©2019 by Jonathan Gruber and Simon Johnson

102 Places for Jump-Starting America

Appendix to Jump-Starting America, with a technical description of the Technology Hub Index System (THIS)

America has many talented people, spread out across a large geographic area. In this appendix, we describe one way to measure the relative potential for various places to develop as technology hubs. What we propose here is far from being the only approach to measuring the strengths and weaknesses of particular locations. This is a suggestive exercise in- tended both to illustrate the enormous opportunity that already exists in all corners of our nation and to help start a more detailed conversation on whether, where, and how to jump-start the creation of more good jobs.


We use data on 382 metropolitan statistical areas (MSAs, which represent cities and the associated commuting communities). Appendix Table A1 summarizes our eight component measures, sources, and notes on construction.

Table A1

The American Community Survey (ACS), conducted by the Census Bureau, is the largest annual survey of households in the United States and collects a rich array of demographic and economic information. The Census Bureau makes data available at various geographical levels through the National Historical Geographic Information System (NHGIS). From this, we download variables at the MSA level. The ACS is largest at decennial years, when there is a census sample. House price data are not available for all MSAs in non-decennial years, so we use the latest available comprehensive values, which are for 2010. For commuting time, we measure the number of workers who report commuting less than thirty minutes to work as a share of all commuters.


For information on university quality at the graduate and undergraduate level, we use a 2005 survey carried out by the National Academy of Sciences. This survey uses a broad set of measures to rank the quality of graduate programs at US universities. We use these rankings to create an indicator for top-twenty programs in each field, and then we count how many such programs exist at every university in each MSA.


To measure the quality of undergraduate education, we use a survey of graduating PhD students collected by the National Science Foundation each year through its National Center for Science and Engineering Statistics (NCSES). This Survey of Doctorate Recipients (SDR) has been fielded since 1973, and it includes a variety of information about doctoral graduates, including where they previously received an undergraduate degree. We provided an official at the NSF with a file containing information on the top twenty graduate programs in each field (as described in the preceding paragraph), and he matched that to their survey. He then identified the undergraduate institution for each student graduating from a toptwenty PhD program over the 2005–2015 period. We counted the number of such students receiving an undergraduate degree from each county in the United States. We aggregated those data to the MSA level.


Patent data were provided to us by Shane Greenstein, as used in his paper with Chris Forman and Avi Goldfarb. They collected the number of patents granted by the US Patent and Trademark Office (USPTO), and from this set, they measure the number of new patents each year (year here is defined as year of application, due to delays in granting patents). We use 2010 values, as these are the most recent that they had available. We normalized this by the number of workers in each MSA (using data from the ACS, this time for 2010), as this measures the entrepreneurial nature of the workforce.


Crime data are from the FBI’s Uniform Crime Reports (UCR) system. We gathered data on violent crimes for every MSA in 2016 and then normalized by population. Violent crime includes murder, rape, robbery, and aggravated assault.


MSAs are a construct of statistical convenience. In thinking about potential technology hubs, it may make sense to combine MSAs. For example, Des Moines and Ames, Iowa, are fifty minutes apart. Des Moines is significantly larger, but Ames has a more highly educated population, high-quality educational institutions, and a higher rate of patenting. By combining them in our data, we create a technology hub that is both large and has strong educational attainment. Using this logic, we created twenty-four pairs of cities and two “triangles” where combining three MSAs seemed appropriate.


We do not claim to have considered all possible combinations. Many MSAs are within an hour of other MSAs and could be combined in various ways apart from our efforts. We also did this work only at the level of MSAs. There may well be combinations of smaller cities or nearby non-MSA areas that create economic development areas with sufficient population, high-quality educational infrastructure, and a good quality of life. We look forward to hearing more about this from readers.


As described in the text, we created our Technology Hub Index System (THIS) by first selecting a set of places in the United States that have—as we define it—sufficiently high population, educational attainment, and quality of life. For population, we choose MSAs with more than one hundred thousand workers age twenty-five to sixty-four. For average education, we choose a college-educated share of the twenty-five-year-old and older population that is greater than 25 percent (this is about 12 percent below the mean national rate of 27.9 percent for this variable). Our cutoff for mean house price is below $265,000, which is about 14 percent above the mean house price of $232,222 in 2010.


The results of this exercise are shown in Appendix Table A2, in which places are listed in order of their overall THIS ranking. The table has columns for:


  • MSA name

  • Total THIS ranking

  • Ranking for total population (one-third weight in THIS)

  • Ranking for each of four education measures (overall education ranking has one-third weight in THIS)

  1. Share of college graduates (one-quarter weight in overall education ranking)

  2. Number of top-twenty graduate programs (one-quarter weight in overall education ranking)

  3. Number of undergraduates going to top-twenty PhD programs (one-quarter weight in overall education ranking)

  4. Patents per worker (one-quarter weight in overall education ranking)

  • Ranking for each of three lifestyle measures (overall lifestyle ranking has one-third weight in THIS)

  1. Average house price (one-third weight in overall lifestyle ranking)

  2. Violent crime rate (one-third weight in overall lifestyle ranking)

  3. Commuting time (one-third weight in overall lifestyle ranking)


The list in Appendix Table A2 is comprised of large urban areas from thirty-six states. States that are excluded are largely due to high house prices (California, Connecticut, Colorado, Maryland, New Hampshire, and Rhode Island), insufficient population in the largest cities (Alaska, Delaware, Hawaii, Maine, Montana, Vermont, and Wyoming), or insufficient education according to our criteria (Nevada). Of course, all such variables can be changed in the long run, with sufficient local political will.

Table A2