Introduction
Various scholars argue that, while left partisan governments continue to be distinctive in their support for generous social policies (Amable et al. Reference Amable, Gatti and Schumacher2006; Jensen and Mortensen Reference Jensen and Mortensen2014; Jensen and Seeberg Reference Jensen and Bech Seeberg2015), partisanship effects are vanishing in economic policies as left governments converge with the right in supporting market deregulation (Mudge Reference Mudge2018; Obinger et al. Reference Obinger, Shmitt and Zohlnhöfer2014; Zohlnhöfer et al. Reference Zohlnhöfer, Obinger and Wolf2008). For example, while left governments used to be associated with greater regulation of financial markets, this partisanship effect has recently all but disappeared, resulting in higher income inequality (Keller and Kelly Reference Keller and Kelly2015; Witko Reference Witko2014). In European countries, this vanishing of partisanship effect is further reinforced by the process of European integration (Billows et al. Reference Billows, Kohl and Tarissan2021; Scharpf Reference Scharpf1999). In particular, the European Commission’s Directorate General for Competition (DG COMP) has assumed growing procedural as well as substantive power (Foster and Thelen Reference Foster and Thelen2024) to move competition policy in all EU countries toward the US model (Bergman et al. Reference Bergman, Coate, Mai and Ulrick2019; Buch-Hansen and Wigger Reference Buch-Hansen and Wigger2010), targeting, in particular, traditionally social democratic countries that are associated with state intervention and protection of workers’ collective rights (Billows et al. Reference Billows, Kohl and Tarissan2021; Foster and Thelen Reference Foster and Thelen2025). For example, EU regulation on mergers and acquisitions has resulted in greater power for transnational corporations and financial capital at the expense of labor (Aydin and Thomas Reference Aydin and Thomas2012; Buch-Hansen and Wigger Reference Buch-Hansen and Wigger2010; Wigger Reference Wigger2012). In other words, the vanishing of partisanship effect on competition policy has marked a partial retreat for left governments from their distinctive egalitarian redistributive agenda (Mudge Reference Mudge2018; Polacko Reference Polacko2021).
This paper offers an argument that differs from the existing literature in several aspects. First, instead of vanishing, the partisanship effect on competition policy has intensified. Second, this partisanship effect has intensified in a novel direction: left partisan governments have become stronger defenders of product market competition than other partisan governments. Furthermore, instead of retreating from redistribution, this new association between left partisanship and market competition has delivered new distributive gains for labor.
To understand the logic behind this argument, I highlight two recent trends in advanced economies. First, rich countries are entering an era of low interest rates, with long-term real interest rates having declined by up to 450 basis points from 1985 to 2015 across the rich world (Rachel and Smith Reference Rachel and Smith2015). Even in the current post-COVID era of high inflation and nominal rates in the Organization for Economic Cooperation and Development (OECD) world, real interest rates (i.e., adjusted for inflation) remain negative in all advanced economies (OECD 2023). Second, because large firms benefit disproportionately from low interest rates, this era of low rates has ushered in a pronounced rise in market concentration, which puts downward pressure on the labor share of income (Autor et al. Reference Autor, Dorn, Katz, Patterson and van Reenen2020; De Loecker et al. Reference De Loecker, Eeckhout and Unger2020; Kroen et al. Reference Kroen, Liu, Mian and Sufi2021; Liu et al. Reference Liu, Mian and Sufi2022).
The current era of low interest rates, in other words, represents a ‘new normal’ (Gagnon et al. Reference Gagnon, Johannsen and López-Salido2021), where workers are getting a less share of the gains from national growth, which opens a new distributive battleground for left partisan governments. By boosting product market competition, left governments can counter the rise of market concentration and defend the labor share of income, thus revitalizing their redistributive agenda for this new era. As a result, when interest rates are low, left partisan governments will be more likely to pursue market competition than other partisan governments. In other words, falling rates lead to rising partisanship effect in competition policy.
