Is Economic Growth Becoming More Concentrated?

We May Be Seeing a Moderate Amount of Re-Concentration

There’s a basic story you hear a lot in economic punditry: that the richest places are getting all the economic gains of the last few years. DC, San Francisco, NYC, they’re the big winners. This story has elements of truth to it. But today, I want to lay out some basic graphs that should give us pause in making those claims too aggressively. This post will be modelled off of my previous post, which argued that city growth is essentially uncorrelated with city size, so, if you didn’t read that post, go back and read it.

My basic thesis will be as follows: Although strengthening economic conditions do show some signs of increasing concentration, the relationship between the size of a city’s economic base and its growth rate remains fairly weak. That is, the big cities are capturing at most a slightly oversized share of the economic growth pie.

The Past is Prologue

Talking about population, I said:

The best predictor of future growth is past growth.

This is almost as true of major economic variables, like employment growth, as it is for population demographics. Let’s do some comparisons. The chart below shows the annual correlation between metro-area employment growth rates, and the prior year’s growth rate, as well as metro-area population growth, and the prior year’s population growth.

As you can see, last year’s employment growth rate does not predict this year’s employment growth rate as well as the related population growth series. But nonetheless, it’s a fairly strong relationship.

This relationship exists for the same reason the population time-series correlation exists: because the facts that drive employment growth tend to be stable. Now, they’re less stable than core population demographics, true! But nonetheless, last year’s performance for a local area is usually an okay predictor of this year’s performance, because economically significant factors tend to be relatively stable over time.

As with population, we can notice that the late 2000s did show a period where this relationship broke down, with new patterns showing up, and now we’re seeing correlation return, again with this new pattern. So maybe we’ve entered a new era where growth patterns are changing?

Employment Is Just Beginning to Predict Growth

The chart below shows the correlation between aggregate level of employment a year prior, and growth rate in a given year, as well as the same statistic for population. In other words, it tells you to what extent employment growth rates are biased toward places with the most pre-existing employment.

As you can see, again, the relationship between population growth and levels and employment growth and levels is quite similar. If “big cities get faster population growth,” then usually “cities with lots of jobs get faster job growth,” and vice versa. The time series for employment appears more volatile, but fundamentally quite similar.

We are indeed, right now, at a historically high level of correlation between size of the employment base and employment growth. Never before since we had data have we seen this much of employment growth be occurring in cities that already have lots of jobs. But, that said, this relationship remains extremely weak: size of the employment base has a correlation of just 0.20, which means it can explain just 4% of the variation in employment growth rates. Plus, this series is volatile and can decline very suddenly. Don’t count on it to last forever.

That said, commentators aren’t wrong to see in the last few years a strengthened relationship between employment growth and the prior size of the employment base. They’re observing a real thing, just, perhaps, being a bit too enthusiastic about the observed effect size.

Let’s compare existing relationships!

A Weak Relationship

The above chart shows you the data on growth rates by metro for population and employment, each line organized by pre-growth levels for their respective variables. As those who read the population article know, I argued that these lines really showed no pattern to a normal person’s eye. Looking at the employment series, it looks a little better! None of the largest 100-largest employment-base cities have seen negative job growth over the last 5 years. Not one! Now, granted, that’s because we’re coming out of recession, but, still. Seems somewhat notable.

So let’s look at what happens if, say, we do the grouping method I said in my population post was misleading; that it could cause you to think you see trends when there’s really nothing there.

As you can see, these average groupings we get are very similar. Yes, it looks like the biggest cities have substantially higher growth… but this is in some sense an optical illusion created by a few cities, and by averaging. Now, yes, the underlying series looks better than the population series, but if you recall our correlation charts earlier, it’s not much better.

For the record, the correlation between size of a city and its employment growth rate is about 0.2, but the correlation between its college-educated population share and its employment growth rate is about 0.27. Considered together, the Multiple-R is 0.31, not much of an improvement. One reason it’s not much of an improvement is that there’s a 0.29 correlation between employment base and the college-educated share. Put simply, to the extent that employment base is associated with the employment growth rate, it may be an artifact of its correlation with the college-educated share of the population.

A History of Job Creation

But if you think about it, I haven’t really disproved the basic claim that growth is getting more concentrated. All I’ve really shown is that any concentration is, at best, only weakly correlated with the size of the economic base. Maybe there is concentration, but it’s not about size.

We can directly measure concentration of job creation! To do so, we just need some measure of job creation that is positive for all metro areas. So “net change in employment” won’t do. Luckily, the Business Dynamics Survey tracks gross changes in employment! So we can see if “job creation” is getting more concentrated.

The numbers on the left axis are really just a kind of theoretical indicator; they have no “tangible” meaning. Rather, bigger numbers indicate job creation is very concentrated, smaller numbers indicate it is less concentrated.

As you can see, job creation was relatively concentrated in the 70s and 80s, but began falling in the late 80s, and by the late 90s had reached a kind of plateau. Job creation got even more evenly spread out in the mid-200s, but then the recession undid that. In the most recent year of data, job creation got more concentrated, the most concentrated since the early 90s. Now, this may tick back down in 2015. We’ll see.

Although job creation remains fairly diffuse compared to the 1980s, it does seem to have become more concentrated since the early 2000s, a common benchmark for comparison. We know that this period has also seen a strengthening in the relationship between employment base and employment growth: from 0 or a slight negative relationship, to, now, a positive relationship of about 0.2.

Combined, this slightly strengthened correlation between employment base and employment growth alongside greater concentration of gross job creation suggests that we are seeing a moderate concentration of economic activity in bigger cities. Let’s not overstate this: it is a moderate concentration. It may not have anything to do with the classic “agglomeration” stories. It may not be permanent. It remains less concentrated than periods in not-so-distant history.

Economic activity appears to be exhibiting a moderate amount of concentration in economically large cities.

A History of Firm Concentration

BDS data doesn’t stop with job creation. It also gives us information about firms, or companies. Before we look at firm concentration data, let’s look at just the history of BDS-tracked firms in metro areas.

As you can see, the period from 2007–2011 saw a pretty remarkable die-off of firms. Firm decline continued for a year longer than employment decline, by the way, and although BDS-tracked employment has basically recovered to pre-recession levels, BDS-tracked firm counts remain 1.6% below pre-recession levels. In other words, the same amount of employees are working at fewer companies. That right there is a kind of concentration. While employees-per-firm are not at their historic peak (22.71, in 2001), they are near it at 22.55.

Okay, so let’s look at Herfindahl-measured concentration.

Here I show concentration, not of job-creation or firm-creation, but just number of jobs and firms. As you can see, jobs used to be more concentrated than firms; they are now less concentrated than firms. Both series show declining concentration from the 1970s to the mid-1990s, but while employment concentration falls all the way to the mid-2000s, firm concentration is basically stable, and then begins rising at a good clip after 2005.

In other words, it seems possible that start-up activity is becoming more concentrated, though I do not have a direct measure of that.

Conclusion

There is some evidence of increasing economic concentration, and some evidence this concentration is more focused on big cities than in the recent past. However, we remain far less concentrated than even just a few decades ago, and the actual evidence of this trend is weak, and may be better explained by omitted variables, like the educational and sectoral composition of cities. It may not be agglomeration that is driving growth, but the presence of universities, or of a strong oil & gas hub, or something like that. Nonetheless, it does seem that population and economic activity are evincing two slightly different trends: even as evidence of population concentration is weak, evidence of economic concentration is somewhat smaller. Mathematically, this means that regional differences are almost certainly exacerbating nation-wide economic inequality. And that is concerning.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.