Friday, August 22, 2014

The Great Unskewing

One way to think about the dynamics of jobs and firms is to examine the distribution of firm growth rates. The high pace of job reallocation we observe implies that firms vary considerably in employment growth rates; some firms are shrinking (or closing their doors), some are operating year after year without changing the size of their workforce, and some are rapidly expanding. We basically know about the first moment of this distribution (the mean or median), which must be slightly positive to accommodate growth of the workforce over time. The second and third moments--the width of the distribution and the degree to which it is skewed up by high-growth firms--are measures of dynamism.

Suppose we sliced the distribution of firms by firm age. The figure below is from our recent JEP paper (click for larger image):

Figure from Decker, Haltiwanger, Jarmin, and Miranda (2014a)

Here you can see the Geico "everybody knows that" fact about young firms "accounting for" aggregate job growth. As a cohort of firms ages, its growth rate distribution converges on zero--so old firms as a group do not create jobs in aggregate (more on cohort dynamics here and here). The second moment can be seen here too: young firms vary widely, but firm cohorts see their growth rate distribution narrow as they age (but note that the distribution never becomes degenerate--there is still lots of churning among old firms). In this sense, old firms are less dynamic than young ones. Finally, observe that young firms are skewed toward high growth. The gap between the median firm and the firm at the 90th percentile is larger than the gap between the median firm and the firm at the 10th percentile.

As part of the general aging of firms (though not entirely because of it), the aggregate firm distribution increasingly looks like the distribution of old firms. The next figure (from a current working paper) plots the gap between the 90th percentile and the 10th percentile firm, for all firms and for only continuing firms (i.e., no entry or exit):

Click for larger image
Figure from Decker, Haltiwanger, Jarmin, and Miranda (2014b)

Observe that this measure of the second moment--the dispersion of firm outcomes--has been falling, so the firm distribution has been tightening up. This might seem like a good thing, particularly if the tightening is happening because fewer firms are shrinking. This is basically the reason that some people are skeptical about the need to be concerned about declining dynamism. Who cares if dispersion is falling? A lot of that churning is unnecessary anyway. The next three figures examine the third moment--skew--by comparing the gap between the 90th and 50th percentiles to the gap between the 50th and 10th percentiles.

Figures from Decker, Haltiwanger, Jarmin, and Miranda (2014b)

While it is true that the bottom of the firm distribution is pulling up, the top is falling down more quickly. The US economy is unskewing, beginning to look more and more like the old timers. When we look at all firms in aggregate or just at the high-tech sector, it looks like the distribution could be totally symmetric soon. Among young firms, there is still a high degree of skew but it is falling (keep in mind that age-0 firms necessarily have positive growth rates, which gives the young firm distribution a boost).

The unskewing of the firm distribution means that we are seeing fewer high-growth firms (or, equivalently, the bar for what constitutes "high growth" has fallen). In partial equilibrium, at least, it's very hard to interpret that result as good news. It is likely to affect how the economy responds to the business cycle, and over time it could have consequences for productivity. For now these are speculative claims, but this is a growing literature. A new paper by Davis and Haltiwanger finds evidence that lower dynamism causes lower employment, not to mention slower wage growth. In my view this evidence greatly complicates the recession stories that dominate the econ blogosphere. It may not be a simple Econ 101 aggregate demand story.

Friday, August 8, 2014

Comment on Cowen

My comment on this post by Tyler Cowen (it never made it through moderation):

- I have been amazed at how many econ journalists and economist bloggers were not already aware of these trends. The Hathaway and Litan briefs have not revealed new facts; just pushed on them in some very helpful ways. Academics have known the basic facts (declining entrepreneurship, aging firms) for some time. The general decline in entrepreneurship was documented in a working paper that started circulating almost 3 years ago (here) and in various Kauffman briefs. Over a year ago I blogged about the aging of the firm distribution. That post had very little traffic even by this blog's standards, but I see now that it wasn't because the topic doesn't interest people.
- The Yglesias argument is just speculation. We've had a hard time finding evidence for it. The aggregate times series should make anyone skeptical. If he wants to push this story, he should use regional variation (for a start). 
- This isn't just about subsistence entrepreneurship, the "orange sellers" and mom-n-pop shops. At least starting around 2000, high-growth entrepreneurs took a hit too. 
- While it's true that the datasets on which these findings are based currently end in 2011, the establishment birth data we see in the BED (which extends through 2013q4) are suggestive (though not conclusive, since establishments aren't firms). Don't expect a massive rebound in new firm formation on the heels of the Great Recession.