Consistent with this argument, the paper draws on data from 10 to 17 OECD countries and shows that as long-term interest rates fall, left governments become more strongly associated with regulatory policies that increase product market competition and reduce concentration. I also show that the embrace of competition by left governments is more consistent with their redistributive concerns than a general ideological shift toward deregulation: when interest rates fall, center-left parties are more likely to oppose market concentration and support redistribution in their manifestos, but no more likely to support a laissez-faire economy; furthermore, the pro-competition policies backed by left governments can involve either lighter regulation (such as fewer administrative barriers on startups) or tougher regulation (such as more vigorous antitrust enforcements or investigations into abuse of dominant market positions). I also draw out the distributive implications, showing that as interest rates fall, left partisanship becomes more strongly associated with lower market concentration and a higher labor share of income. In other words, the current era of depressed interest rates has given rise to a partisan distributive effect that has not been commonly observed in the literature.
The rest of the paper is organized as follows. The next section lays out the paper’s main theoretical argument. Sections 3 and 4 provide the empirical analysis. I conclude and outline directions for future research in section 5.
The main argument
As an extensive literature points out, a confluence of factors including globalization, economic stagnation, and budgetary crises have increased pressure on left governments to embrace orthodoxy in economic policy and support market deregulation. As left governments converge with the right on deregulation, government partisanship has also become a less informative predictor of competition policy in advanced economies (Amenta Reference Amenta2019; Keller and Kelly Reference Keller and Kelly2015; Mudge Reference Mudge2018; Polacko Reference Polacko2021; Witko Reference Witko2014; Zohlnhöfer et al. Reference Zohlnhöfer, Obinger and Wolf2008). In the EU, this vanishing of partisanship effect through convergence toward deregulation is further reinforced by the European Commission’s DG COMP. Since the 1990s, the DG COMP has put pressure on strongly corporatist states such as Scandinavian countries and Austria to roll back state aid. Furthermore, through its discretionary investigation of potential abuse of market positions, the DG COMP has granted growing regulatory lenience on merger activities (Billows et al. Reference Billows, Kohl and Tarissan2021; Buch-Hansen and Wigger Reference Buch-Hansen and Wigger2010). Nevertheless, even in the EU, national governments continue to retain significant influence over competition policy, because since the late 1990s the lack of resources at DG COMP has led to increasing decentralization in the implementation of competition policy to member states (Dierx and Ilzkovitz Reference Dierx and Ilzkovitz2021). This process of decentralization accelerated with Regulation (EC) No 1/2003, which allowed for joint enforcement of competition policy by European and national competition authorities (Kassim and Wright Reference Kassim and Wright2007; Townley and Türk Reference Townley and Türk2019). For example, since Regulation No 1/2003, individual member states have accounted for more than 85% of all antitrust cases in the EU as national governments gained increasing experience in making competition policy (Dierx and Ilzkovitz Reference Dierx and Ilzkovitz2021). Given their continued ability to influence competition policy both within Europe and without, how far would left governments go in embracing the rightwing agenda of deregulation?
As I explain next, the depressed long-term interest rates across the rich world today have generated a new source of distributive disadvantage for workers: market concentration, which can be countered by tougher market competition. This new distributive challenge may therefore reignite a partisanship effect in competition policy, but in a novel direction: left governments may start to support market competition more strongly than other partisan governments.
It is widely documented that real long-term interest rates have steadily declined in the rich world. Based on data from the Jordà-Schularick-Taylor Macrohistory database (Jordà et al. Reference Jordà, Schularick, Taylor, Eichenbaum and Parker2017), Figure 1 plots the real long-term interest rates for the 17 OECD countries included in this paper.
Figure 1. Real interest rates, 17 OECD countries.
