Serguey Braguinsky and Lee Branstetter are from Carnegie Mellon University. Andre Regateiro is portuguese, from IST, working with both of them in the School of Public Policy and Management, under the Carnegie Mellon Portugal Program. Together they have published last year an elegant paper with a strong title: “The incredible shrinking portuguese firm” (2011), in the prestigious NBER working paper series. It is an outcome of the CMU Portugal program, involving both international scholars in the analysis of portuguese problems, and portuguese PhD students in top research teams. And the paper has been cited in The Economist (3 March 2012). Definitely this is good.
The authors claim that Portugal firm size distribution shows a pronounced shift for the left, at least during last two decades. They claim this shift not to exist in other advanced industrial countries, and that it cannot be fully ascribed to expanding data coverage or other “natural causes” (like the shift from industry to services along the same period of time). They believe this is a Portugal unusual and distinctive shift, and they claim to be the first paper to document “this surprising change”. And they also claim that it reflects the unusual and distinctive labor market regime in Portugal, that shrinks firms and lowers aggregate productivity of the economy.
Well - we may have some problems.
First, the so called “unique” (and “incredible”) portuguese shrinking firm does not seem to be so unique (and not so incredible). See, for instance, our previous post about Choi and Spletzer (2010). Braguinsky et al (2011) do not cite this paper - although Chou and Spletzer do cite them (in a subtle footnote). So the shrinking firm does not seem to be so specific to Portugal: during last decade, it happened both in Portugal and US - and may be also in other countries. In next figure we plot data from both papers during last decade: size of firms index, rebased to 2001=100. Portugal declive is more pronounced than US one.
But Braguinsky et al do not control for age of the firms. No age cohorts have been extracted from the data (although that may be feasible). We have seen before (here and here) how important this is and how misleading it can be to ignore them. Are the companies smaller and smaller because they start smaller and / or they grow less and less?
Second, Braguinsky et al paper argues that the reason for smaller firms is due to restrictive regulations of labor market, a negative incentive to growth of portuguese firms. They show that a “shift to the left” in size distribution is compatible with the impact of legal thresholds of firm size on their analytical models. So their hypothesis has not been refuted - what, of course, does not means it is true (remember Karl Popper).
The authors present a list of laws supporting small firms (and, consequently, assumed to be disincentives for large firms). But it seems the list (annex B of their paper) only includes entries in the regulation, from 1983 to 2005, not exits. Some of the laws cited have meanwhile been revoked along the period, or they have been superseded.
The minimum we can say is that the authors have ignored, or discarded, other possible reasons for the shift, without any justification. For instance, the technology hypothesis previously advanced by Chou and Spletzer (2010) for the US economy.
We have our own "grounded" ideas about the impact of labor restrictions on growth of firms in Portugal. I have around twenty five years of “field experience” in top management of companies in Portugal, from small to medium and to very large manufacturing firms listed in stock market, exactly during the period of time of this analysis. I agree that labor laws do indeed condition the operational flexibility of firms in Portugal, but in practice they are more onerous and difficult for small companies than to large companies. Labor regulations probably make an important contribution to early mortality of small companies. But I do not feel, based on my personal experience of management in different types of firms, that labor regulations have been the main obstacle to growth of firms in Portugal during that period of time. Problems of financing, yes, very probably. Geography, far from large consumer markets, it may also be (in some cases). What my experience tells me is that easy access to financing and lack of non banking sources of financing (venture capital, business angels, large capital markets), or difficult access to them, may be a much more more important reason “not to grow” (and probably to die earlier).
The “up or out” model of Haltiwanger et al (1910) can be useful if applied here. The technology reasons for the shrinking firm could be very credible indeed. We can easily construct a narrative, parallel to the legal one suggested by Braguinsky et al in their annex B, about the “incredible” update of technologies used by portugueses business firms during last decades, after Portugal joined the European Union (1986). Several generations of programs of european funds (from PEDIP to QREN) helped an huge renew and update of technologies and business infrastructures, especially in manufacturing firms. Sure, this did not go without consequences.
Labor restrictions hypothesis can be very attractive inn the actual political climate. And the authors explicitly recognize this when they say that “given the difficult choices facing Portugal’s new government, the message of this paper may prove to be a timely one”. This statement is a surprising one in an academic paper. Of course, ultra liberal politicians will be very happy to claim that it is now “scientifically proved” that labor laws are the reason for portuguese lower productivity. But Braguinsky et al have not proved it - they have claimed it and suggested theoretical considerations to support a similar shrinking effect due to labor regulations.
By the way, is it reasonable to apply Braguinsky et al claim to USA, and to claim restrictive labor regulations during last decade, dominated by Bush rightist policies, to explain the shrinking american firm? I do not think so.
Without controlling for age cohorts and considering alternative sources to explain the variation in the data, their paper may be elegant, but it is far from convincing.
Annex:
To be fair, full data plot (see next figure) shows different medium term trends in Portugal and USA, suggesting different mechanisms may be involved in the variation of the size of firms in different periods of time.
