Minsky Theory of Inflation: An Empirical Analysis of OECD Countries

Abstract

This article investigates the dynamics of the four-decade-long disinflationary trend combined with cyclical inflation in OEDC countries. Minsky’s inflation model is adopted as the analytical framework and its relevance is demonstrated for modern consumption-driven economies that have gone through the globalization of the production process and the weakening of workers’ bargaining power since the 1980s. We find that unit labor cost and demand channels through household, business, and government spending are significant in explaining inflation in twenty-eight OECD countries during both the post-Bretton Woods period and the Great Moderation period. We also find evidence suggesting that central banks are not capable of achieving price stability either with their key instrument or by exploiting the trade-off between unemployment and inflation. Overall, our results contend that a more effective strategy would be demand management via fiscal policy or credit guidance and an income policy that directly affects labor bargaining power.

Keywords:

• inflation
• household saving
• wage
• panel data

JEL Classification Codes:

• E31
• E21
• E25
• C23

Disclosure Statement

No potential conflict of interest was reported by the author.

Notes

1 Minsky adopted Kalecki’s (1954) representation of profits. Minsky (2008) also defined profits as “the difference between sales revenues and technologically determined costs.” This means that wages of nonproductive (technologically irrelevant or ancillary/overhead) workers (even in consumption-good industries) are considered as profits. It implies that a higher proportion of nonproduction workers to total employment will lead to a higher markup due to higher demand from those incomes and higher expenses related to ancillary activities (i.e., research, development, sales, marketing, and finance), putting upward pressure on prices.

2 For more detail on the contrast between Minsky’s inflation theory and neoclassical/post-Keynesian models, see Kim (2021).

3 See Brown 2007; Cynamon and Fazzari 2008 and 2013; Keen 2009; Glick and Lansing 2010; Mian and Sufi 2010 and 2018; Eggertsson and Krugman 2012; Andersen et al. 2014; Bunn and Roston 2015; Kalantzis 2015; Ryoo 2016; Scott and Pressman 2019.

4 In a related article, Kim (2020) studies the empirical impact of household credit on U.S. inflation.

5 Core instead of headline inflation was chosen because the latter has many outliers and more noise, as indicated in table 1. However, as a robustness test, results are also presented for the CPI headline inflation in table 3.

6 “There are several reasons why inflation is time-dependent. First, if prices are set in a forward-looking manner under conditions of nominal rigidity, it is optimal for firms to set higher prices in advance when they rationally expect the aggregate price level to rise. Second, inflation inertia arises under conditions of indexation. This is observed when wages and prices of goods and services (most frequently, but not exclusively, prices of non-tradables like public utilities, home rents, and other services) are indexed to past inflation” (Calderón and Schmidt-Hebbel 2008).

7 We chose annual data instead of quarterly data because the quality of quarterly data is still poor for many OECD countries and quarterly data are sometimes constructed artificially from annual data (see also, e.g., Giannone, Reichlin, and Small 2008). Moreover, using annual data instead of quarterly data has the potential advantage that we take out the short-term cycles from the data and focus on medium to long term cycles which may be more relevant for inflation and its determinants. Last, quarterly data for the import price index and the household saving ratio are simply not available for most OECD countries.

8 In our baseline specification, the number of lags has been chosen to be equal to two, but the results are robust to the choice of another lag length from 0 to 4.

9 Estimation results are available from the author upon request. Furthermore, we replaced the growth rate of import prices with other alternatives to account for the changes globalization brought to the domestic inflation process. The relevant literature suggests three alternative variables: (1) international oil prices (West Texas Intermediate and Brent crude oil), proposed by Cologni and Manera (2008) and Wen, Zhang, and Gong (2021); (2) a trade openness measure, used by Temple (2002), Gruben and McLeod (2004), and Borio and Filardo (2007); and (3) exchange rates adopted by Shambaugh (2008), Comunale and Kunovac (2017), and Forbes, Hjortsoe, and Nenova (2018). We found some evidence that trade openness and exchange rates contribute to inflation with statistical significance, although the results are not robust to different estimation methods. On the other hand, oil prices do not seem to have the expected impact on inflation dynamics. Our results for the unit labor cost, the investment ratio and the saving ratio are, nevertheless, robust to whichever alternative is used for import prices. Estimation results for the alternative variables for import prices are also available from the author upon request.

10 For example, Calderón and Schmidt-Hebbel (2008) have a Phillips curve-based model estimated over OECD countries with R squared = 0.75 at its highest, which is smaller than our Minskyian baseline model with R squared = 0.86.

11 The OECD’s calculation of NAIRUs is not only based on a Kalman filter procedure but also adjusted by policy reforms, especially in Europe.

12 Data for NAIRU and thus the unemployment gap is available from 1984 so the total sample starts from 1984.

13 The positive coefficients of unemployment are not surprising to heterodox economists, such as Anwar Shaikh. Shaikh (2016) argues that high unemployment can be associated with high inflation when the capacity-utilization rate falls (which leads to high unemployment) slower than purchasing power (demand pressure).

14 Estimation results are available from the author upon request.

15 Data for structural fiscal balances is available from 1979 so the total sample starts from 1979.

16 Estimation results that exclude each country for a robustness test are available from the authors upon request. We also re-estimate separate regressions on G-7 countries (Canada, France, Germany, Italy, Japan, the UK, and the United States) and on non-G-7 countries to explore whether our results are dominated by highly advanced economies, but it did not alter estimation results.

Additional information

Notes on contributors

Hongkil Kim

Hongkil Kim is in the Department of Economics, Georgia State University, Atlanta, Georgia.