A global race to the bottom: The neo-Goodwinian aggregative-systems estimation of income distribution and capacity utilization interactions

Vechsuruck, Tanadej Article A global race to the bottom: The neo-Goodwinian aggregative-systems estimation of income distribution and capacity utilization interactions PSL Quarterly Review Provided in Cooperation with: Associazione Economia civile, Rome Suggested Citation: Vechsuruck, Tanadej (2024) : A global race to the bottom: The neo-Goodwinian aggregative-systems estimation of income distribution and capacity utilization interactions, PSL Quarterly Review, ISSN 2037-3643, Associazione Economia civile, Rome, Vol. 77, Iss. 308, pp. 59-87, https://doi.org/10.13133/2037-3643/18186 This Version is available at: https://hdl.handle.net/10419/324101 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by-nc-nd/4.0/ PSL Quarterly Review This work is licensed under a Creative Commons Attribution – Non-Commercial – No Derivatives 4.0 International License. To view a copy of this license visit http://creativecommons.org/licenses/by-nc-nd/4.0/ vol. 77 n. 308 (March 2024) A global race to the bottom: The neo-Goodwinian aggregative-systems estimation of income distribution and capacity utilization interactions TANADEJ VECHSURUCK* Abstract: The neoliberal reforms since the 1980s have resulted in rapid globalization paralleled by worsening income distribution. In this paper, I first show that most countries worldwide (58 of 81) have experienced a decline in the labor share of income, or the wage share, during 1950-2019. Second, I estimate the demand and distributive regimes from 81-country panel data based on the neo-Goodwinian model. At the global level, the short-run estimation shows that the distributive regime appears to be Marxian/profit-squeeze and the demand regime exhibits profit-led. I further separate the estimation into two groups: advanced and developing countries. The estimation still confirms the profit-led/profit-squeeze regimes in both groups, even though the demand and distributive regimes are stronger in advanced economies. In the long run, the results reveal a global race to the bottom: a decline in the long-run wage share. Neither positive nor negative gain is founded on capacity utilization in advanced and developing countries. University of Rhode Island, Kingston, USA, email: [email protected] How to cite this article: Vechsuruck T. (2024), “A global race to the bottom: The neoGoodwinian aggregative-systems estimation of income distribution and capacity utilization interactions”, PSL Quarterly Review, 77 (308), pp. 59-87. DOI: https://doi.org/10.13133/2037-3643/18186 JEL codes: C23, E12, E25, O47, O57 Keywords: Panel analysis, post-Keynesian, neo-Goodwinian model, distributive-demand dynamics, globalization Journal homepage: http: //www.pslquarterlyreview.info Income distribution has become one of the hottest economic topics in the past few decades (Atkinson, 2015; Galbraith, 2012; Milanovic, 2005, and 2016). Piketty (2014) revived the interest in functional income distribution (the shares of income between labor and capital) 1 among mainstream economists. Several scholars on both the mainstream and the heterodox sides, including Piketty himself, found that the labor share of income or the wage share has fallen across the world (UNCTAD, 2012; Karabarbounis and Neiman, 2014; IMF, 2017; Dao et al., 2019; Suzuki et al., 2019; Autor et al., 2020; Paul, 2020; Stansbury and Summers, 2020). This phenomenon contrasts with one of Kaldor’s stylized facts (Kaldor, 1957), that the share between labor and capital should be constant across time. Although the main culprit of this labor share decline is still * This paper is developed from the first chapter of Vechsuruck (2018), the author’s PhD dissertation. The author wants to thank the managing editor and two anonymous referees for their valuable comments and suggestions. The usual disclaimer applies. 1 Functional income distribution is separated into the income share of labor (labor share or wage share) and the income share of capital (capital share or profit share). The sum of both shares is always 1 (or 100%). For instance, if the wage share equals 60% (or 0.6), the profit share is 40% (or 0.4). Note that the wage share, the labor share, the labor income share, and the labor share of income are the same. All terms will be used interchangeably in this paper. Articles 60 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review inconclusive, the possible drivers include rapid globalization in trade and finance, automation, financialization, and welfare state retrenchment (Amsden and Hoeven, 1996; Crotty et al., 1998; Jayadev, 2007; Onaran, 2009; Onaran and Galanis, 2012; Stockhammer, 2013). The characteristics of dynamics between aggregate demand and income distribution have become one of the main research questions in post-Keynesian economics. In contrast to neoclassical economists, post-Keynesian economists believe that a change in income distribution impacts aggregate demand. The neo-Goodwinian model (Barbosa-Filho and Taylor, 2006), one of the post-Keynesian macroeconomic business cycle models, considers the economy to be composed of two regimes: the demand regime and the distributive regime. The demand regime captures the causal effect from income distribution to effective demand. In a closed economy without a government, the demand regime is considered profit-led 2 when a positive effect of a higher profit share on investment dominates the negative effect of a higher profit share on consumption. The opposite case is called wage-led, when the positive effect of a higher wage share on consumption is significant enough to offset the negative effect of a higher wage share on investment. The distributive regime, on the other hand, captures the causal effect from effective demand to income distribution. The distributive regime could act in two main ways. If increasing economic activity hurts the wage share, we say the distributive regime is wage-squeeze, forcedsaving, or Kaldorian. 