region continues to be more highly specialized in manufacturing as compared to the U.S.

At a more granular level, the chart below illustrates the many metropolitan statistical areas (MSAs) with higher job concentrations in manufacturing than the U.S. as a whole As seen, these include very populous MSAs, such as Detroit and Milwaukee, but also many smaller MSAs, such as Decatur, Illinois, Jackson, Michigan, and Cedar Rapids, Iowa. Even some of those MSAs with smaller shares, such as Flint, Michigan, have diversified out of manufacturing only under the pain of wholesale loss of jobs, people, and income. (And Flint’s economy continues to decline).[2]

A more systematic illustration of how manufacturing has been a large part of the destiny of the Midwest can be seen in the charts below. On the horizontal axes can be found the share of manufacturing employment for all MSAs of population 100,000 and greater in 1969. The vertical axes measure subsequent (post-1969) total job growth and per capita income for each MSA. The inverse correlations are striking; the general tendency indicates that manufacturing-oriented cities fared worse as measured by growth of income and total employment. Further analysis of these cities (not shown) again indicate that the depressing growth tendency of yesteryear’s manufacturing orientation has not discriminated by population size; both big and small MSAs were similarly affected.[3]

Have any other industry concentrations or preconditions been important in determining the economic fate of Midwestern cities? Other researchers, such as Edward Glaeser, have emphasized that educational attainment of the adult population has been a strong causal determinant or precondition of MSA growth. The reasons for this finding are varied. It may be that this measure represents local workers and community leaders who were most capable of reinventing their home city when damaging shocks to the local economy took place. Alternatively, a large share of college-educated workers may simply reflect that the MSA already enjoyed an economic diversification into other key industries (employing college-educated workers) that did well after 1969. In any event, our analysis suggests that taking “percent of adult population with a college degree or more” into account explains more of the performance variation among Midwest MSAs. Together, the “share of manufacturing jobs” along with the “percent of adult population with a college degree” explain as much as 40 percent of the variation in Midwestern MSA economic growth after 1969.

Looking at these past determinants of growth, can we identify any room for local policy actions to shape the local economy? In the analyses above, we have used very simple measures of a local economy’s “industry mix” to suggest that historical development may have been destiny for many Midwestern towns and cities. Indeed a more careful accounting of each place’s historical industry mix might yield more telling findings and insights. For example, towns steeped in steel production, automotive, or television electronics may have had even less control over their destiny in recent decades. However, a more expansive view indicates that our measures of “industry mix” (above) explain only 40 or less percent of the 1969-2010 variation in performance among Midwest MSAs. This leaves much more performance to be accounted for. It seems that some places have improved their own economic performance through deliberate development policies such as work force training, tax incentives to business investment, land use reform and re-development, or public infrastructure investments.

For these reasons, the Chicago Fed’s “Community Development and Policy Studies” (CDPS) area has launched an investigation into how industrial cities in the Seventh District have taken deliberate steps to fashion their own destinies in more favorable ways. According to CDPS Business Economist Susan Longworth, an initial project step will be to “develop comprehensive community profiles of cities throughout the Federal Reserve’s Seventh District that had populations of at least 50,000 and had 25% or more of their employment in manufacturing in 1960. Research includes in-depth qualitative information, combined with the best available quantitative analyses of the trends and issues impacting these communities to identify policies and programs that promote (or inhibit) economic growth and vitality in industrial cities.”


[1]The declining share of manufacturing is overstated because manufacturing companies have outsourced functions (and jobs) to local service sectors. These include the hiring of factory workers who are on the payrolls of temporary employment firms, as well as outsourcing of maintenance, payroll, and transportation workers to outside (service) firms.(Return to text)

[2]Some well-performing nonmanufacturing MSA economies are fashioned around state government capitols and major universities, such as Ann Arbor, Michigan, Madison, Wisconsin, and Columbus, Ohio.(Return to text)

[3]Some might wonder whether manufacturing orientation continues to influence community growth in more recent years. Our analysis of 1990 to date continues to show such influence widely across the Midwest, although there is tendency of a weakening correlation.(Return to text)

Net Migration and Regional Adjustment

By Britton Lombardi and Bill Testa

How do regions adapt following economic calamities that displace large numbers of workers? In the best case scenario, they reinvent themselves. For example, economist Ed Glaeser documents the several times that Boston rose from the ashes following collapse of its shipping and old line manufacturing industries. In a similar vein, Detroit hopes to rebuild as an economy based on logistics, energy, and high-tech manufacturing—especially next generation automotive technology and production.

