Does the Secular Stagnation hypothesis match the data? Evidence from the USA

Abstract

This paper focusses on Secular Stagnation in the United States. It answers to two research questions. First: is Secular Stagnation a fact? Second: conditional to the previous reply, how does the literature meet the qualitative and quantitative evidence? I focus on historical annual data related to the USA from 1870 onwards. The contribution to the debate is manifold. Firstly, I provide a careful systematization of the concept and explain why this definition is crucial for the topic analysis. Secondly, I talk about Secular Stagnation in terms of labor and multifactor productivity growth: their decline since the 1970s is not comparable to any previous shortfall. Thirdly, the definition allows to evaluate whether present-day debates on the phenomenon are supported by data, but it also lets to disentangle the most relevant channels through which Secular Stagnation emerges. Therefore, I carry out a comparative validation exercise on the different lines of thought. Finally, as long as Secular Stagnation is confirmed as empirical evidence, also policy implications can be drawn from the different theoretical frameworks. Despite the several contrasting approaches and viewpoints, I trace out a complementary in policy implications.

Keywords:

• Secular Stagnation
• negative interest rates
• GDP and productivity slowdown in growth
• stagnation policy

JEL Classifications:

• E20
• E43
• E60
• O11
• O30
• O40

Disclosure statement

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

Notes

1 “He saw the concept as rooted in J. S. Mill's notion of the stationary state, suggesting that the term “mature economy” described Mill's formulation of the stationary state as a low-investment but high-consumption economy. However, unlike Mill's stationary state, Hansen's secular stagnation featured chronic unemployment'' (Backhouse and Boianovsky 2016, p. 951; italics in original).

2 I restrict the analysis to the United States from 1870 onwards for two reasons. Firstly, the literature on Secular Stagnation focusses most on the US economy. Secondly, the economic science has identified the USA as the world’s technological frontier since early XX century (Pagano and Sbracia 2014). Additionally, I smooth time series with the Hodrick-Prescott (HP) filter to focus on the long-run and trend component only. I recognize that thinking the cyclical and trend components as additive is a very simplifying hypothesis. Yet, although long criticised, the performance of the HP filter was recently positively re-assessed in Franke et al. (2022).

3 I set the time windows of the peak-to-peak averages as in Gordon (2010) to ease the comparison. This holds for all subsequent tables too. Moreover, tables include some data on the COVID-19 period. We should handle this information with great care: the long-period character of this analysis should not focus too much on a too-recent phenomenon whose economic consequences are still very unclear.

4 GDP per capita growth can be broken down into the sum of two components: the growth of GDP per hours worked and the growth of labour utilisation, i.e., per capita hours worked. GDP per capita is a reliable indicator for productivity only to the extent that labour utilisation is constant through time.

5 I have carried out Bai and Perron structural break tests for each variable of interest. Results are available upon request.

6 Author’s estimates do not consider the composition adjustments that correspond to the aggregation of different components of capital and labour. Hence, no differences between ICT and non-ICT capital, and between skilled and unskilled workers. TFP is computed as the Solow residual of a standard Cobb-Douglas production function. The careful reader will recognise that TFP relies on, at least in the original formulation, the notion of exogenous technical progress. Technical progress is not exogenous at all (Dosi and Nelson 2010). Moreover, the very concept of TFP is controversial and I refer to Shaikh (1974) for further details. I would like to remark that I adopt TFP as a descriptive tool for it is hard and risky to detect evidence of Secular Stagnation by simply considering a unique macroeconomic indicator. I must rely on several instruments, imperfect and much disputed as they might be. Further details in the Appendix.

7 This evidence agrees with several contributions to the analysis of growth trajectories: see Maddison (2010).

8 The lack of historical data does not allow to argue whether the significant decline in potential output since the 1960s represents a back-to-normal dynamics as for GDP per capita. I limit myself to back up its overall decreasing pattern in growth.

9 The increasing number of working women and the entrance of the baby-boomers into the labour force from 1965 to 1990 pushed per-capita hours upward, despite the same period experienced a decrease in labour productivity. Ramey (2020) provides further support to this argument.

10 From this point of view, Schumpeter better understood what was going on for his theories reflected more successfully the technological spirit of the first part of the XX century (Schumpeter 1942; Field 2006).

11 Summers refers to the Wicksellian interest rate, the one “at which the demand for loan capital and the supply of savings exactly agree, and which more or less corresponds to the expected yield on the newly created capital” (Wicksell 1935, p. 193; italics in original). The idea of a negative natural rate is not new in economics, as was already mentioned by Klein in a discussion with Pigou on Hansen’s work (Backhouse and Boianovsky 2016).