3 However, if higher demand, which usually causes lower unemployment, lifts the bargaining power of labor and leads to a rising wage share, the distributive regime is labeled as profit-squeeze or Marxian. In this paper, based on the neo-Goodwinian model, I examine the interactions between income distribution, in terms of the wage share, and capacity utilization, in terms of the output gap, along the lines of the aggregative-systems approach (see below). To the author’s knowledge, this is the first panel data analysis of a neo-Goodwinian model that examines both advanced and developing countries. The primary dataset is the latest Penn World Table (PWT) 10.0. I created an unbalanced panel dataset of the wage share and output gap for 81 countries (47 developing and 34 advanced) covering 1950-2019, to investigate the global dynamics of output and distribution in the short and the long run. First, comparing the first five years and the last five years of data available for each country, I find that 58 of 81 countries experienced a decline in their wage shares. Second, I apply the neo-Goodwinian framework to test the unbalanced panel data. Several recent studies relied on the conventional Hodrick-Presscott filter to obtain the potential output and the output gap. However, the filter has been exposed to criticism, as it may create spurious cycles (Cogley and Nason, 1995) or force the long-term output gap to be zero (Blecker, 2016), ruling out the variations in the output gaps (see appendix B for more details on the filters). I therefore use five filters to estimate the potential output and obtain the output gap: the Hodrick-Presscott (HP), the Baxter-King frequency (band-pass), the moving average, the Beveridge-Nelson (BN), and the Hamilton. Different filters are used to avoid selection bias and ensure the robustness of the results. Using the panel data econometric regression based on the standard seemingly unrelated regression (SUR) model, I find that, in the short run, the global distributive regime is profit-squeeze or Marxist and the global demand regime is profit-led. The results are robust across different output gaps. I also estimate the panel and allow for coefficients 2 The profit-led/wage-led definition was coined by Taylor (1991); the meaning is comparable to the terms in Bhaduri and Marglin (1990), exhilarationist/stagnationist, respectively. 3 Kaldor (1956) hypothesized that an economy needs to shift distribution in favor of capitalists during the booming period, so they can have sufficient funds to make investments in the following period. Marx, on the other hand, emphasized the role of the reserve army of unemployed, which makes labor's bargaining power, and thus the real wage, vary procyclically. T. Vechsuruck 61 varying between two groups of countries: advanced and developing countries. In both groups, the demand and distributive regimes are still profit-led/profit-squeeze. However, both regimes are stronger in advanced economies than in emerging economies. Lastly, the long-run estimation reveals no long-term gain or loss on the output gap, but the wage share has a long-term decline. The decline in the wage share is more perverse in developing countries. These results suggest that, although countries implement wage repression for short-term benefits, there are unclear gains in the long run. The global race to the bottom results in a long-term decline in only the labor share. The structure of this study is as follows: After the introduction, the theoretical model section explains the construction of the neo-Goodwinian model. The subsequent section, empirical analysis, is separated into four subsections. The first and second subsections explore the wage share and the output gap data. The third subsection explains the econometric models on which this study is based and analyzes the findings. The fourth subsection considers the implications of the results. Finally, the last part summarizes the essence of this work. 1. The Neo-Goodwinian model Following Marx’s idea on social conflict and the Lotka-Volterra mathematical model on the competition between species, Goodwin (1967) constructed a macroeconomic model to show that a business cycle can be explained by two endogenous variables: income distribution, or the predator, and employment, or the prey. Based on the literature from Keynes, Kalecki, and Steindl, post-Keynesian works, including Bhaduri and Marglin (1990), Dutt (1984), Taylor (1991), Rowthorn (1981), Foley and Michl (1999), and Blecker (1989), among others, have developed a Keynesian economic growth theory in which effective demand is emphasized as crucial to economic growth. The importance of income distribution is revived along the lines of classical economics dating back to Smith, Ricardo, and Marx. The neo-Goodwinian model presented below closely follows the business cycle model developed in Barbosa-Filho (2001), Barbosa-Filho and Taylor (2006), and Taylor (2004), in which income distribution and capacity utilization are endogenized. 4 In particular, this heterodox business cycle model incorporates effective demand into social conflict to examine how income distribution interacts with business fluctuations. The concept of distribution-demand interactions has been further scrutinized in a nonlinear fashion (Tavani et al., 2011; Nikiforos and Foley, 2012). Suppose a closed economy produces only one good and a government has no role. A society is divided into two classes: capitalists, whose income is mainly derived from profit, and workers, whose income is mainly derived from wage. Capacity utilization (𝑢) is defined as real output (𝑋) over existing capital or potential output (𝐾). Wage share (𝜓) is defined as real wage (ω) over labor productivity (𝜉). By differentiating 𝑢 and 𝜓 with respect to time (given that 𝑥=𝑥󰇗 𝑥 when 𝑥 is continually differentiable), we have: 𝑢=𝑋 −𝐾  (1) ψ =ω−ξ󰆹 (2) 4 This model was originally called the structuralist Goodwin model (Barbosa-Filho and Taylor, 2006). Stockhammer (2017) and Blecker and Setterfield (2019) later popularized the titles the neo-Goodwin cycles and the neo-Goodwinian model. Note that, in this paper, they are all the same. 62 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review The growth rate of utilization relies on the difference between the growth rates of output and capital (suppose there is no capital depreciation), whereas the growth rate of the wage share is the difference between real wage growth and labor productivity growth. The relationship between utilization and wage share can be further scrutinized using the Two-Species Model (Shone, 2002, chapter 14) since they can be considered two species demonstrating either rivalry or predation. Output, capital, real