While interest rates in these 17 countries stood near 5% in 1995, two decades later, they have become either close to zero or negative in all countries (and they remain negative in the current post-COVID era of high inflation and high nominal rates (OECD 2023)). However, to what extent are long-term rates, a key independent variable for this paper, endogenous to governments’ policy decisions? As Carvalho et al. (Reference Carvalho, Ferrero and Nechio2016) and Gagnon et al. (Reference Gagnon, Johannsen and López-Salido2021) point out, because the decline in long-term rates has persisted for more than two decades, changes in long-term rates are unlikely to be driven by accommodative monetary policy of individual governments, which mainly affects higher-frequency short-term rates. Instead, long-term rates are more strongly driven by secular or global forces orthogonal to the influence of individual governments (such as the aging of the workforce and the global savings glut) and are in this sense more exogenous to domestic economic and political conditions (Del Negro et al. Reference Del Negro, Giannone, Gionnoni and Tambalotti2019; Rachel and Smith Reference Rachel and Smith2015).
For example, the aging of the workforce tends to increase household saving, which puts downward pressure on long-term rates. Relatedly, because younger workers are comparatively more productive in physically repetitive and dexterous manual tasks that are more substitutable with capital, when the aging of the workforce increases the relative scarcity of such workers, firms will substitute labor with capital, which further drives down the return on capital (Acemoglu and Restrepo Reference Acemoglu and Restrepo2022; Carvalho et al. Reference Carvalho, Ferrero and Nechio2016; Stähler, Reference Stähler2021). To this effect, simulations by Gagnon et al. (Reference Gagnon, Johannsen and López-Salido2021) suggest that, had the demographic structure of the workforce in rich countries remained at their state in 1960, real interest rates would have stayed at their level in the 1980s. Furthermore, the Asian financial crises of the late 1990s spurred growing demand from emerging economies for safe and liquid assets, which further depressed long-term rates in the rich world (Del Negro et al. Reference Del Negro, Giannone, Gionnoni and Tambalotti2019; Rachel and Smith Reference Rachel and Smith2015). Discussions above suggest that there are indeed theoretical reasons to regard long-term interest rates as a largely exogenous independent variable relative to government policy decisions. Relatedly, as I discuss later in the empirical section, it is also possible to use the intensity of workforce aging as an instrument for long-term interest rates.
As recent research points out, a decline in long-term rates tends to increase market concentration. Because large firms tend to have better access to credit and take on more leverage, they tend to benefit more from the low borrowing cost that results from depressed real interest rates (Chatterjee and Eyigungor Reference Chatterjee and Eyigungor2019; Kroen et al. Reference Kroen, Liu, Mian and Sufi2021). Furthermore, because large firms tend to maintain a large base level of investment, they also benefit more when low real rates boost the present value of future profits from a given increase in investment (Liu et al. Reference Liu, Mian and Sufi2022). As a result, as long-term rates become lower, investment tends to become more concentrated in the largest firms of the economy, leading to increases in overall market concentration.
As Kroen et al. (Reference Kroen, Liu, Mian and Sufi2021) point out, because firms need time to form investment expectations, it is long-term rates (beyond the influence of individual governments) rather than policy rates that affect market concentration. For example, the authors find that, starting from 2%, a reduction of long-term interest rates by 10 basis points will increase the property, plants, and equipment (PPE) acquisition of the top 5% firms in the economy by 2%. As rates approach zero, there is even more capacity expansion from the top. Similarly, Morlacco and Zeke (Reference Morlacco and Zeke2021) show that a 100-basis-point reduction in interest rates will prompt the top 20% firms in the economy to increase their Selling, General, and Administrative expenditure by 35% relative to the bottom 20% firms. More broadly, Chatterjee and Eyigungor (Reference Chatterjee and Eyigungor2019) show that the fall in long-term interest rates between 1997 and 2015 reduced US firm entry in the economy by 27% and increased the top 1% firms’ output share by 11%.