The last two points are responses to other comments on the post. This is a good example of an issue on which the blogosphere (or, at least, the blogs with readers) and journalists are way behind academia, though there are a handful of journalists who have been on the case for several years.

Sunday, August 3, 2014

Entrepreneurial decline: Mom-n-pop or high-growth?

A lot of people respond to data on declining rates of entrepreneurship by suggesting that this is about the decline of mom-n-pop shops associated with retail sector consolidation. From a recent paper:

Most business startups exit within their first ten years, and most surviving young businesses do not grow but remain small. However, a small fraction of young firms exhibit very high growth and contribute substantially to job creation. 

If we're just seeing fewer slow growers and likely failures, and if that "small fraction" of young firms that accounts for so much job creation has been unaffected, then less entrepreneurship isn't alarming. Big firms pay better than small firms, and the rise of big box retail has probably made the economy more productive. So Noah Smith and others have pushed back against the growing chorus of declining dynamism alarmism.

In a preliminary, incomplete working paper, coauthors and I attempt to determine whether the decline of entrepreneurship has involved high-growth firms. One way to do this is to study the distribution of growth rates of young firms (i.e., firms age 5 or less). Figure 10 from that paper is below (click for larger image):

Figure from Decker, Haltiwanger, Jarmin, & Miranda (2014)
"The secular decline in business dynamism in the U.S."

The lines indicate the employment-weighted 90th percentile of employment growth rates. That is, only 10 percent of young firm employment is at firms with growth rates that exceed the red line (the series are actually HP filtered). For young firms, the growth rate required to be in the 90th percentile was somewhat constant during the 80s and 90s, suggesting that the decline in entrepreneurship we see in those decades may not have been affecting high-growth firms. But starting around 2000, the top of the growth rate distribution falls. We take this as suggestive evidence that high-growth entrepreneurship began declining around 2000. You can also see that mature firms had a fairly constant distribution before 2000, so that the overall trend was largely a composition effect.

So at least by blog standards I think it's fair to assume that the pre-2000 decline of entrepreneurship may have been largely about the death of the mom-n-pop, but since then even high-growth entrepreneurship has taken a hit. It turns out that the year 2000 is significant for other measures of dynamism, too. The high-tech sector, the information sector, finance, and publicly traded firms all had turning points around that time (see Figure 6 from the working paper), with flat or rising rates of job reallocation and within-firm volatility prior to ~2000 and declining dynamism after ~2000 (and dynamism trends aren't the only thing that changed around 2000; an example). So the story of declining dynamism in the US is partly a story of convergence, with roughly the year 2000 marking a point at which classes of firms that were previously unaffected by the aggregate patterns finally joined the trend.

Friday, August 1, 2014

"Economic dynamism and productivity growth"

Dietz Vollrath has a nice response to our work on dynamism. Basically his argument is that the decline in dynamism doesn't seem to be affecting aggregate productivity. Yesterday I posted a comment at his blog, but comment moderation over there seems to be on hiatus, so I'm reposting my comment here:

In the paper, we are pretty careful not to draw conclusions about whether this is a good thing or a bad thing, since (as you suggest) it’s unclear what is optimal in terms of young firm activity and job flows. A lot of journalists have looked at our work and related work and run with stories about doom and gloom, but we’re not ready to draw any kind of conclusions. 
Likewise, I think it’s a little premature to draw the conclusions you draw about productivity. First, just looking at the aggregate data without a plausible counterfactual is, as you know, pretty tricky. I know you’re aware of the evidence on productivity and reallocation (which we review in the paper); I think our prior, at least, should be that this matters. 
Second, the Fernald chart you include is not inconsistent with a connection between productivity and dynamism. Note the trend break in the early 2000s. This is consistent with the dynamism data in Figure 3 from our paper–observe that the dynamism trend is pretty flat during the 1990s. Moreover, as we show in some newer work, dynamism measures for, eg, the high tech sector and public firms actually increase until the early 2000s, at which point they head downward (see and the paper linked therein). It’s not unreasonable to assume that strong dynamism in key sectors helped boost productivity in the 1990s, and once those sectors experienced reversal in the early 2000s the aggregate trend moderated somewhat. 
As a side note, you mention retail. As we discuss in the paper, there is lots of evidence that consolidation in retail has been productivity enhancing. So, certainly, there are countervailing trends at work.