Age cohorts are fundamental to understand dynamic behaviour in business demography, we discussed in previous post. A paper by Haltiwanger, Jarmin and Miranda (2010), published as a NBER working paper ("Who creates jobs? Small vs. large vs. young"), deals with the old controversy of job creation by small versus old firms. Their main finding is that once we control for firm age (through age cohorts), there is no systematic relationship between firm size and growth. Controlling for age had a dramatic impact on the analysis.
The high importance of small firms in job creation is a very popular and “politically correct idea” between policy makers. But once again it is misleading.
Haltiwanger and colleagues used a large set of data, from 1976 to 2005 (equivalent to 70 million firm-year observations) and they show that small companies are not the job creators. However start-ups, new entries, make important contributions to job creation - and indeed also to job destruction during their (usually) short life. But net job creation by start ups makes a significant contribution to net employment.
Typically a new entry (start up) begins small and grows. This paper gives evidence for a “up and out” dynamic of young firms: new firms begin small, and they grow or they disappear (exit). They either grow fast or they disappear. But in manufacturing sectors the exit pattern during early stages is much stronger.
From young to mature, along time (age), usually firms go from small to larger - may be medium size and, in some (rare) cases, they become very large corporations.
In early stages, firms are small and they have small, incremental, effects on job creation and destruction. But lots of them can create a sizable effect. While they are small, even small absolute changes in size can give high relative rates of change. Usually they are financially weak and dependent of external sources of funds (venture capital, equity). They often have an high R&D intensity, but their share of total R&D activities is marginal.
Later, in the mature phase, if they survive, firms are larger, their relative rates of change are smaller, but their share of total employment is larger, their contribution for job destruction is also larger when they collapse, and they make substantial contributions to R&D activities (even if their R&D intensity is smaller).
It is easy to chart these important changes in business behavior as firms go from young / small to mature / large. But starting point also influences firm behavior (path dependency): technologies used and related operational / organizational aspects are different for a start up today and for a start up, in a similar business and environment, twenty years ago (for instance).
Mortality is high for small firms, including start ups in early stages: a robust finding from the authors is that small firms are more likely to exit than large firms, even controlling for age.
Business start ups have a small contribution for total employment in any given year: roughly 3%. But these new jobs become significant in the overall net changes in employment, from job creation and destruction. They account for almost 20% of gross job creation.
Young firms exhibit high rates of job creation and destruction. They have very high job destruction rates from exit (closing of firms). After five years, around 40% of the jobs initially created by start ups have been eliminated by exit of the firms. But if the firm survives this initial and testing period, young (then small) firms will grow more rapidly than their more mature counterparts. Anyway - large, mature firms account for a large fraction of the jobs.
The authors make some relevant conclusions:
- We think our findings help interpret the popular perception of the role of small businesses as job creators in a manner that is consistent with theories that highlight the role of business formation, experimentation, selection and learning as important features of the U.S. economy. Viewed from this perspective, the role of business startups and young firms is part of an ongoing dynamic of U.S. businesses that needs to be accurately tracked and measured on an ongoing basis.
- Our findings suggest that policies targeting firms based on size without taking account of the role firm age are unlikely to have the desired impact on overall job creation.
- It may be, for example, that the volatility and apparent experimentation of young businesses that we have identified is critical for the development of new products and processes that are in turn used by (and perhaps acquired by) the large and mature businesses that account for most economic activity.
Our own conclusions:
- Analyzing dynamic data about firms by isolation of age cohorts gives new insights. Overlapping of different dynamic behaviours (firms of different ages) can create a misleading “confounding” effect (well known in statistics) that can easily induce wrong conclusions from time series data about firms.
- Young companies role is very important in economy change (innovation), so is entrepreneurship. But young companies do not mean young people. Young companies by mature people are specially important (see previous post).
- Small and large firms have different roles in the innovation ecosystem (as predicted by William Baumol, and also recognized by Joseph Schumpeter in his later phase).
Choi and Spletzer (2011) ("The declining average size of establishments: evidence and explanations") have shown a declining trend in the average size of US firms and establishments during last decade, after a reverse of trend during 2000/2001 (from 1993 to 2001 the trend was crescent, and also long term trend from 1977).
The declining trend occurs in almost all industries and can be explained by the age of the firms and establishments. They found that new establishments and start ups are smaller: they start small and they stay smaller hen they mature.
Although not driven by a particular industry, the decline in manufacturing size of firms had a substantially larger effect than in other industries.
They advance the hypothesis that new entries (births) adopt new modes of production, more based on technology and less based on labor. Industries more technology intensive really show a more declining trend.
An important point of their paper is the critical role of age cohorts in order to understand the evolution of size of firms along time.
Their figure 6B shows two important things:
- establishments growth along time: initial size was around 7 for entries during 2001, but ten years later their average size was around 14 (a compounded annual rate of growth around 8%)
- establishments start smaller and smaller: average initial size was around 4.7 in 2010, versus around 7 ten years earlier, in 2000.
It is interesting to compare with their figure 6A, size by age, not age cohorts: it shows a declining trend for units of the same age, along time. This means that three years old units were around 12 people in mid 90’s, but less than 8 in 2011. Figure 6A is a consequence of overlapping of different curves from figure 6B.
