Firm Dynamics and Growth Measurement in France
GAICIL Insights
If the US statistical blind spot were an American quirk, it could be dismissed. This paper shows France has it too, and in some ways worse.1 Firm Dynamics and Growth Measurement in France, by Philippe Aghion, Antonin Bergeaud, Timo Boppart and Simon Bunel, applies to France a methodology developed for the United States to estimate missing growth from creative destruction. When products disappear, statistical agencies often impute inflation from surviving products, understating growth if new products improve more than survivors.
The estimates are substantial. From 2004 to 2015, about 0.5 percentage point of real output growth per year is missed by the statistical office, about the same as in the United States. In absolute terms, missing growth from imputation is slightly lower in France than in the US, implying that, if anything, the growth differential between the two countries over the past fifteen years has been slightly underestimated. Yet in France missing growth is closer to one third of total true productivity growth, a larger share than in the US. A third of real progress, invisible in the official record.
Geography sharpens the picture. Missing growth varies across French sectors and regions and is higher where creative destruction and measured growth are higher, suggesting differences in geographic economic dynamism may have been underestimated. Similar aggregate estimates for France and the United States also conceal noticeable differences in plant dynamics, with different market shares for entrants and exiters. The most dynamic places look less dynamic than they are.
The pattern travels. Official statistics systematically understate growth generated by creative destruction, in both the United States and France at similar magnitudes, suggesting the bias may be widespread across advanced economies. A shared blind spot in national accounts shapes every debate built on them.
Where does this leave AI? Artificial intelligence will widen the blind spot. AI is expected to generate significant creative destruction as new products and services replace existing ones. If statistical methods understate improvements embodied in new products, AI's contribution to growth may be underestimated. Accurate measurement is essential for evaluating AI's economic impact and designing policy.
The regional finding is the most policy-relevant. Missing growth concentrates where creative destruction and measured growth are highest. Dynamic regions and sectors, the places where AI is likely to have the greatest impact, may be those where statistics understate growth most. Policymakers relying on official statistics may underestimate these areas' dynamism, and may allocate support accordingly. A region-by-region AI strategy built on biased numbers will tilt against the leaders.
Cross-country comparison earns its keep. Similar aggregates can hide different underlying dynamics, such as differences in entry and exit, which help explain how economies generate growth and how policy can support it. For AI, cross-country comparisons can reveal how institutions affect diffusion and the creative destruction it generates.
The policy implication is consistent. If true growth exceeds measured growth, returns to innovation may be larger than they appear, strengthening the case for policies supporting innovation and creative destruction. Statistical agencies should develop methods better capturing new products' value, a task more urgent as AI accelerates product turnover. National statistics offices are part of the AI policy apparatus whether they know it or not.
Plant-level data make it possible. Tracking surviving, entering and exiting plants allows a transparent estimate of missing growth. Investment in detailed microdata enables such analysis and improves understanding of the economy. Microdata infrastructure is cheap relative to the policy errors it prevents.
Read as a whole, creative destruction generates growth that official statistics partly miss, about half a percentage point per year in France as in the United States. The missing growth concentrates in the most dynamic sectors and regions. As AI drives new waves of creative destruction, accurate measurement will be essential for recognizing its contribution and supporting the innovation on which future growth depends.
One question to put to European policymakers comparing themselves unfavorably with the US: part of the gap may be measurement, and part of Europe's own dynamism may be hidden. Fix the statistics before drawing conclusions from them.
Introduction
In this paper we use the same methodology as Aghion et al. (2017a) to compute missing growth estimates from creative destruction in France. We find that from 2004 to 2015, about 0.5 percentage point of real output growth per year is missed by the statistical office, which is about the same as what was found in the United States. We look at how missing growth varies across French sectors and regions, and we look at the underlying establishment and firm dynamics. In particular we show that the similar missing growth estimates between France and the United States hide noticeable differences in plant dynamics between the two countries.
Takeaway
In this paper we used plant-level information to compute missing growth estimates for France. We first found that in absolute terms missing growth from imputation is slightly lower in France than in the U.S.: this implies that, if anything, the growth differential between the U.S. and France over the past fifteen years has been (slightly) underestimated by the statistics. Second, while in the U.S. between a third and a fourth of true productivity growth was missed according to ABBKL, in France missing growth from imputation is closer to one third of total true productivity growth between 2004 and 2015. Third, we found that missing growth is higher in French sectors or regions with higher rates of creative destruction and with higher measured growth rates, suggesting that geographical economic dynamism differences could have been underestimated in the case of France. Fourth, we found that the similarity between French and U.S. in terms of missing growth hides differences in the market shares of new entrants and exiters between those two countries: both the employment share of new entrants and that of exiters are higher in France than in the U.S., but the differences in these employment shares between the two countries (almost) cancel each other out when computing missing growth estimates. The analysis in this paper suggests several avenues for future work.1
This paper includes replication of the model developed in ABBKL which could be easily extended to other countries. Preliminary results in the UK and Japan show that missing growth seem to lie within the same range in these countries, and similar computations are ongoing in Italy and Sweden. Another extension would be to look at how the French Statistical Office could improve its measurement of productivity growth. Finally, our results have implications for growth, fiscal and labor market policies and their impact on growth. For example minimum wage policy uses estimates of the yearly level of inflation. Should the fact that inflation is overestimated, lead us to revisit minimum wage policy?
Notes
- Philippe Aghion et al., Firm Dynamics and Growth Measurement in France, 16 J. Eur. Econ. Ass'n 933 (2018), https://doi.org/10.1093/jeea/jvy031. Takeaway source: Conclusion, reproduced from a free author or repository copy (working-paper version). ↩
Cite as: Philippe Aghion et al., Firm Dynamics and Growth Measurement in France, 16 J. Eur. Econ. Ass'n 933 (2018), https://doi.org/10.1093/jeea/jvy031
Header image: DSC02328 Vordernbergerstraße Leoben at the Steel Mill, 2023-08.jpg by Aciarium, CC BY 4.0, via Wikimedia Commons.