Wednesday, July 30, 2014

The role of entrepreneurship in US job creation and economic dynamism

That is the title of my first quarter-of-a-publication, in the Journal of Economic Perspectives. The issue includes symposia on entrepreneurship, development, and academic production, as well as some other content. JEP articles are always ungated, courtesy of the AEA. I suspect that the excellent Timothy Taylor will do a full write-up of the issue soon. This particular journal is meant for generalists, and as such the papers tend to be largely nontechnical and digestible (and interesting!).

Our article has been in circulation as a working paper in various versions for maybe 2 or 3 years, and during that time the general topic of declining entrepreneurship and dynamism generally has become pretty widely known outside of academia. I have blogged on this stuff too many times to link. We're also already working on a follow-up study that looks at things like high-growth entrepreneurship, differences between public and private firms, some specific sectors, and trends in the nature of "shocks" hitting firms. I've blogged a bit about that newer paper, here and here; it is still in early stages.

In the JEP paper, we describe some data on the dynamics of young firms, how the growth rate distribution of firm cohorts evolves as they age, the role of young firms in productivity growth (see also here), and some long-term trends that are pretty widely known now. We do some simple accounting exercises to determine the degree to which composition effects are driving long-term trends in gross job flows. Some basic insights and findings:

- New firms experience a strong "up or out" dynamic--a few grow very quickly and survive, while the rest shrink and fail (see Figure 1 in the paper, below, click for larger version). As such, many of the jobs created by startups are destroyed in short order. This is pretty well known in this literature.

I do not own this image

- The growth rate distribution of young firms is highly skewed, with some firms growing very quickly and pulling up the mean. Among older firms, the growth rate distribution is symmetric with a mean and median of zero (see Figure 2 in the paper).

- Startups, and reallocation more generally, play a huge role in productivity growth. We discuss this in some detail, and I covered it a bit here; we really just review existing research.

- In shift-share analysis, the aging of the firm distribution "accounts" for about one third of the decline in gross job flows. Changing industry composition (away from manufacturing and toward retail and services) works the "wrong way", since we have moved toward more activity in more volatile industries. When we absorb age, size, and industry composition effects, we "explain" about 15 percent of the decline (note, though, that this is not causal analysis). This means that the decline is happening within cells, and a good explanation for it has yet to be found. As such, policy implications of what we know right now are unclear.

- The decline in dynamism is relentless, indefatigable, indisputable, and undeniable (Yorke 2006), and it is ubiquitous across industry and geography. This suggests that simple policy explanations may not get us very far.

- We conceptualize the question in terms of standard models of firm dynamics, which would suggest that a decline of this kind means either (a) a decline in the volatility of shocks that drive firm outcomes, or (b) a decline in the responsiveness of firms to these shocks (which could be driven by, e.g., technology or policy changes). Our newer working paper sheds some light on this problem.

We write,

We do not yet fully understand the causes of the decline in indicators of business dynamism and entrepreneurship, nor in turn, their consequences. Improving our understanding of the causes and consequences should be a high priority. . . .  
The declining pace of startups, job creation, and job destruction is mirrored in other measures of the dynamism of American society. . . . Taken together, there appears to be less scope for the US economy to adjust to changing economic conditions through the migration of workers, the reallocation of jobs across producers, and through the switching of workers across a given allocation of jobs.

The paper is reasonably short, nontechnical, and (I think) focused enough to be worth looking through. A lot of this literature consists of papers where you drink through a fire hose of data, but here we've tried hard to be concise (thanks in large part to excellent editors). We started with dozens of figures and tables and whittled down to just a few. When I first encountered the firm dynamics literature, I was blown away by the richness and diversity of market economies that shows up in the administrative micro data. Hopefully this paper will get others thinking about the topic.

I'm very excited about this paper. It builds a lot on work that has been done by people other than me, primarily including my coauthors John Haltiwanger, Ron Jarmin, and Javier Miranda, but also Stephen Davis, Lucia Foster, Chad Syverson, and others whom are listed at the end of the text. These people, along with others like Erik Hurst, have done and are doing a lot of really interesting work in empirical firm dynamics. In my view this is the best stuff happening in macro these days, as it utilizes large amounts of micro data on firms and establishments to explore big macroeconomic questions. For my involvement in this project I thank my generous coauthors and a series of consecutive luck shocks.

BED: 7.3 million jobs created, 6.5 million destroyed in Q4 2013

From the BLS:

From September 2013 to December 2013, gross job gains from opening and expanding private sector establishments were 7.3 million, an increase of 290,000 jobs from the previous quarter. . . . Gross job losses from closing and contracting private sector establishments were 6.5 million, a decrease of 34,000 jobs from the previous quarter.

