Assessing the losses in euro area potential productivity due to the financial crisis

Abstract

In this article, we show that the recent financial crisis has significantly affected the potential total factor productivity (TFP) of the four largest euro area economies, as well as that of the rest of the euro area. We used a reduced-form equation of TFP, based on an approach recently developed by Cahn and Saint-Guilhem (2010). Our empirical findings show that the permanent impact on potential TFP varies across countries from –3.9 points to –1.3 points in Q2 2012. When these losses are incorporated, TFP gaps develop closely in line with capacity utilization rates (CUR). Moreover, in the case of France, including CUR in our TFP model improves the quasi-real-time reliability of TFP gap estimates.

Keywords:

• production function
• total factor productivity
• financial crisis
• capacity utilization

JEL Classification:

• E22
• E23
• E32
• O47

Acknowledgements

We are grateful to Annabelle Mourougane and seminar participants at the AFSE meeting 2013 and the Banque de France seminar for their useful comments and suggestions. The views expressed in this article do not necessarily reflect those of Banque de France.

Notes

¹ Excluding the financial crisis of 2007–2008.

² Intuitively represents the remaining share in the available capital stock of capital of vintage j.

³ This transformation does not change their elasticity (it only changes the estimated intercept). For the CUR, it allows us to disentangle the cyclical part related to the TFP gap from the long-run reference value of the CUR (measured by its mean) related to the potential TFP. For the age of capital, this transformation would be useful if we were interested in a long-run potential TFP, which would also exclude fluctuations in the age of capital.

⁴ If we relax the constraints relating to the coefficients of Equation 2′ by regressing the TFP on an intercept, its lag and lags of all regressors (ecur, ecage, the time trend, breaks and the qualitative indicator), we get a linear equation and we can apply the Bai–Perron test. Another possible approach would have been to test break dates with the nonlinear equation by using the test of Boldea and Hall (2013).

⁵ Since this break date is exogenous, testing its significance does not require tests of endogenous break dates (Bai–Perron or Boldea–Hall tests) and we simply use the standard distribution of the Student test.

⁶ L is the lag operator.

⁷ K_t = KM_t + KB_t with KM_t = (1–0.024)*KM_t–1 + IM_t and KB_t = (1–0.004)*KB_t–1 + IB_t, at each date t.

⁸ Historical data on investment are derived from Maddison (1994) databases for France and Germany, from Prados (2003) for Spain and from Baffigi (2011) for Italy.

⁹ We have performed robustness tests on residuals which suggest homoscedasticity and nonautocorrelation. These additional results are available upon request from the authors.

¹⁰ We use the terminology of Orphanides and Van Norden (2002), who distinguish the quasi-real-time reliability from the real-time one: the quasi-real-time reliability is only based on revision related to the sample availability, while the real-time reliability also takes into account data revisions.