wage, and labor productivity are constructed as a linear function consisting of our two species, utilization and wage share, as follows: 𝑋 =𝛼0+𝛼𝑢𝑢+𝛼𝜓𝜓 (3) 𝐾 =𝛽0+𝛽𝑢𝑢+𝛽𝜓𝜓 (4) 𝜔=𝛾0+𝛾𝑢𝑢+𝛾𝜓𝜓 (5) 𝜉󰆹=𝛿0+𝛿𝑢𝑢+𝛿𝜓𝜓 (6) Theoretical foundations justify the signs of all coefficients αj,βj,γj,δj (see appendix A for more details). Then equation (3) and equation (4) are substituted into equation (1). Equation (5) and equation (6) are substituted into equation (2). Also, let ϕj=αj−βj and θj=γj−δj for j = 0,u or ψ. We can obtain: 𝑢󰇗 =𝑢(𝜙0+𝜙𝑢𝑢+𝜙𝜓𝜓) (7) 𝜓󰇗=𝜓(𝜃0+𝜃𝑢𝑢+𝜃𝜓𝜓) (8) Equation (7) and equation (8) can be constructed as the utilization nullcline and distributive nullcline, respectively, after they are equated to zero. The slopes of both nullclines, the stationary solution, the long-run solution, and the stability analysis are elaborated in appendix A. The slope of the utilization nullcline depends on the sign of 𝜙𝜓 or the difference between the effects of wage share changes on output and capital. When the sign of 𝜙𝜓 is positive, the utilization nullcline is positively sloped or the demand regime is wage-led. The negative 𝜙𝜓 causes the utilization nullcline to be negatively sloped, or the demand regime is profit-led. The slope of the distributive nullcline, on the other hand, largely rests on the sign of 𝜃𝑢 or the difference between the effects of utilization changes on real wage and labor productivity. The positive 𝜃𝑢 results in a positively sloped distributive nullcline, or the distributive regime is profit-squeeze. The distributive regime is considered as wage-squeeze when 𝜃𝑢 is negative, which causes the slope of the distributive nullcline to be negative as well. Figure 1a illustrates the system with profit-squeeze distributive and profit-led utilization regimes. The system manifests counterclockwise predator-prey dynamics, where the wage share is a predator, and capacity utilization is the prey. At the beginning of the business cycle, a reduction in the wage share induces a higher investment that overshadows a fall in consumption. An increase in capacity utilization strengthens labor’s bargaining power, eventually leading to a higher labor share, which would set the stage for an economic slowdown, ending the cycle. The new cycle will start when a lowering wage share stimulates aggregate demand again. The dynamic behavior can be characterized as spiral sink as it converges to the long-run steady state. In this system, a prolabor distributive shock, or a leftward shift of the distributive nullcline, will improve T. Vechsuruck 63 the wage share while worsening utilization. A positive demand shock, or a rightward shift of the utilization schedule, will improve wage share and utilization in this system. Wage-led and wage-squeeze dynamics regimes are shown in figure 1b. The system exhibits clockwise predator-prey dynamics, where a predator is instead performed by capacity utilization while the wage share turns out to be the prey. The business cycle starts when an increase in the wage share boosts the economy. Higher consumption is large enough to compensate for a reduction in investment. The economy expands until the profit share starts to increase. The period of recession stalls the economy until the wage share rises again, and the new cycle begins. Likewise, the dynamic behavior is still considered as spiral sink. Nevertheless, a prolabor shock in this system will improve both labor share and utilization, whereas a positive demand shock will improve only utilization but discourage the labor share. In figure 1c, the distributive curve shows the forced-saving/Kaldorian characteristic of the profit-led demand regime. In this case, the distributive schedule must cut the demand schedule from above to make the system stable (Taylor, 2004). In other words, the distributive schedule must be steeper than the demand schedule to have a positive determinant for the Jacobian matrix (see appendix A). As a result, the system also embraces counterclockwise predator-prey, spiral sink dynamics, as in the first case. However, whereas a prolabor distributive shock causes the same effect as the case above, a positive demand shock will, in this case, cause the wage share to suffer. In the next section, the panel data analysis is utilized to see how the distributive and demand regimes look in advanced and developing countries. Figure 1 – Three scenarios of the neo-Goodwinian model 1a – Profit-led/profit-squeeze 64 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review 1b – Wage-led/wage-squeeze 1c – Profit-led/wage-squeeze Note: Figure 1a represents the profit-led/profit-squeeze neo-Goodwinian model with stable wage share dynamics. Figure 1b represents the wage-led/wage-squeeze neo-Goodwinian model with stable wage share dynamics. Figure 1c represents the profit-led/wage-squeeze neo-Goodwinian model with unstable wage share dynamics. T. Vechsuruck 65 2. Empirical analysis This section will translate the neo-Goodwinian model described earlier into the empirical model. Regarding the theoretical model above, we recognize that any econometric model that attempts to empirically test the model requires two endogenous variables’ time series data: labor share and capacity utilization. The two variables are defined below. 2.1. Wage share data For the wage share data in this paper, I use the labor shares (LABSH) from the Penn World Tables (PWT) 10.0 dataset, since it covers more than 100 countries across continents, and many series are dated from 1950 up to right before the COVID-19 pandemic in 2019. According to Feenstra et al. (2015, pp. 21-27), the dataset also adjusts for self-employment income, which is prevalent in developing countries following Gollin (2002), Timmer et al. (2012), and their estimations. 