This ‘concentration effect’ of falling rates, however, can be countered with stronger product market competition, to the potential advantage of workers. For example, when startups face less administrative burden or incumbents are more likely to be targeted with antitrust investigations, firms with dominant market shares will be less able to exploit their financing advantage to expand at the expense of smaller firms and absorb a larger share of total employment (Chatterjee and Eyigungor Reference Chatterjee and Eyigungor2019; Kroen et al. Reference Kroen, Liu, Mian and Sufi2021; Rinz Reference Rinz2022). As employment opportunities become less concentrated in a few employers, employers’ ability to exploit their market power and mark down wages will also decline (Gouin-Bonenfant Reference Gouin-Bonenfant2022; Mertens Reference Mertens2022). Less wage markdowns, in turn, will translate into a greater labor share of income (Azar et al. Reference Azar, Marinescu and Steinbaum2022; Manning Reference Manning2021; Naidu et al. Reference Naidu, Posner and Weyl2018).
Above and beyond the direct effect on market structure, pro-competition policies can also indirectly alter the intensity of wage competition between employers. As Gouin-Bonenfant (Reference Gouin-Bonenfant2022) points out, firms face stronger wage competition when they are closer to each other in productivity. Because concentration in value-added tends to reinforce concentration in productivity advantage, when the market share in value-added becomes less concentrated due to fewer barriers to entry or less regulatory protection of incumbents, productivity also tends to become more evenly dispersed across firms, thus exposing employers to stronger wage competition and further reducing their power to pay workers below their marginal revenue product. Policies that boost market competition, in other words, can help prevent employers from taking advantage of today’s low-rate environment to exercise market power and suppress labor’s share of income.
As noted earlier, existing literature suggests that the partisanship effect in competition policy has gradually vanished as left governments increasingly embrace a neoliberal framework of market deregulation, which often implies a lessening of competition standards (Bergman et al. Reference Bergman, Coate, Jakobsson and Ulrick2010, Reference Bergman, Coate, Mai and Ulrick2019; Foster and Thelen Reference Foster and Thelen2025; Mudge Reference Mudge2018; Obinger et al. Reference Obinger, Shmitt and Zohlnhöfer2014; Zohlnhöfer et al. Reference Zohlnhöfer, Obinger and Wolf2008). Discussions in this section suggest a different possibility: the current era of low interest rates has created an environment of market concentration that gives employers potential market power to suppress the labor share of income, and in response, left governments as the traditional defenders of redistribution may start to support market competition more strongly than other partisan governments. Falling rates, in other words, will lead to rising partisanship effect, and do so with a novel twist, as pro-competition policies become distinctively associated with left partisanship. In this process, left partisanship will also become increasingly associated with a less commonly discussed distributive outcome: a higher labor share of income.
H1 As interest rates become lower, left government partisanship will become increasingly associated with low regulatory barriers to product market competition.
H2 As interest rates become lower, left government partisanship will become increasingly associated with a higher labor share of income.
Empirical analysis: competition policy
Because the data on competition policy and market concentration cover multiple countries and years, there will likely be unspecified country- and time-specific effects that are potentially correlated with the regressors. By violating the standard assumptions of independent observations in ordinary least squares estimators, such heterogeneity may lead to biased and inconsistent estimates (Kollmeyer Reference Kollmeyer2017, pp. 109). For this reason, one estimation strategy I use is two-way fixed effects (country and year). This strategy is effective in absorbing the various unobserved events in competition policy and interest rates that are unique to each country, such as the elevated long-term rates in Ireland and the lack of clear interest rate movements in Japan and Switzerland shown in Figure 1, or unique to each year, such as the appointment of Mario Monti as Commissioner of Competition in 1999 and the formal abandonment of the market dominance standard in EU competition policy in 2004.