12 By the way, implicit is the belief that investments are a positive function of the discrepancy between the profit rate and the interest rate. Yet, this discrepancy has widened since the 1970s, questioning further this view (Bresser-Pereira 2019; Storm 2017, 2022).

13 Di Bucchianico (2020) moves further criticism to the negative natural rate view along similar lines.

14 “Americans owe $1.2 trillion in college debt, and an increased fraction of the next generation may choose not to complete college as they are priced out of the market for higher education” (Gordon 2015, p. 57).

15 Supply-side influences of income distribution have been analysed also by heterodox and non-neoclassical schools of thought. Barrales-Ruiz et al. (2022), Kiefer et al. (2020), and Rada, Schiavone, and von Arnim (2022), Rada et al. (2023), among the others, focus on the effects of decreasing labour share and workers bargaining power on potential output dynamics. In particular, Kiefer et al. (2021) provides a new measure of potential output growth derived from a specification that builds on the interactions between the wage share and the rate of capacity utilization. With a focus on the US economy, the results clearly confirm a positive long-run relationship between economic growth and wage share, although no explicit causal linkage is emphasized. Anyway, we should remind the empirical evidence suggesting that the current slowdown of output growth resembles what US capitalism experienced in the seven decades before the Golden Age itself. In other words, the US GDP growth rate has averaged around 2 percent since late nineteenth century, with the exception represented by Golden-Age years.

16 The fall of the investment share did not occur, cf. Table 6. A different issue is the reallocation of funds from long-run physical investments to short-term financial assets. More on that below.

17 However, this is an old argument by David (2007) that explained the low TFP growth of the 1980s and early 1990s. Whether the same argument still holds today, after almost 40 years of “re-organization”, is something to be examined with great care.

18 Hein (2016), among the others, prefers talking about Stagnation Policy than Secular Stagnation. Engaging in a dispute on proper labeling is beyond the scope of the present work; anyway, my focus concerns to Secular Stagnation as a precise stylized fact, while Stagnation Policy is about the rationales behind it. Furthermore, it is important to note that Bresser-Pereira (2019) does not identify the period after the 1970 as Secular Stagnation sensu stricto, but only as a period with low growth and general uneasiness. Interestingly enough, Bresser-Pereira (2019, 21) defines this phenomenon as a “condition of negligible or no economic growth in a market-based economy”, connecting this lexicon to the long-term decline in productivity growth occurred in rich countries, very much alike the definition given throughout the present work.

19 Table 6 shows the decrease of physical investments growth since the 1950s and the constancy of the investment share in GDP. This suggests, by the way, that the composition of the capital stock has shifted toward short-lived assets. The endogeneity of GDP can shed light on both occurrences: debt-led consumption compensated for the negative effects on consumption expenditure and income multiplier caused by the reduction of the wage share. This enabled the investment-to-GDP ratio to stay constant even though private investment growth declined in the same period.

20 “In the US the overall trade union membership was reduced substantially from 1980 to 2017; it declined from 20.1% in 1983 […] to 11.9% in 2010. In 2014, the union membership rate was 11.1%, down 0.2 percentage point from 2013; in June 2016 it was 10.7%, which is half of what it was in 1980” (Arestis, 2021; 359).

21 Globalisation certainly played a key role. The integration of China and India in the global economy doubled the available workforce worldwide. The huge increase in the labour supply helped shift the bargaining power to the benefits of capitalists and corporations since too many workers were available for not-enough jobs (Akyuz, 2018).

22 Heterodox and non-neoclassical theories often assume the positive influence of an increase in the profit share on investments. This topic is beyond the scope of the paper. Yet, it is important to note that the empirical evidence on this point is rather uncertain (Onaran et al. 2011). Other demand-side determinants are more important (Girardi and Pariboni 2020).

23 We should admit that this vision of the productivity growth process relies on quite a schematic representation of the mechanisms that drive technical change. The latter seems to occur in a deterministic and automatic way without a proper conception of the process behind. Llerena and Lorentz (2004a, 2004b) and more recently Borsato and Lorentz (2023a, 2023b) contribute to filling this gap by matching the Post-Keynesian literature with the evolutionary Schumpeterian approach à la Dosi (1982, 1988) and Nelson and Winter (1982).

Additional information

Funding

This work was supported by Università degli Studi di Siena [PhD Grant (2017-2020)].

Notes on contributors

Andrea Borsato

Andrea Borsato is a Postdoctoral Researcher at CNRS BETA, University of Strasbourg, University of Lorraine, Strasbourg, France