5 Table 1 summarizes the list of countries and the wage share for each country. Overall, there are 81 countries from all regions around the world. Countries are categorized into advanced and developing countries, according to the IMF (2023). The low-income developing countries are excluded. Selected countries must have at least 10 years of wage share data. Overall, there are 81 countries: 34 advanced economies and 47 developing economies. The average wage share is 0.52 (0.58 for advanced and 0.47 for developing countries). Unfortunately, only a handful of countries have the full range of data for wage share, but most countries have different ranges, from more than 68 years to only 14 years. Figure 2 shows advanced countries that experience a long-term decline in their wage shares, including Australia, Canada, France, Netherlands, and the United States. In Australia, for example, the wage share peaked in 1974 at 0.72 before it bottomed out in 2008 at 0.57. Figure 3 shows the trends for selected developing economies. Many countries, such as Bolivia, India, South Africa, China, and Mexico, exhibited a declining trend before 2008, with the shares picking up afterward. In India, for instance, the wage share decreased from 0.7 in the 1970s to 0.48 in 2007. In China, the wage share declined from 0.6 in 2002 to 0.55 in 2010 before increasing to 0.59 in 2016. This U-shaped trend in China is confirmed by several studies, including those by Zhou (2015), Qi (2020), and Vechsuruck (2023). In Mexico, the wage share did not have a clear trend. However, the absolute level of the wage share was already low, at less than 0.4, the lowest in these five countries. 5 According to Feenstra et al. (2015), three methods were used to construct a ‘best estimate’ labor share. First, when mixed income data are available, they calculate the labor share of income as Compensation of Employees over GDP – Mixed Income. This method applies to almost half the countries in the sample (Feenstra et al., 2015, sec. Appendix C). Second, for a few countries whose unadjusted labor share exceeds 0.7, the unadjusted labor share is used since it already includes the self-employed labor income. This method applies only to a few countries in the dataset. Third, when the mixed income data are not available and the unadjusted labor share is below 0.7, the estimation of labor share for this group of countries compares two methods. The first method estimates the labor share of income as compensation of employee multiplied by the total number of wage employees over total employees over GDP. The second method uses the value added of agriculture as a proxy for mixed income. It follows the calculation that the labor share equals compensation of employees + mixed income over GDP. The final step is to pick the lower number from the two methods as the labor share of income of the country. Overall, 127 countries are covered (out of 167). In 2005, the average labor share was 0.52, which is lower than the 0.7 that Gollin (2002) preferred or the two-thirds (0.67) rule of thumb. 66 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review Table 1 – Wage share summary of all countries Country Period Advanced or developing Mean Median Min Max First 5year average Last 5year average Change Angola 2002-2018 Developing 0.29 0.29 0.23 0.36 0.26 0.33 0.07 Argentina 1993-2013 Developing 0.40 0.39 0.31 0.54 0.4 0.48 0.08 Armenia 1991-2017 Developing 0.65 0.64 0.55 0.75 0.74 0.57 –0.17 Australia 1959-2018 Advanced 0.63 0.62 0.57 0.72 0.68 0.59 –0.09 Austria 1995-2018 Advanced 0.58 0.58 0.55 0.63 0.62 0.58 –0.04 Azerbaijan 1994-2017 Developing 0.33 0.32 0.21 0.57 0.48 0.26 –0.22 Bahrain 1992-2010 Developing 0.33 0.34 0.30 0.38 0.35 0.3 –0.05 Belarus 1990-2015 Developing 0.55 0.55 0.43 0.63 0.49 0.57 0.08 Belgium 1985-2018 Advanced 0.62 0.62 0.59 0.64 0.63 0.6 –0.03 Bolivia 1970-2015 Developing 0.53 0.53 0.45 0.72 0.53 0.47 –0.06 Brazil 1992-2017 Developing 0.55 0.55 0.49 0.58 0.53 0.52 –0.01 Bulgaria 1995-2018 Developing 0.48 0.48 0.39 0.54 0.46 0.52 0.06 Canada 1970-2018 Advanced 0.68 0.67 0.63 0.77 0.76 0.66 –0.1 Chile 1996-2009 Developing 0.46 0.48 0.38 0.52 0.5 0.41 –0.09 China 1992-2016 Developing 0.58 0.57 0.55 0.61 0.58 0.58 0 Colombia 1992-2018 Developing 0.48 0.48 0.45 0.51 0.49 0.49 0 Croatia 1995-2018 Developing 0.64 0.64 0.59 0.71 0.67 0.59 –0.08 Czech Republic 1992-2018 Advanced 0.52 0.52 0.51 0.55 0.52 0.53 0.01 Denmark 1995-2018 Advanced 0.64 0.64 0.62 0.67 0.65 0.62 –0.03 Dominican Rep. 1991-2016 Developing 0.54 0.52 0.43 0.67 0.64 0.44 –0.2 Ecuador 1970-2013 Developing 0.48 0.47 0.35 0.68 0.57 0.66 0.09 Egypt 1996-2015 Developing 0.37 0.38 0.31 0.42 0.4 0.35 –0.05 Estonia 1994-2018 Advanced 0.59 0.59 0.55 0.66 0.64 0.58 –0.06 Finland 1975-2018 Advanced 0.62 0.61 0.56 0.71 0.68 0.58 –0.1 France 1950-2018 Advanced 0.65 0.65 0.61 0.69 0.68 0.62 –0.06 Gabon 1972-2004 Developing 0.37 0.37 0.28 0.50 0.37 0.33 –0.04 Georgia 1999-2018 Developing 0.35 0.37 0.23 0.45 0.36 0.43 0.07 Germany 1991-2018 Advanced 0.64 0.63 0.59 0.68 0.67 0.63 –0.04 Greece 1996-2018 Advanced 0.53 0.54 0.48 0.55 0.49 0.53 0.04 Guatemala 2001-2018 Developing 0.51 0.50 0.48 0.55 0.53 0.49 –0.04 Hong Kong 1980-2017 Advanced 0.49 0.49 0.45 0.52 0.47 0.52 0.05 Hungary 1995-2018 Developing 0.59 0.59 0.55 0.65 0.62 0.56 –0.06 Iceland 1995-2018 Advanced 0.61 0.61 0.51 0.68 0.63 0.6 –0.03 Indonesia 2000-2014 Developing 0.45 0.45 0.44 0.47 0.45 0.46 0.01 India 1975-2017 Developing 0.62 0.62 0.48 0.75 0.74 0.52 –0.22 Iran 1994-2016 Developing 0.32 0.31 0.25 0.41 0.38 0.32 –0.06 Iraq 1997-2010 Developing 0.20 0.21 0.09 0.32 0.14 0.27 0.13 Ireland 1995-2018 Advanced 0.46 0.48 0.32 0.56 0.53 0.35 –0.18 Israel 2000-2018 Advanced 0.56 0.55 0.54 0.60 0.58 0.54 –0.04 Italy 1980-2018 Advanced 0.54 0.52 0.50 0.60 0.59 0.52 –0.07 Jamaica 1970-2018 Developing 0.56 0.58 0.46 0.63 0.58 0.6 0.02 Japan 1980-2017 Advanced 0.58 0.58 0.55 0.63 0.62 0.56 –0.06 Jordan 1970-2009 Developing 0.48 0.49 0.45 0.50 0.49 0.46 –0.03 Kazakhstan 1990-2016 Developing 0.48 0.45 0.38 0.61 0.52 0.4 –0.12 Korea 1970-2017 Advanced 0.56 0.56 0.50 0.65 0.63 0.52 –0.11 Latvia 1994-2018 Advanced 0.52 0.52 0.46 0.62 0.56 0.54 –0.02 Lithuania 1995-2018 Advanced 0.51 0.51 0.46 0.58 0.54 0.51 –0.03 Luxembourg 1995-2018 Advanced 0.56 0.56 0.54 0.60 0.55 0.55 0 Mauritius 1990-2010 Developing 0.48 0.48 0.43 0.55 0.53 0.43 –0.1 Mexico 1993-2018 Developing 0.38 0.38 0.36 0.43 0.4 0.37 –0.03 Mongolia 1995-2018 Developing 0.40 0.40 0.33 0.46 0.42 0.41 –0.01 Morocco 1998-2018 Developing 0.49 0.49 0.47 0.51 0.5 0.49 –0.01 T. Vechsuruck 73 regime is profit-led, ranging from 0.01% to 0.07%. These results are in the same range as the two studies above suggested for the U.S. and OECD economies. 