On the other hand, as Imai and Kim (Reference Imai and Kim2019; Reference Imai and Kim2021) point out, despite having become the default method for estimating time-series cross-section data, two-way fixed effects are only strictly equivalent to the difference-in-differences estimator under the simplest setting of two groups and two time periods; under the more general multi-country multi-year setting as encountered in this paper, this equivalence cannot be assumed. As Garritzmann and Seng (Reference Garritzmann and Seng2020) point out, because country fixed effects remove a substantial amount of variation in government partisanship, the decision to include such fixed effects can make a material difference to the findings by significantly reducing the likelihood of yielding partisanship effects. In addition, because government partisanship tends to alter with the change of cabinets rather than annually, using year alone as the time unit may further reduce the likelihood of finding partisanship effects (Schmitt Reference Schmitt2016, pp. 1453-4; Schmitt and Zohlnhöfer Reference Schmitt and Zohlnhöfer2019). For these reasons, to increase the robustness of the findings I also follow Garritzmann and Seng (Reference Garritzmann and Seng2024) and use the mixed-effects model as an alternative estimator, which adjusts for the nested structure of the data with variation at the cabinet, country, and year levels.
As noted earlier, because long-term rates are strongly driven by a secular demographic trend of workforce aging beyond the control of individual governments, there are theoretical reasons to treat this variable as largely exogenous to government policy decisions. Furthermore, precisely because the gradual aging of the workforce drives down long-term rates while being largely exogenous to government policies, I further test the robustness of the findings by alternatively using a two-stage least squares (2SLS) estimator (Bollen Reference Bollen2012; Winship et al. Reference Winship, Muller, Morgan, Best and Wolf2014), with the intensity of workforce aging (measured as the proportion of workers 55 and older) as an instrument for long-term interest rates. As an alternative instrument, I also use Acemoglu and Restrepo’s (Reference Acemoglu and Restrepo2022) measure of workforce aging, which is the ratio of older workers to younger/middle-aged workers (between the ages of 20 and 55). According to Acemoglu and Restrepo, this ratio more directly captures the relative scarcity of younger workers, which is the key mechanism that induces firms to substitute labor with capital, thereby driving down the long-term return on capital.
More formally, when long-term rates are instrumented with either of these two measures, the Wald F-test statistics in the 2SLS estimates are always above the 10 percent critical value threshold suggested by Stock and Yogo (Reference Stock, Yogo and Andrews2005), rejecting the null hypothesis of weak instruments. Furthermore, because long-term rates are instrumented parsimoniously (with a single instrument), the 2SLS estimates have exact identification. Since these estimates by definition do not face overidentifying restrictions, Sargan’s (Reference Sargan1958) and Basmann’s (Reference Basmann1960) tests of overidentifying restrictions are not warranted.
While some elements of the paper’s empirical analysis (for example, on left party manifestos and some measures of competition policy) will extend, on the margin, the time coverage to the mid-1970s, it is important to acknowledge that the main timeframe is restricted to the post-1995 period due to limited data availability for market concentration, which is the common conceptual thread throughout the paper. For cross-national data on the level of overall market concentration in the economy, I rely on the European Central Bank’s Competitiveness Research Network (CompNet) dataset,Footnote 1 which draws on administrative as well as public sources from individual countries and seeks to cover all nonfinancial corporations for each country included in the dataset (Autor et al. Reference Autor, Dorn, Katz, Patterson and van Reenen2020, pp. 659). The CompNet dataset aggregates data from all individual firms, and provides aggregate, country-level, measures of market concentration. Because the coverage of CompNet starts from the year 1995, unless noted otherwise, most of the paper’s empirical findings should be treated as applying to the post-1995 period.
