Economy versus lifestyle in the inter-metropolitan migration of the young: a preliminary look at the 2000 Census.

• Author: Gottlieb, Paul D.
• Position: Report

Abstract

This paper a measures net migration into (out of) 100 U.S. metropolitan areas between 1990 and 2000 by people aged 15-24 in 1990, modeling variation as a function of economic growth in the late 1980s, amenity variables in 1990-1993, and each metropolitan area's technology base in 1990. Multi-collinearity among independent variables is addressed by creating oblique factors to reduce seven variables to three; and through use of path analysis to specify direct and indirect causation (correlation) explicitly. Results suggest that economic growth and technology were more important drivers of youth migration in the 1990s than amenities. Because the study used comprehensive, but relatively traditional measures of place amenities, it was unable to directly test recent hypotheses about effects of hip, "Bohemian" culture and ethnic or gender preference diversity on the migration decisions of people in this age cohort.

INTRODUCTION

For economic development professionals of each era, there seems to be one indispensable policy lever for creating economic growth. In the 1960s and 70s, it was the need to drive down tax and labor costs. In the 1980s and 1990s, it was sports venues and downtown revitalization. It may be too early to tell, but this is shaping up to be the decade when development practitioners feel compelled to do one thing above all others: attract educated young people to their jurisdictions. Consider, for example, the recent proliferation of state programs designed to prevent the "brain drain" of young people trained in local universities (Chronicle of Higher Education, 1998; McLaughlin, 1999; Associated Press, 2000; Indiana Fiscal Policy Institute, 2000; Tornatsky et. al., 1998).

This new conventional wisdom is grounded in academic research as well as several broad, stylized facts. Many economists argue that human capital has replaced physical capital as the primary driver of economic growth (Lucas, 1988; Romer, 1990; Mathur, 1999). This means that the policy maker's first goal must be to attract smart people--not plants and equipment--to his/her own patch of ground.

The research university has also received considerable attention as a source of entrepreneurial growth in regions (Beeson and Montgomery, 1993; Bania, Eberts, and Fogarty, 1993). Educated young people are both an integral part of the research process at these institutions and an inevitable by-product of their broader mission (Florida, 1999; Bound, et al., 2002). If we look at metropolitan areas in the 1990s, we see positive correlations among per-capita income growth, concentration in industries typically regarded as high-tech--such as information science and biotechnology--and a high proportion of younger workers (1). It is logical to suppose that younger workers are disproportionately concentrated in cutting-edge sectors of the economy. They will migrate toward--and also propel forward--those regions that succeed in the so-called New Economy (2).

What factors, then, affect migration patterns of the young and educated? Universities are one answer: Ann Arbor had 11,000 more 20-21 year olds in 2000 than we would expect by aging the same cohort living there in 1990--and 8,700 fewer 30-34 year-olds (3). High-paying job opportunities are a useful, if not always necessary, condition for attracting graduates. Some young people are clearly attracted by the "bright lights" of cities like New York and San Francisco, and will be happy to serve cappuccinos until they get their big break.

In several recent reports and a widely-read book, geographer Richard Florida highlighted the role played by America's young "creative class" in the recent high-tech boom, and hypothesized that lifestyle amenities are a crucial location factor for this group of workers (Florida, 2000, 2002; Florida and Gates, 2001). While Florida is not the first to write about amenities as a source of economic growth (see, e.g., Ullman, 1954; Gottlieb, 1994, 1995), he has provided a valuable service in distinguishing amenities particularly attractive to the young from the standard list of museums, crime, and climate that journalists typically use to rank places (Savageau and Loftus, 1997; Money Magazine, annual).

It would clearly be useful to distinguish among several hypothesized causes of youth migration. This paper is the first to address this using data from the 2000 Census. It creates a standardized measure of net migration of the young into or out of 100 U.S. metropolitan areas between 1990 and 2000. Factors hypothesized to increase net in-migration are divided into three categories: 1. economic growth factors that increase opportunities to find a job or earn a high income; 2. technology factors that make a region attractive to the young and mobile, regardless of how fast it is growing; and 3. amenity factors that make a region attractive to the young regardless of #1 and #2.

The problem with this statistical framework should be obvious: we would expect there to be strong correlations among independent variables (4). Presence of such correlations has long been recognized to reduce our ability to measure the relative importance of each causal factor, e.g., by calculating standardized beta coefficients (Darlington, 1968). Ultimately this problem is insoluble, as it is a feature of the world we are analyzing and not a failure of statistical technique. Nevertheless, we will try to use accepted regression methods to separate migration drivers as best we can, reporting several tests and correlations to make our exploratory approach transparent.

STUDY APPROACH

For this study, we collected data on net migration into or out of 100 U.S. metropolitan areas between 1990 and 2000 (5). Our primary focus is on people aged 15 to 24 in 1990 / 25 to 34 in 2000. This cohort definition ensures that most individuals sampled were out of school by 2000, and have therefore made "economic" location choices. One can get additional insights by separating those likely to be moving from high school to college from those moving from college to work, or from work to work. That analysis is deferred to a future publication (6).

Next we assembled variables describing the three categories of migration drivers: economic growth, technology base, and lifestyle amenities. University graduates per-capita and the proportion of young residents were collected as additional control variables expected to affect youth migration. If structural or very slow to change (e.g., high-tech economy), independent variables were measured in 1990. If dynamic growth variables, they were measured from 1985 to 1990 (7). Both types of variables could then be interpreted as possible causes of subsequent 1990-2000 migration patterns. A correlation matrix (available by request) shows that there is indeed considerable inter-correlation among the variables. Of the 55 possible correlations in the matrix of raw independent variables, nine exceeded .30.

The next step in the analysis was to regress the dependent variable on all of the independent variables taken together. Our main goal here was to use F-tests to determine if broad categories of variables (such as the three variables that together describe amenities) were statistically significantly related to youth migration as a group.

If a metropolitan area experienced net in-migration of the 15 to 24 year old cohort, the dependent variable is positive. If a metropolitan area experienced net out-migration of the cohort, the dependent variable is negative. Either way, the net migration variable is standardized so that city size and raw body counts do not skew the results. (Appendix for a complete description of the logic and technique for creating net migration variables.)

The larger the dependent variable, then, the more "attractive" the destination (8). Reasonable predictions for the sign of the regression coefficient on each category of independent variable are as follows:

* 1985-90 economic growth: positive because a substantial proportion of this cohort will be job-seekers by 2000.

* 1990 technology base: positive for both job-seekers and college attendees within the cohort.

* 1990 amenities and climate: positive for the entire cohort.

* 1990 population: positive either because it proxies additional cosmopolitan amenities or because of a generalized gravity effect (9).

* 1990 proportion of the population in their 20s: positive for everyone in the cohort, on the ground that young people will move to places known to contain other young people.

* 1990 per-capita university graduates: negative, because most of the cohort aged 15-24 in 1990 is presumed to be out of school by the year 2000. College metros like Boston and Ann Arbor will experience net out-migration among those moving from university to first job. They will experience net in-migration among those moving from high school to university, but these represent a minority of our cohort, the youngest of whom were already 25 years old in 2000. For members of our cohort moving from high school to a post-college job over the decade--or not attending college at all--the sign of this coefficient is ambiguous, since the scale of universities at origin or destination is presumably neither relevant to the migration decision nor predictive of aggregate flow.

Many researchers would...