9 Figure 5 simulates the results from the first column of table A1 (HP filter) to create trajectories of the system when the distributive nullcline is positively sloped (profit-squeeze) and the utilization nullcline is negatively sloped (profit-led). For simplicity, the long-run output gap is assumed to be zero. The counterclockwise convergence to the long-run equilibrium (zero GDP gap) seemingly slows in the very first years but speeds up in later years. This implies that any negative output shock might create prolonged stagnation in the total system before it can reach recovery years. Figure 5 – Trajectories from the HP output gap in table A1 The linear trends for both the long-run coordinates are introduced to test if the long-run equilibrium might move downwards or upwards. The ψ0 ∗ coefficient is reinterpreted as the 1970 wage share equilibrium and u0 ∗ as the 1970 utilization equilibrium. The trends can be negative or positive. The equations can be specified as: 9 Note that this aggregative estimation can be biased for short-run effects and play down the long-run effects. Blecker (2016) stressed the time horizon differences. He argued that aggregate demand tends to be profit-led in the short run and wage-led in the long run because consumption positively responds to a higher wage share more in the longer run. Rolim (2021) agreed and added that most structural analysis still emphasized more the short-run effects and suggested that the cointegration test can be used to detect the existence of the long-run relationship between income distribution and consumption. The economy can become more and more wage-led in the long run if consumption is more positively responsive to a higher labor share. 74 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review 𝜓𝑡−𝜓𝑡−1 =𝛼0(𝜓𝑡−1−(𝜓0 ∗−𝛼1𝑢0 ∗+(𝜓1 ∗−𝛼1𝑢1 ∗)(𝑑𝑎𝑡𝑒−1970))−𝛼1𝑢𝑡−1)+𝜖𝑡 (11) 𝑢𝑡−𝑢𝑡−1 =β0(ψ𝑡−1−(ψ0 ∗−β1𝑢0 ∗+(ψ1 ∗−β1𝑢1 ∗)(𝑑𝑎𝑡𝑒−1970))−β1𝑢𝑡−1)+υ𝑡 (12) where 𝜓1 ∗ is a long-run wage share trend, and 𝑢1 ∗ is a long-run utilization trend. The estimation results are presented in table A2. Most coefficients correspond to the previous results. However, the last two coefficients pose two crucial points. First, the sign of the long-run utilization trend is inconclusive. The coefficients from the band-pass and the moving average gaps are negative, whereas those from the BN and Hamilton filters are positive. The coefficient from the HP filter is insignificant at any level. This ambiguous result implies that there are neither positive nor negative effects of the output gap in the long run. Second, the long-run wage trend is negative and significant across all filters except for the BN gap. In other words, there is a longterm downward movement of wage share. These results suggest that there is a collective effort to suppress labor income even though the benefit is not visible in the long run. 2.3.2. Advanced vs. developing country results The estimation when the world economy is divided into developed and developing groups of countries is presented in table A3. The magnitudes across different filters vary. However, some patterns are conclusive. First, the wage slope coefficients are all positive across all filters. The distributive regime is still Marxist/profit-squeeze. However, the results show that this pattern is always stronger in advanced than in developing countries. For example, from the HP gap, a percentage increase in the output gap results in a 2.64% increase in the wage share in developing countries, but it increases to 3.54% in advanced countries. These results can reflect stronger labor institutions or unions, tighter labor markets with less informal employment, and a larger welfare state in the advanced countries that allow wages to increase in tandem with economic upswings. The profit-led demand regime can be observed across the board (negative utilization slopes). This regime is also stronger in developed countries than in developing countries based on all different filters. For example, from the HP gap, a percentage increase in the wage share results in a 0.08% (or 1/12.74) increase in the output gap in advanced countries, whereas it leads to a 0.07% (or 1/14.42) increase in developing countries. The results of the long-run utilization intercept are still ambiguous. The coefficients are negative for the band-pass and the moving average gaps, but they are positive for the HP, BN, and Hamilton gaps. These unclear results confirm the result above, that there is no movement of the output gap in the long run. In sum, both distributive and demand regimes in developing countries are profit-led/profit squeeze. However, the regimes in developing countries are weaker than in advanced countries. To better illustrate this point, figure 6 compares the demand regimes and distributive regimes between two groups of countries. This diagram is based on the coefficients from the HP gap. The dotted lines represent developing countries’ distributive/demand regimes, and the dashed lines represent advanced countries’ two regimes. As analyzed above, the advanced countries’ distributive curve is stronger or steeper, indicating the higher bargaining power of labor unions in advanced countries that are more likely to be able to pressure for higher wages amidst the economic upturns. The steeper slope of developing countries’ demand regime in this 𝑢−𝜓 plane automatically translates into the flatter slope of the regime in the 𝜓−𝑢 plane. In other words, in the short run, higher utilization can be achieved more for developed countries for every percentage reduction in wage share, given other conditions. T. Vechsuruck 75 Figure 6 – Comparison of demand-distribution regimes in advanced vs. developing countries (HP output gap) Lastly, the estimation with linear trends is separated between developing and advanced countries to analyze if there is any difference in terms of their coefficients and long-term trends. The results are in table A4. Similar to the results in table A3, it reveals that the long-run utilization trends are still inconclusive for both groups of countries. On the other hand, the long-run wage trend is negative and significant for all filters. The results also show that the wage trend is worse or more negative in developing countries than in advanced countries. This finding implies that, even though all the countries have attempted to suppress wages, the labor share in developing countries has suffered more in the last few decades. 