For data on interest rates, I use the nominal long-term interest rates variable from the Jordà-Schularick-Taylor Macrohistory database (Jordà et al. Reference Jordà, Schularick, Taylor, Eichenbaum and Parker2017), which I convert to real interest rates by subtracting the rate of inflation. In estimating regulatory barriers to competition, I start by controlling for the overall size of government using total government spending. I also follow Chang and Berdiev (Reference Chang and Berdiev2011) and Schmitt and Zohlnhöfer (Reference Schmitt and Zohlnhöfer2019) by controlling for capital account openness (Chinn and Ito’s (Reference Chinn and Ito2008) index) and trade openness (imports and exports as percentage of GDP). Because various scholars point out that adverse economic conditions tend to trigger regulatory reforms, I also follow the literature and control for the unemployment rate, economic growth, and inflation (Galasso Reference Galasso2014; Smith and Urpelainen Reference Smith and Urpelainen2016; Zohlnhöfer et al. Reference Zohlnhöfer, Engler and Dümig2017). Furthermore, because regulatory changes are sometimes stymied by either organized labor (Schmitt and Zohlnhöfer Reference Schmitt and Zohlnhöfer2019) or political veto points (Pitlik Reference Pitlik2007), I also control for union density, the extent of wage bargaining coordination (Visser’s (Reference Visser2019) index of wage setting coordination) as well as political veto points (an additive index of presidentialism, bicameralism, federalism, and referenda) from Brady et al. (Reference Brady, Huber and Stephens2020).
Before examining left governments’ regulatory responses, I start by briefly examining the effect of falling interest rates on market concentration. Model 1 of Table 1 presents 2SLS estimates of the effect of long-term rates (instrumented with the proportion of workers 55 and older) on market concentration (measured as the combined value-added share of the ten largest firms in the economy, available from CompNet). Consistent with expectation, Model 1 suggests that falling rates indeed result in greater market concentration: a one-percentage-fall in long-term rates will increase the top ten firms’ combined share of value-added in the economy by 2.4 percentage points.
Table 1. Left partisanship and regulatory barriers to product market competition
* p < 0.1; ** p < 0.05; *** p < 0.01.
Next, I examine how this environment of falling rates influences center-left parties’ programmatic intentions through the lens of their election manifestoes. According to this paper, by increasing the structural potential for employers to exploit market power and squeeze workers’ share of income, falling rates should prompt center-left parties to more strongly prefer the curtailing of market concentration and the defense of egalitarian distributions. These sentiments are captured in several variables from the Manifesto Project Dataset (Volkens et al. Reference Volkens, Burst, Krause, Lehmann, Matthieß, Regel, Weßels and Zehnter2021). In the dataset, variable per403 measures the percentage of party manifesto texts expressing opposition to product market concentration, including ’preventing monopolies’, ‘defense of small businesses against disruptive powers of big businesses’, and ‘calls for increased consumer protection’ (Volkens et al. Reference Volkens, Burst, Krause, Lehmann, Matthieß, Regel, Weßels and Zehnter2021, p. 14).
Separately for each country and year in the dataset, I calculate the overall proportion of center-left party manifesto texts opposing market concentration by taking the average of variable per403 across parties identified in the Manifesto Project Dataset as either Social Democratic Parties (‘30’ on the party family variable parfam) or Socialist or other left parties (‘20’ on the party family variable). Analogously, I calculate the proportion of center-left manifesto texts supporting the principle of equality (variable per503), including ‘fair distribution of resources’, ‘protection of underprivileged groups’ and ‘removal of class barriers’, and the proportion supporting the welfare state, which I follow Garritzmann and Seng (2020; Reference Garritzmann and Seng2024) and calculate as per504 (texts emphasizing welfare expansion) minus per505 (texts emphasizing welfare limitation).
This paper’s theory implies that the three separate dependent variables above (‘opposing market concentration’, ‘supporting equality’, and ‘supporting the welfare state’) are all similarly driven by an increase in center-left parties’ redistributive concern in an environment of falling long-term rates. As a result, if each dependent variable is estimated separately in its own equation, the potential correlation of error terms across the three equations will be ignored, leading to inefficient and potentially biased estimates. To properly account for this correlated error structure, Model 2 follows Betts (Reference Betts1997) and Piva et al. (Reference Piva, Santarelli and Vivarelli2005) by using the Seemingly Unrelated Regression (Zellner Reference Zellner1962) estimator, which estimates the three dependent variables in a joint system of equations with identical regressors and correlated error terms. Consistent with expectation, the coefficients for interest rates are negatively signed for all three outcomes: center-left parties are more likely to express opposition to market concentration and support for equality/welfare states when long-term rates are lower. Based on the estimates, each standard deviation decline in long-term rates will increase the proportion of center-left party manifesto texts opposing market concentration by .78 percentage points, which is equivalent to one third of standard deviation in this outcome. The effect on manifesto texts supporting equality or the welfare state is of similar magnitude.
