Abstract
Secularly declining GDP investment shares are often explained by the widespread fall in the relative price of investment goods. Granger non-causality tests applied to longer-term time series for a large number of industrial countries tend to reject that explanation.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The concept of ‘secular stagnation’ has a fairly long tradition in economics (as recently documented e.g. by Backhouse and Boianovsky 2015). Canuto et al. (2014) review some of the current views on secular stagnation. A more extensive presentation of the opinions held by prominent contributors to the debate is collected in a recent VoxEU volume edited by Teulings and Baldwin (2014). Conferences, featuring prominent speakers, have abounded since addressing the issue. The contributions to the IMF’s and the Austrian National Bank’s conferences on the topic (both held in June 2015) are accessible under: http://www.imf.org/external/np/seminars/eng/2015/secularstag/index.htm; http://www.oenb.at/en/Publications/Economics/Economics-Conference/2015/economics-conference-2015.html.
More recent reviews include e.g. Gourinchas et al. (2016) and Hall (2017).
The secular stagnation cannot be blamed on depletion of natural resources. If anything, natural resources have become less scarce. This is reflected in the secular decline in the relative prices of raw materials (see e.g. Mollick et al. 2008).
This is well exemplified by US data. From 1965 through 1982 the US relative price of investment goods rose by 0.8% per year. Since then that price has been declining, on average by 1.4% per year.
The nominal investment share divided by the relative price index (investment goods over all GDP) gives the ratio of real investment over real GDP. Both items tend to follow declining trends, at least since the early 1990s. For example, the German (nominal) GDP investment share fell from 24.9% in 1991 to 19.9% in 2015. The ‘real investment ratio’ (calculated at 2010 prices) fell from 21.2 to 19.8% accordingly. In macroeconomic analyses the (nominal) investment share has a well-defined meaning (inter alia as a counterpart of the national saving rate) and plays a role in e.g. the calculation of the contributions of fixed capital formation to real GDP growth rates. In contrast, the operational meaning of the ‘real investment ratio’ seems more elusive, at least to the present author. Anyway, the overall conclusions on the causal effects of relative prices on the investment share generally agree with those for the ‘real ratio’ (see Footnote 7). Incidentally, the fact that the decline in the ‘real investment ratio’ is associated with the declining relative price of capital goods might suggest that the latter may constitute a sort of ‘Giffen commodity’.
Granger causality is understood as follows: Assume one considers two stationary time series X and Z. X is said to Granger-cause Z if Z can be better predicted using the histories of both X and Z than it can by using the history of Z alone. The same applies to X being Granger-caused by Z. Absence of Granger causality can be tested by estimating the Vector Auto regression (VAR) model with two equations:
\( Z_{t} = a_{0} + a_{1} Z_{t - 1} + \cdots + a_{p} Z_{t - p} + b_{1} X_{t - 1} + \cdots + b_{p} X_{t - p} + u_{t} \)
\( X_{t} = c_{0} + c_{1} Z_{t - 1} + \cdots + c_{p} Z_{t - p} + d_{1} X_{t - 1} + \cdots + d_{p} X_{t - p} + v_{t} \)
The parameters ‘a’ and ‘b’ remain to be estimated; u and v are error terms. Ho: b1 = b2 = ···bp = 0 is a hypothesis that X does not Granger-cause Z. Similarly, Ho: c1 = c2 = ···cp = 0 is a hypothesis that Z does not Granger-cause X. Testing Ho is in terms of the usual Wald test statistics. The results may (and often do) depend on the number of lags (p) taken. That number can be selected on the basis of so-called information criteria (such as Schwarz’s or, alternatively, Akaike’s) and paying attention to the additional properties of the VAR model (its stability and absence of residual autocorrelation).
The P values were derived from auxiliary VAR models. In each case the number of lags in the auxiliary VARs was determined on the basis of the usual Akaike Information Criterion. In addition, the VARs underlying the probabilities reported are stable and free of residual serial autocorrelation.
Interestingly, substituting the ratio of real investment to real GDP for the nominal investment share does not change the conclusions materially. The hypothesis on the ‘real investment ratio’ not Granger-causing the relative investment price is rejected, at 5% significance level, in 11 cases (including Germany, UK, Canada and US). The hypothesis on the relative price not Granger-causing the real investment ratio is rejected in 7 cases including Turkey, Bulgaria, Denmark, Estonia, Greece, France and Spain.
The Toda–Yamamoto procedure proceeds in steps. The first is the selection of the lag length (p) for the VAR (in levels, not in their first differences). The selection is guided by the information criteria (Schwarz etc.) and the properties of the VAR (stability, absence of residual autocorrelation). The second step requires the determination of the order of integration of the original series considered (e.g. via ADF tests). Suppose the maximum of these orders is m (i.e. one of the series is integrated of order m, the other is integrated of order not greater than m). Third, the VAR (in levels) with (p + m) lags is estimated. Finally, one runs the standard Wald test that the coefficients of only the first p lagged values of X are zero in the equation for Z and does the same for the coefficients of the lagged values of Z in the equation for X. Rejection of the null implies rejection of Granger non-causality (i.e. delivers evidence of Granger causality). The probabilities reported in Tables 3 and 4 were derived from the auxiliary VARs with m = 1.
Of course, investment goods may have become cheaper also on account of supply-side developments (including falling relative production costs in the sectors supplying the investment goods and services, e.g. construction services). Observe though that the hypothesis of primary interest here is about the causation running from relative investment prices to investment/GDP shares.
References
Backhouse RE, Boianovsky M (2015) Secular stagnation: the history of macroeconomic Heresy. Blanquie Lecture, Rome
Buiter W, Rahbari E, Seydl J (2014) The Long-Run Decline in Advanced-Economy Investment. citi Research, 11 June
Canuto O, Nallari R, Griffith B (2014) Sluggish postcrisis growth: policies, secular stagnation and outlook. The World Bank Economic Premise, no. 139
Gordon JR (2015) Secular stagnation: a supply-side view. Am Econ Rev Pap Proc 105(5):54–59
Gourinchas PO, Portes R, Rabanal P (2016) Secular stagnation, growth, and real interest rates. IMF Econ Rev 64(4):575–580
Hall RE (2017) The Anatomy of stagnation in a modern economy. Economica 84(1):1–15
Mollick AV, Faria JR, Albuquerque PH, Leon-Lendesma MA (2008) Can globalisation stop the decline in commodities’ terms of trade? Camb J Econ 32(5):683–701
Podkaminer L (2015) The euro area’s secular stagnation and what can be done about it. A post-Keynesian perspective. Real World Econ Rev 70:1–15
Podkaminer L (2017) The slowdown in labour productivity growth is an effect of economic stagnation rather than its cause. Acta Oecon 67:67–77
Rachel L, Smith TD (2015) Secular drivers of the global real interest rate. Bank of England staff working paper no. 571
Teulings C, Baldwin R (2014) Secular stagnation: facts, causes, and cures, VoxEU.org eBook
Thwaites G (2015) Why are real interest rates so low? Secular stagnation and the relative price of investment goods. Bank of England Staff Working Paper No. 564
Toda HY, Yamamoto T (1995) Statistical inference in vector autoregressions with possibly integrated processes. J Econom 66(1–2):225–250
Acknowledgements
The author thanks two anonymous referees for helpful comments.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Podkaminer, L. The decline in investment shares is not caused by falling relative prices of capital: a note. Empirica 46, 369–380 (2019). https://doi.org/10.1007/s10663-018-9401-2
Published:
Issue Date:
DOI: https://doi.org/10.1007/s10663-018-9401-2