2.4. Discussion: A global race to the bottom A nation may aim to reduce a unit labor cost to become more competitive internationally. However, once every nation commits to the same strategy, it can result in unfavorable outcomes for all nations as a whole. Robinson (1947) argued that an increase in the balance of trade is tantamount to an increase in investment, which usually leads to an increase in employment. As the global market does not grow fast enough to accommodate all sales, each nation seeks to increase the share in the market that will benefit its people; but this comes at the expense of other nations, because the balance of trade of the world as a whole must be zero. This zero-sum game means an increase in exports for one country implies an increase in imports in another. In other 76 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review words, under international competition, countries aim to increase their employment by exporting unemployment to the rest of the world. Therefore, a so-called beggar-my-neighbor game is played between nations, such as during the interwar period (Rothermund, 2002, pp. 6-9). After one nation succeeds at the expense of others, the other nations will retaliate. The principal devices to increase a trade balance entail import restrictions, export subsidies, exchange rate depreciation, and wage cuts. For instance, an exchange rate depreciation or a fall in money wages would stimulate a primary increase in employment in export industries, assuming the Marshal-Lerner condition holds. Put simply, there are four suits in the pack, and a country tries to play a higher card out of any suit to be ahead of others (Robinson, 1947, p. 69). The long-run consequence of the global race to the bottom can be illustrated by the profitled/profit-squeeze neo-Goodwinian model extended from figure 1a. The comparative statics is shown in figure 7. When countries try to gain competitiveness by using any one of the four cards above, it can be interpreted as the antilabor distributive shock or a rightward shift of the upwardsloping distributive curve (nullcline). The steady-state moves from point A to point B. With a lower long-run wage share, the profit-led regime should allow the economy to move toward a higher long-run capacity utilization because investment responds positively to a rise in profit share. However, this might not be the case regarding the weakening profit-investment nexus recently manifested in many countries. Instead, it could imply that a negative demand shock may occur, and the utilization curve is simultaneously shifted to the left. Investment in this situation does not positively respond to the lower wage share, which implies that the long-run utilization might increase only a little or not at all. The outcome can be simply a fall in wage share without any gain in capacity utilization in the long run, shown by the steady state moving from point B to point C. The global economy has been trapped in the so-called secular stagnation (Hein, 2016). Figure 7 – The race to the bottom in the neo-Goodwinian model T. Vechsuruck 77 This global shift corresponds to the econometric results shown above; in the short run, the demand is profit-led and the distribution is profit-squeeze. However, in the long run, the race to the bottom, captured by falling wage shares, yields neither positive nor negative outcomes on the long-run capacity utilization. The shift in the demand regime dominates the shift in the distributive regime. The short-run, cyclical behavior can be reconciled with the medium-term and long-term trends in falling labor shares and slowed-down economic growth observed worldwide in the past few decades (Blecker, 2020). 3. Conclusion In this study, I show that in the era of rapid globalization of the past few decades, there is evidence of the race to the bottom across the world. First, from the PWT 10.0, I studied the relationship between income distribution and capacity utilization for 34 advanced and 47 developing countries. Of 81 countries, 58 have experienced a decline in the wage share. The average wage share in developing countries is also lower than in advanced countries. Second, I employed the neo-Goodwinian model to estimate the interactions between the wage share and the output gap across the globe. The output gap is obtained through five filters to prevent any bias and to strengthen the results’ robustness: the Hodrick-Presscott (HP), the Baxter-King frequency (band-pass), the moving average, the Beveridge-Nelson (BN), and the Hamilton. The panel data estimation shows that the world system exhibits a counterclockwise oscillatory convergence to the equilibrium point where wage share is the predator and capacity utilization is the prey. The distributive curve is upward-sloping, which represents a Marxian/profit-squeeze regime. The demand curve is downward-sloping, which implies the regime is profit-led. The results have been confirmed both in advanced and developing countries. However, the demand and distributive regimes are stronger in advanced economies than in developed countries. In the long run, there is no positive gain in utilization but the decline in wage share is intensified, especially in developing countries. The outcome could be interpreted as a result of the global beggar-my-neighbor game in which nations attempt to suppress labor income shares while the global demand becomes stagnant. The global race to the bottom benefits countries in the short run but not in the long run. Appendices A. Theoretical model Regarding equation (7) and equation (8), a nontrivial stationary solution, where 𝑢󰇗 =0 and 𝜓󰇗=0, yields the utilization and distributive nullclines of the system, respectively: 𝑢󰇗 =0→𝑢=−𝜙0 𝜙𝑢−𝜙𝜓 𝜙𝑢𝜓 (A.1) 𝜓󰇗=0→𝜓=−𝜃0 𝜃𝜓−𝜃𝑢 𝜃𝜓𝑢 (A.2) 78 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review On the 𝑢−𝜓 plane, the slopes of demand and distributive curves are important as they signify the characteristics of each regime. The slopes can be derived as: 𝑑𝜓 𝑑𝑢|𝑢󰇗=0 =−𝜙𝑢 𝜙𝜓=𝛽𝑢−𝛼𝑢 𝛼𝜓−𝛽𝜓≶0 (A.3) 𝑑𝜓 𝑑𝑢|𝜓󰇗=0 =−𝜃𝑢 𝜃𝜓=𝛿𝑢−𝛾𝑢 𝛾𝜓−𝛿𝜓≶0 (A.4) The long-run solution or stable node, where wage share and utilization are constant, can be solved by equating two nullclines or equation A.1 and equation A.2 to have: 𝑢∗=𝜃0𝜙𝜓−𝜙0𝜃𝜓 𝜙𝑢𝜃𝜓−𝜃𝑢𝜙𝜓 𝜓∗=𝜙0𝜃𝑢−𝜃0𝜙𝑢 𝜙𝑢𝜃𝜓−𝜃𝑢𝜙𝜓 (A.5) To determine the dynamics and the stability of the system at the stationary points where 𝑢󰇗 = 𝜓󰇗=0, the Jacobian matrix, trace, and determinant can be obtained as: 𝐽=(𝜙𝑢𝜙𝜓 𝜃𝑢𝜃𝜓) 𝑇𝑟(𝐽)=𝜙𝑢+𝜃𝜓 𝐷𝑒𝑡(𝐽)=𝜙𝑢𝜃𝜓−𝜃𝑢𝜙𝜓 (A.6) From this system, the stability condition, as well as the characteristics of each curve, cannot