I like this data series, with some caveats.* If you're not familiar with this series, note that gross flows are large relative to net flows. Roughly speaking, think of the Great Recession as involving about 8.5 million net job losses. Entering and expanding business establishments create at least half that many jobs even in terrible quarters, but a recession is characterized by even larger numbers of jobs being destroying by shrinking or closing establishments.

I like to slice the data by extensive margin (opening or closing business establishments) and intensive margin (expanding or contracting business establishments). Figure 1 reports the flows of employment associated with opening and closing establishments, and Figure 2 reports actual numbers of establishments that opened or closed (click for larger images).

Figure 1

Figure 2
New establishments continue to boost net employment, keeping positive job flows ahead of closures for several quarters in a row. The latest quarter shows a slight uptick in gross flows of establishments, which many might consider to be a positive sign. But in general, total reallocation on the establishment extensive margin has been pretty constant since the end of the recession.

Next, the intensive margin. Figure 3 reports employment flows from expanding and contracting establishments, and Figure 4 reports establishment counts for these categories (click for larger images).

Figure 3

Figure 4

I would say that these numbers look promising. Establishment growth is helping drive employment recovery, and shrinking establishments are providing less "drag" over time. Total reallocation seems to be trending upward as well, so in my view these data look reasonably good.

Overall, this is a pretty good report.

Now some usual thoughts: gross flows give us an idea of where jobs are being created and destroyed, which fleshes out the net job numbers that are more popular (and timely). For policymakers, it matters whether job market problems are being driven by establishment turnover or job flows in existing establishments. In my (hasty) view, these latest numbers suggest that both margins are firing reasonably well, which was less the case a few quarters ago. Further, my prior is that the slight upward trend in total reallocation among continuing establishments is a good sign and may boost productivity somewhat.

More broadly, these data help dissuade us from thinking in representative agent terms, which is what the net numbers incline people to do. It's tempting to think that net numbers tell us about the experience of most businesses, but in reality there is a lot of heterogeneity among firms and reallocation proceeds at a high pace. In my view this complicates macro analysis somewhat, rendering simple "aggregate demand/supply" heuristics somewhat tricky.

Some previous BED posts are here.

*The BED are quarterly data provided from the BLS based on state UI data. They are released with a lag of about 8 months. Like the BDS (the dataset I usually use here), the BED basically covers the universe of private nonfarm employers; unlike the BDS, the BED is available at higher frequency and is released more quickly. BED has other drawbacks compared to the BDS, such as a more limited ability to track firms.

The BLS effectively expanded the sample definition in the first quarter of 2013, and it does not appear that they have done anything to fix the time series. This is very unfortunate as it limits the usefulness of looking at time series in ways that are difficult to fully grasp. The 2013q1 observation was the most obviously affected, as it reported all establishments that were added to the sample as establishment openings. For openings data, I have replaced the 2013q1 observation with the average of 2012q4 and 2013q2. I haven't dug into the data enough to know whether users can manually correct for this over the longer run. See BLS discussion here, on the bottom of the page ("Administrative Change Affecting..."). Please, BLS, do something about these time series.

It is also important to note that these numbers are seasonally adjusted, and any guess at net numbers based on the difference between two seasonally adjusted series is very, very rough. Non-SA numbers are available on the BLS website.

These numbers track business establishments, which are different from firms. Costco is a firm; your local Costco store is an establishment. Most firms consist of only one establishment. The BED is not ideal for tracking firms, as it has limited ability to correctly link establishments to the firm level.

Tuesday, July 29, 2014

Exceptional industries

Six-digit NAICS (2002) codes that include the word "except" in their titles, with the warning that industry code humor is subtle and beyond the reach of most:

