Is There a Declining Trend in Capacity Utilization in the US Economy? A Technical Note

ABSTRACT

Recent contributions have mentioned the possibility of a declining trend in capacity utilization in the US since the 1970s. However, no consensus has emerged on the empirical evidence. The aim of this paper is to identify if such a declining trend in capacity utilization exists in the US economy: New empirical evidence is shown confirming that this is the case, at least since 1989.

KEYWORDS:

• Capacity utilization
• growth
• growth models
• federal reserve board

JEL CLASSIFICATION:

• c22
• d24
• e11
• e22

Acknowledgments

The author would like to thank three anonymous referees for their constructive criticism. The author would also like to thank Professor Riccardo Pariboni, Matteo Deleidi and Alejandro González for their valuable comments and advice on a very early draft, Ryan Decker (FRB) and Justin Pierce (FRB) for their assistance in using the databases employed in this paper. Remaining errors are the sole responsibility of the author.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 A substantial element of the first step of this debate was continued in the journal ‘Political Economy - Studies in the Surplus Approach’ (http://www.centrosraffa.org/politicalEconomy.aspx) during the 1980s.

2 Board of Governors of the Federal Reserve System (US), Capacity Utilisation: Manufacturing [CAPUTLB00004SQ], quarterly, seasonally adjusted, retrieved from FRED, Federal Reserve Bank of St. Louis; https://fred.stlouisfed.org/series/CAPUTLB00004SQ

3 In Appendix A.1. some econometric tests are presented to prove this claim.

4 See Nikiforos (2019b) for a critical review of this measure.

5 Source: https://www.federalreserve.gov/Releases/g17/Meth/MethCap.htm and https://www.federalreserve.gov/pubs/Bulletin/2000/0300secnd.pdf

6 Nikiforos might be labelled under a Neo-Kaleckian framework, in which the utilization rate is endogenous and responsive to demand pressures in the long-term. See Nikiforos (2013, 2016, 2018, 2019a) for theoretical and empirical evidence on this issue, Girardi and Pariboni (2019) and Gahn and González (2019a) for critical remarks.

7 https://www2.census.gov/programs-surveys/qpc/technical-documentation/questionnaires/watermark_form.pdf?#

8 Shaikh might be labelled as a classical Marxist/Harrodian scholar, who argues that there is a tendency for firms to obtain a normal rate of capacity utilization in the long-run equilibrium. In this view, Keynesian results and policy conclusions apply in the short-run, but, apparently, not in the long-run.

9 We are not claiming here that McGraw-Hill and FRB measures are the same, but both share a survey-based estimation method and therefore, this critique is common to both.

10 https://www2.census.gov/programs-surveys/qpc/technical-documentation/questionnaires/watermark_form.pdf?#

11 The question of the survey has been changing through time (see Doyle 2000; Morin and Stevens 2004; Nikiforos 2016 and Fiebiger 2018).

12 https://www2.census.gov/programs-surveys/qpc/technical-documentation/questionnaires/watermark_form.pdf?#

13 The ‘engineer’ concept of capacity utilization Y/Y^∗ implies a notion in which the denominator expresses the maximum technical possibilities of the plant or firm, even if this is not profitable.

14 https://www.census.gov/programs-surveys/qpc/data/tables.html

15 In Appendix A.2., we show that we cannot reject the null hypothesis of both series being equal (different only in levels).

16 Even better than the Average Workweek of Capital (Foss 1963; Taubman and Gottschalk 1971; Foss 1981a; Foss 1981b; Foss 1984; Foss 1985; Shapiro 1986; Orr 1989; Mayshar and Solon 1993; Shapiro 1996; Foss 1997; Beaulieu and Mattey 1998; Gorodnichenko and Shapiro 2011) because NER takes into account, at least partially, the speed of operation.

17 This adjustment is also performed by the US Census using the same methodology.

18 Some econometric evidence is also shown in Appendix A.2.

19 Only mentioned in a few articles (Shapiro 1996; Morin and Stevens 2004; Bansak, Morin, and Starr 2007; Gorodnichenko and Shapiro 2011) and briefly analysed in Doyle (2000) and Petri (2003, 2004). See Belzer et al. (1991, 1993) for estimations.

20 One of the referees raised the issue that the presence of a declining trend in capacity utilization might imply that the latter cannot be considered an exogenous variable in the long-run, as some versions of the Neo-Kaleckian model suggest. It is not the purpose of this article to provide a thorough investigation of the theoretical implications for growth theory of the empirical findings reported here. This note is meant to provide a preliminary technical assessment on a declining trend of capacity utilization in the US economy. Nonetheless, it is interesting to notice few things: First, the fact that one variable is considered exogenous does not mean that it cannot change, but that it is relatively persistent; for the method to be useful it is enough that the speed of change of the endogenous variables is of a higher order than the rate of change of the exogenous magnitudes; the given variables may then be assumed to be fixed for the purpose of explaining the long-period position, even though they may change slowly over time (Dvoskin and Lazzarini 2013, 118, emphasis added in italics). Finally, it can be envisaged at least two scenarios in which a declining trend in utilization is compatible with its exogeneity with respect to growth: a) If the process of adjustment towards normal utilization is very slow and b) if any of the exogenous determinants of the normal rate is changing. In the end, the evidence presented in this paper it is not in itself enough to accept or reject any model.

21 If bounds were not sufficiently far away, the analysis must consider this issue (see Cavaliere and Xu 2014).

22 Thanks to Alejandro González who raised this issue.

23 See Appendix A.3. for data description.

24 These tests were performed without deterministic trends in order to be useful for the Appendix A.2.1., A.2.2. and A.2.3. Cointegration analysis was discarded because of the small number of observations.

25 Given that we cannot reject the presence of a unit root process, for the first sub-sample we differentiate the data so the equation to be tested is: $\mathrm{\Delta }\mathrm{N}\mathrm{E}{\mathrm{R}}_{\mathrm{t}}={\gamma }_1\mathrm{\Delta }\mathrm{N}\mathrm{E}{\mathrm{R}}_{\mathrm{t}-1}+\dots +{\gamma }_{\mathrm{n}}\mathrm{\Delta }\mathrm{N}\mathrm{E}{\mathrm{R}}_{\mathrm{t}-\mathrm{n}}+{\text{β}}_1\mathrm{\Delta }\mathrm{F}\mathrm{U}{\mathrm{R}}_{\mathrm{t}-1}+\dots +{\text{β}}_{\mathrm{n}}\mathrm{\Delta }\mathrm{F}\mathrm{U}{\mathrm{R}}_{\mathrm{t}-\mathrm{n}}+ϵ$(2)

The same procedure will follow for all of Appendix A.2.

26 FUR: only last quarter, not s.a.; FRB: only last quarter, s.a.

27 NER: only last quarter, not s.a.; FRB: only last quarter, s.a.