be determined a priori, since the slopes of both curves, the trace, and the determinant of the Jacobian matrix fundamentally depend on how the wage share and utilization affect output, potential output, real wage, and labor productivity along the economic cycle. In the next step, some economically meaningful closures will be analyzed to have phase diagrams of the system. Note that the focus is more on the cases where the nullclines are stable in isolation (𝜙𝑢,𝜃𝜓<0) and the system is locally stable (𝑇𝑟(𝐽)<0 and 𝐷𝑒𝑡(𝐽)>𝑜), which implies that only the signs of 𝜙𝜓 and 𝜃𝑢 are left to be explored. Considering the demand regime, with the Keynesian stability condition, 𝛼𝑢 is assumed to be negative to decelerate economic growth in the long run as saving is growing faster than investment. Capital accumulation has responded positively to utilization because, by profit rate accounting, an increase in utilization, given the rates of profit share and organic composition of capital, leads to an increase in the profit rate. Similarly, a rise in profit share can boost profitability and investment demand. Both arguments thus implicitly mean that 𝛽𝑢>0 and 𝛽𝜓<0. From equation A.3, we can determine the sign of the nominator as positive. The sign of the slope now depends only upon the sign of 𝛼𝜓 or whether the demand schedule is wage-led or profit-led. As the demand regime is determined by whether or not the size of a positive effect of an increasing wage share on consumption can dominate the negative of an increasing wage share on investment, the economy is always wage-led (positive slope) when demand is wage-led (αψ>0). On the other hand, if demand is profit-led (𝛼𝜓<0), the overall demand regime is inconclusive. Barbosa-Filho and Taylor (2006) found that, in the United States, the size of the negative effect on demand outperforms the negative effect on investment and capital accumulation (|𝛼𝜓|>|𝛽𝜓|), which T. Vechsuruck 79 forces the slope to be negative, or a profit-led regime. If the opposite case holds, the demand regime is wage-led. The distributive regime is slightly more complicated when we try to determine the signs of each coefficient. According to Marx's Reserve Army of Labor hypothesis, an economic upswing will increase labor's bargaining power as the unemployed are depleted. The real wage therefore tends to vary procyclically (𝛾𝑢>0). Labor productivity is also assumed to react positively to utilization as firms invest more in new technology while they see improving profitability (𝛿𝑢>0). We can see that the sign of the nominator of equation A.5 cannot be determined a priori. According to Barbosa-Filho (2001), suppose that the real wage growth rate is a negative function of the real wage level and a positive function of labor productivity. We thus have a negative relation between wage share and real wage (𝛾𝜓<0). Further, since we emphasize the case in which the distributive nullcline is stable in isolation (𝜃𝜓<0), 𝛿𝜓 must be positive. If the 𝛿𝜓 sign is negative, the sign of 𝜃𝜓 will depend on the difference between 𝛾𝜓 and 𝛿𝜓. In the prior case (𝜃𝜓<0), the denominator is forced to be negative. The sign of the distributive curve will rely only upon the nominator sign. In particular, if 𝛿𝑢>𝛾𝑢, we will have a forced-saving/Kaldorian distributive regime (negative slope distributive nullcline). If 𝛿𝑢<𝛾𝑢, we will instead have a profit-squeeze/Marxian distributive regime (positive slope distributive nullcline). For the stability condition, we disregard the saddle point case, so the determinant of Jacobian matrix in equation A.6 must only be positive. B. Filters used for obtaining output gaps In the neo-Kaleckian and neo-Goodwinian literature, it is common to use the filter to obtain the long-term trend of the real output and calculate for the capacity utilization or output gaps. Also, filtering methods can be used to have a long-run economic growth trend, although some methods are exposed to controversies (see for example Nikiforos, 2016, 2020, 2021; José Gahn, 2020; Haluska, 2020). The Hodrick Presscott (HP) filter has been one of the most popular filters to separate the longterm trend from the short-term fluctuations in the macroeconomic time series. However, there are several criticisms regarding the use of the HP filter for obtaining the potential output. Cogley and Nason (1995) claimed that the HP filter can potentially create spurious cycles. In addition, Hamilton (2017) argued that the HP filter has three main issues. First, it generates series with spurious dynamic relations. Second, filtered values at both ends of the sample differ greatly from those in the middle. Third, it produces values for the smoothing parameter that are very different from common practice. Blecker (2016) compared U.S. utilization rates obtained from the HP filter with those from a survey by U.S. firms. He found that the utilization rates from the HP filter downplayed the adverse effect of the 2008 financial crisis. Avritzer (2022) examined the relationship between the long-run capacity utilization, economic growth, and the wage share for the US. By applying different filters, including HP, moving average, Hamilton, and band-pass, she found that the long-run capacity utilization is endogenous to the long-run income distribution. Even though she claimed that the results are not much different across different filters, she showed that the band-pass and the HP showed a negative relationship between the long-run wage share and long-run capacity utilization, whereas the Hamilton and the moving-average filters did not show any significant results. I therefore follow Avritzer (2022) by using different filters to estimate the econometric relationship. The different filters should prevent bias when using any filter alone and they enhance the robustness of the results. The filters used are: 80 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review 1. Hodrick-Prescott filter (HP filter). This standard filter is often used in macroeconomics to obtain the output trends and cycles suggested by Hodrick and Prescott (1997) after the working paper was circulated during the 1980s. Since the data are at an annual level, I use the smoothing parameter (lambda) equals 100. 