  • 111120 Oilseed (except soybean) farming
  • 111219 Other vegetable (except potato) and melon farming
  • 111320 Citrus (except orange) groves
  • 111334 Berry (except strawberry) farming
  • 115114 Postharvest crop activities (except cotton ginning)
  • 213115 Support activities for nonmetallic minerals (except fuels) mining
  • 236115 New single-family housing construction (except operative builders)
  • 236116 New multifamily housing construction (except operative builders)
  • 311611 Animal (except poultry) slaughtering
  • 313312 Textile and fabric finishing (except broadwoven fabric) mills
  • 315223 Men's and boys' cut and sew shirt (except work shirt) manufacturing
  • 316213 Men's footwear (except athletic) manufacturing
  • 316214 Women's footwear (except athletic) manufacturing
  • 316993 Personal leather good (except women's handbag and purse) manufacturing
  • 321213 Engineered wood member (except truss) manufacturing
  • 322121 Paper (except newsprint) mills
  • 325414 Biological product (except diagnostic) manufacturing
  • 326113 Unlaminated plastics film and sheet (except packaging) manufacturing
  • 326130 Laminated plastics plate, sheet (except packaging), and shape manufacturing
  • 326150 Urethane and other foam product (except polystyrene) manufacturing
  • 326211 Tire manufacturing (except retreading)
  • 331419 Primary smelting and refining of nonferrous metal (except copper and aluminum)
  • 331422 Copper wire (except mechanical) drawing
  • 331491 Nonferrous metal (except copper and aluminum) rolling, drawing, and extruding
  • 331492: Secondary smelting, refining, and alloying of nonferrous metal (except copper and aluminum)
  • 331513 Steel foundries (except investment)
  • 331522 Nonferrous (except aluminum) die-casting foundries
  • 331524 Aluminum foundries (except die-casting)
  • 331525 Copper foundries (except die-casting)
  • 331528 Other nonferrous foundries (except die-casting)
  • 332211 Cutlery and flatware (except precious) manufacturing
  • 332211 Cutlery and flatware (except precious) manufacturing
  • 332812 Metal coating, engraving (except jewelry and silverware), and allied services to manufacturers
  • 332993 Ammunition (except small arms) manufacturing
  • 333414 Heating equipment (except warm air furnaces) manufacturing
  • 333997 scale and balance (except laboratory) manufacturing)
  • 334612 Prerecorded compact disc (except software), tape, and record reproducing
  • 336330 Motor vehicle steering and suspension components (except spring) manufacturing
  • 337125 Household furniture (except wood and metal) manufacturing
  • 337214 Office furniture (except wood) manufacturing
  • 339911 Jewelry (except costume) manufacturing
  • 423810 Construction and mining (except oil well) machinery and equipment merchant wholesalers
  • 423860 Transportation equipment and supplies (except motor vehicle) merchant wholesalers
  • 424430 Dairy product (except dried or canned) merchant wholesalers
  • 424720 Petroleum and petroleum products merchant wholesalers (except bulk stations and terminals)
  • 423810 Construction and mining (except oil well) machinery and equipment merchant wholesalers)
  • 423860 Transportation equipment and supplies (except motor vehicle) merchant wholesalers)
  • 424430 Dairy product (except dried or canned) merchant wholesalers
  • 445110 Supermarkets and other grocery (except convenience) stores
  • 452111 Department stores (except discount department stores)
  • 453998 All other miscellaneous store retailers (except tobacco stores)
  • 484220 Specialized freight (except used goods) trucking, local
  • 484230 Specialized freight (except used goods) trucking, long-distance
  • 512131 Motion picture theaters (except drive-ins)
  • 524128 Other direct insurance (except life, health, and medical) carriers
  • 531120 Lessors of nonresidential buildings (except miniwarehouses)
  • 533110 Lessors of nonfinancial intangible assets (except copyrighted works)
  • 541370 Surveying and mapping (except geophysical) services
  • 561621 Security systems services (except locksmiths)
  • 621111 Offices of physicians (except mental health specialists)
  • 621330 Offices of mental health practitioners (except physicians)
  • 622310 Specialty (except psychiatric and substance abuse) hospitals
  • 713210 Casinos (except casino hotels)
  • 721110 Hotels (except casino hotels) and motels
  • 721214 Recreational and vacation camps (except campgrounds)
  • 811310 Commercial and industrial machinery and equipment (except automotive and electronic) repair and maintenance
  • 812320 Drycleaning and laundry services (except coin-operated)
  • 812910 Pet care (except veterinary) services
  • 812921 Photofinishing laboratories (except one-hour)
  • 813990 Other similar organizations (except business, professional, labor, and political organizations)
  • 923130 Administration of human resource programs (except education, public health, and veterans' affairs programs)

So, for example, there are two kinds of vegetables & melons: potatoes, and other vegetables & melons. There are two kinds of hotels/motels: casino hotels, and the others. There are two kinds of foam products: polystyrene, and the others. The metals examples are instructive, demonstrating the value of iron/steel, aluminum, and copper in manufacturing.

Now go impress people at parties.