However, in many more instances, rather than a quick turnaround, cities and regions undergo a painful adjustment period that leaves them smaller than in their heyday. In the process of adjustment, economists have shown that, following an economic shock, unemployment rises for several years before returning to more normal levels. Using the steel- and auto- related job losses in the early 1980s as an example, a paper discusses the long-run effects of massive job losses. The authors found that after an initial spike in the unemployment rate in the impacted local economies, the rate converged back to the national average after five or six years. However, high out-migration (and low in-migration) led to this reduction in unemployment, rather than an increase in new jobs. Similarly, a landmark study found that those U.S. multi-state regions that experienced downward employment shocks returned to a more normal path of employment growth within five to ten years. However, the lost jobs were never recovered.[1]

One lesson from such studies is that the ability to migrate from distressed areas can be helpful in restoring the lives and livelihoods of households that undergo economic displacement. Recently, there has been concern that the migration mechanism is broken or impaired. William Frey and others have documented a fall in interstate migration during the recent recession, along with a longer-term decline over recent decades. Nonetheless, much of the long-term decline is gradual and mobility in the U.S. remains high relative to other nations and continents.[2]

In the current environment of high unemployment and a weak housing market, some analysts have attributed dampened mobility of the labor force to falling house prices and “negative equity” or underwater households—that is, homeowners who owe more on their existing home mortgage than their home is worth on the market. Some economists posit that negative equity in a home may hamper the ability of the homeowner to move to a new labor market because they are “locked-in” to their existing house by their inability to raise funds to pay off their mortgage. A staff report from the New York Fed found that negative equity does reduce the probability of moving. However, in a rebuttal paper using the same data but differing coding of movers, the author concludes that negative equity does not make homeowners less mobile. If anything, a homeowner with extremely negative equity is actually slightly more likely to move than an individual with smaller negative equity. Using statistical analysis, the author finds that negative equity actually increases the probability of moving by 1 to 3 percentage points; this represents an increase of 10 to 18 percent of the overall probability of moving. Dan Aaronson, has conducted some preliminary research that finds that differences in migration rates of homeowners and renters barely changed during the recession and early parts of the recovery. Since renters should not suffer “house lock,” this evidence suggests that the migration process has not been hampered by falling home prices to a large degree. Similarly, Mark Partridge et al find that the migration falloff in the recent decade may fundamentally reflect heightened risk aversion of U.S. households.


So, how does this all play out in the Seventh District? Despite some impairments to mobility, interstate migration appears to be well underway in the region, especially in those states, such as Michigan, that have experienced the most damaging employment shocks. Michigan has been an outlier in its high unemployment rate, which has been well above those of the rest of the District states since 2004.

Click to enlarge.

As seen below, Michigan started experiencing net out-migration during the recession in the early 2000s. However, the out-migration rate picked up speed around 2004/2005 as Michigan’s unemployment rate started to drift higher than those of the other Seventh District states. Net out-migration from Michigan remained relatively high, at around 8 per 1,000 people, from 2007 through 2009.

Click to enlarge.

To check more recent trends, we turn to data released from United Van Lines and Atlas Van lines as they track interstate movements of U.S. households around the country. They have been tracking this data for a long time and so provide some historical data. In the chart below, we combine both companies’ data to try to produce comprehensive results and remove any anomalies that one company may experience in their sales.[3] We plot the outbound migration gap, or the extent to which the number moving out of the state exceeds those moving in. In the early 2000s, Michigan’s gap is in line with the rest of the Seventh District but, again, starting around 2004/5, Michigan’s gap greatly widened compared with the other states’ gaps, consistent with the unemployment observations above. The gap continued to grow during the recent recession; it recently began to trend back down, but still remains at higher than the rest of the District states’.

Click to enlarge.

Net out-migration is not a preferred solution to economic shocks for many reasons. Moving can be costly in terms of out-of-pocket expenses, and households can suffer from the severing of social ties as well as difficulties in learning how to negotiate daily living in new communities. Nonetheless, mobility can be a helpful part of the adjustment process when it allows workers and households to improve their standard of living and well being following a negative shock to local employment. For this reason, and in response to today’s difficult employment situation, researchers are asking whether public policy should intervene to assist those unemployed workers who would benefit from re-location, but who lack the funds to re-locate in search of jobs. One recent idea has been explored by Jens Ludwig and Steven Raphael. They “propose the creation of a “mobility bank” that would help finance the residential moves of U.S. workers relocating either to take or search for work….in depressed areas of the country.” These loans would be “amortized over a fairly long period (10 years), and repayment terms be contingent on the borrower’s post-move employment and income.” The paper suggests that the program costs would compare favorably to alternative federal programs designed to achieved re-employment.


[1]In contrast, Timothy J. Bartik finds persistent effects of shocks on unemployment rates and labor force participation; See here(Return to text)

[2]See here (Return to text)

[3]There may be some sample bias of individuals that actually use movers versus those that rent a truck and move themselves. (Return to text)