2. Baxter-King frequency filter (band-pass filter). The band-pass filter, suggested by Baxter and King (1999), is a linear filter that calculates the two-sided weighted moving average where cycles in a “band,” or intermediate values, are “passed” through or extracted, given specified lower and upper bounds. According to Benati (2001), the band-pass filter allows us to target a specific frequency band while discarding all the others. However, the band-pass filter may distort key business cycle stylized facts as captured by the cyclical component of GDP and may create entirely spurious stylized facts (p. 7). Here I set the upper and lower bounds equal to the standard levels at 2 and 8, respectively. This fixed length filter requires that, for every weighted moving average, I use the same number of lead and lag terms at 3. 3. Moving average. I decided to calculate a moving average of five years centered on obtaining the trend of GDP. 4. Beveridge-Nelson filter (BN filter). I use the Kamber et al. (2018) modification of the Beveridge and Nelson (1981) decomposition that imposes a lower signal-to-noise ratio on an AR model, which resulted in a persistent output gap with large amplitude. Unlike the HP or bandpass filter, this modified BN filter also requires fewer estimation revisions to match observable data. Kamber et al. (2018) used AR(12) for the quarterly data, so I use AR(3) for my annual data. 5. Hamilton filter. Hamilton (2017) proposed a much simpler way to extract trends and cycles from a time series. He suggested using a linear time series model shifted ahead by ℎ periods regressed against lags of the series of 𝑝 periods. In most studies, the data are quarterly with 𝑝 = 4 and ℎ = 8, but here I have data at an annual level. Because Hamilton stated that a 2-year horizon should be a standard benchmark if we are interested in business cycles (p. 838), I choose a combination of 𝑝 = 2 and ℎ = 2. So a modified autoregressive AR(2) model can be expressed as: 𝑦𝑡=β0+β1𝑦𝑡−2+β2𝑦𝑡−3+υ𝑡 (B.1) 𝜐𝑡=𝑦𝑡−(𝛽󰆹0+𝛽󰆹1𝑦𝑡−2+𝛽󰆹2𝑦𝑡−3) (B.2) where υ𝑡 estimates cyclical components and the fitted values are the trends. Note that the Hamilton filter may produce a very noisy measure of potential GDP. Quast and Wolters (2022, p. 152) showed that the filter does not evenly cover typical business cycle frequencies from 6 to 32 quarters. It mutes short and amplifies medium length economic cycles. The extracted GDP trend or potential output is therefore not smooth. C. Econometric model Given that the rate of change is defined as 𝛥𝑥 𝑥=𝑥𝑡−𝑥𝑡−1 𝑥𝑡−1 , the pure or original Goodwin model, rather than in differential equation form, can be estimated by the following difference-equation specification: T. Vechsuruck 81 𝜓𝑡−𝜓𝑡−1 =𝛼0𝜓𝑡−1(𝑢𝑡−1−𝑢0 ∗)+𝜖𝑡 (C1) 𝑢𝑡−𝑢𝑡−1 =β0𝑢𝑡−1(ψ𝑡−1−ψ0 ∗)+υ𝑡 (C2) where 𝜖 and 𝜐 are error terms, 𝛼0 is wage share scaling, 𝛽0 is gap scaling, and 𝜓0 ∗ is a long-run wage intercept. The long-run gap intercept is 𝑢0 ∗ and restricted at zero. The original version of the Goodwin model often cannot provide strong results, as it shows closed orbits around a unique fixed point (Kiefer and Rada, 2015, p. 7). Therefore, there must be some adjustments to the equations. The general Goodwin model is an adapted version of the original Goodwin model above and can be shown as: 𝜓𝑡−𝜓𝑡−1 =𝛼0(𝜓𝑡−1−(𝛿1+𝛿2𝑢𝑡−1))+𝜖𝑡 (C.3) 𝑢𝑡−𝑢𝑡−1 =β0(ψ𝑡−1−(δ3+δ4𝑢𝑡−1))+υ𝑡 (C.4) At the steady state, 𝛥𝜓 and 𝛥𝑢=0, errors are gone, and 𝜓𝑡−1 and 𝑢𝑡−1 turn into 𝜓0 ∗ and 𝑢0 ∗, respectively. We have: 𝜓0 ∗=𝛿1+𝛿2𝑢0 ∗ (C.5) ψ0 ∗=δ3+δ4𝑢0 ∗ (C.6) Then we solve for 𝛿1 and 𝛿3 to have: 𝛿1=𝜓0 ∗−𝛿2𝑢0 ∗ (C.7) δ3=ψ0 ∗−δ4𝑢0 ∗ (C.8) Then we plug back 𝛿1and 𝛿3 into equation (C.5) and equation (C.6) to have: 𝜓𝑡−𝜓𝑡−1 =𝛼0(𝜓𝑡−1−(𝜓0 ∗−𝛿2𝑢0 ∗)−𝛿2𝑢𝑡−1)) + 𝜖𝑡 (C.9) 𝑢𝑡−𝑢𝑡−1 =β0(ψ𝑡−1−(ψ0 ∗−δ4𝑢0 ∗)−δ4𝑢𝑡−1))+𝜐𝑡 (C.10) where 𝛼0 is a wage share scaling, 𝜓0 ∗ is the long-run wage trend, 𝛿2 (or 𝛼1) is a wage slope, 𝑢0 ∗ is the long-run gap trend, 𝛽0 is a gap scaling, and 𝛿4 (or 𝛽1) is a gap slope. These equations are estimated, and the results are shown in table A1. For advanced/developing countries differences, the wage and utilization slopes are allowed to vary across different groups of countries. 82 A global race to the bottom: The neo-Goodwinian aggregative-systems estimation PSL Quarterly Review Table A1 – Econometric results for the globe HP Band-pass Moving average BN Hamilton Wage slope (𝛼1) 2.86 (32.12) 5.83 (–60.65) 5.36 (98.55) 3.21 (18.96) 0.64 (33.63) Utilization slope (𝛽1) –14.04 (–39.53) –39.31 (–62.74) –45.39 (–105.98) –95.53 (–6.26) –32.36 (–30.65) Wage share scaling (𝛼0) –0.07 (–36.82) –0.07 (–68.09) –0.07 (–106.92) –0.02 (–21.95) –0.07 (–49.64) Utilization scaling (𝛽0) –0.03 (–41.25) –0.02 (–58.9) –0.02 (–101.5) –0.002 (–6.15) –0.02 (–30.83) Long-run wage intercept (𝛼0 ∗) 98.95 (703.56) 98.2 (884.45) 98.52 (854.15) 88.57 (191.25) 97.92 (930.75) Long-run utilization intercept (𝑢0 ∗) 0.16 (6.69) –0.1 (–33.82) –0.09 (–75.46) 1.01 (37.75) 0.74 (34.62) Schwarz Bayesian Criterion (SBC) –830.17 –5526.54 –5027.69 –4606.72 4238.9 Akaike Information Criterion (AIC) –843.58 –5539.5 –5040.83 –4620.04 4225.77 Note: All coefficients are significant at a 1% significance level (t-statistics in the parentheses), except as denoted otherwise. Asterisks * and ** indicate a significance level of 10% and 5%, respectively. Table A2 – Econometric results with linear trends for the globe HP Band-pass Moving average BN Hamilton Wage slope (𝛼1) 2.72 (30.98) 5.57 (57.67) 5.15 (85.38) 2.89 (15.2) 0.61 (32.56) Utilization slope (𝛽1) –13.41 (–39.2) –35.95 (–58.63) –43.06 (–79.21) –93.15 (–6.3) –36.94 (–25.58) Wage share scaling (𝛼0) –0.07 (–35.67) –0.07 (–64.9) –0.07 (–93.39) –0.02 (–18.52) –0.08 (–46.2) Utilization scaling (𝛽0) –0.03 (–41.06) –0.02 (–57.1) –0.02 (–78.72) –0.002 (–6.19) –0.02 (–25.83) Long-run wage intercept (𝛼0 ∗) 102.16 (192.87) 102.17 (223.1) 101.74 (232.26) 71.56 (30.64) 100.69 (217.16) Long-run utilization intercept (𝑢0 ∗) 0.22 (2.05)** –0.02 (–0.63)a –0.06 (–5.62) –1.1 (–8.82) 0.34 (3.2) Long-run wage trend (𝜓1 ∗) –0.1 (–6.95) –0.13 (–10.04) –0.1 (–9.03) 0.55 (9.37) –0.08 (–6.97) Long-run utilization trend (𝑢1 ∗) –0.002 (–0.64)a –0.003 (–4.54) –0.001 (–3.08) 0.07 (20.52) 0.01 (4.36) Schwarz Bayesian Criterion (SBC) –836.6 –5532.66 –5023.54 –4589.27 4239.7 Akaike Information Criterion (AIC) –854.47 –5549.93 –5041.06 –4607.02 4222.18 Note: All coefficients are significant at a 1% significance level (t-statistics in the parentheses), except as denoted otherwise. Asterisks * and ** indicate a significance level of 10% and 5%, respectively. a insignificant at any level.