In Model 3, the dependent variable is the proportion of center-left party manifesto texts expressing support for a laissez-faire economy (variable per401 in the Manifesto Project). Here, in contrast to the previous outcomes, interest rates do not have any statistically significant effect. In other words, when interest rates fall, center-left parties are more likely to express opposition to market concentration and support for redistribution but no more likely to express support for a laissez-faire economy, which implies that their embrace of stronger market competition is more consistent with redistributive intentions than a general ideological shift toward deregulation.
For data on competition policy, I follow a long line of literature (Potrafke Reference Potrafke2010; Schmitt and Zohlnhöfer Reference Schmitt and Zohlnhöfer2019; Smith and Urpelainen Reference Smith and Urpelainen2016; Zohlnhöfer et al. Reference Zohlnhöfer, Engler and Dümig2017) and use the OECD’s economy-wide product market regulation (PMR) index to measure the extent of regulatory barriers to competition (OECD 2023). Because the index is five-yearly, I linearly interpolate the gaps in between. This index is measured on a zero-to-six scale, with a lower value reflecting fewer regulatory barriers to competition. Instead of the composite score on overall barriers to competition (which captures various policy areas not directly connected to market concentration, such as the scope and governance of state-owned enterprises, command and control, and communication and simplification of rules and procedures), I use the disaggregated scores for specific policy mechanisms that more transparently affect the extent of market concentration. In particular, I focus on (1) administrative burdens on startups, (2) regulatory protection of market incumbents, and (3) regulatory exemptions of large firms from antitrust investigation.
I start by providing some basic descriptive statistics on how the making of competition policy by left governments differs between low- and high-rate environments. For example, when interest rates are above median (2.6 percentage points), the mean score on the index of administrative burden on startups for governments with above-median left representation is 2.07, but when interest rates are below median, the mean score is only 1.83, a statistically significant difference that is equal to almost one half of the standard deviation of the index. Similarly, on regulatory protection of incumbents, the mean scores are 3.49 and 3.03 under high and low interest rates respectively; on regulatory exemptions from antitrust, the scores are .97 and .74 respectively. On both policy measures, the difference in means is again statistically significant. On all three policy measures, in other words, left governments are associated with lower regulatory barriers to competition when interest rates are lower. Next, I examine this pattern more formally through the specification of interaction effects between left partisanship and interest rates.
According to hypothesis H1, as interest rates become lower, left government partisanship will become increasingly associated with lower regulatory barriers to competition. In other words, on each of the three regulatory barriers above, there should be a positively signed interaction effect between left partisanship and interest rates. Before estimating the interaction effects, I check for potential violation of the linear interaction effect (LIE) assumption underlining multiplicative interaction models, which states that the marginal effect of left partisanship on competition policy should change with the level of interest rates at a constant rate. More specifically, I follow Hainmueller et al. (Reference Hainmueller, Mummolo and Xu2019, p. 170) and split the data into three equally sized groups based on the level of interest rates and then plot regulatory policy scores against left partisanship for each of the three groups, creating a linear interaction diagnostic (LID) plot. Figure 2 displays the LID plot for regulatory protection of market incumbents, against left government seat share (share of seats in parliament held by left parties as a percentage of all seats held by the government), overlaying linear and LOESS fits. As the figure shows, the conditional expectation of regulatory protection given left partisanship is indeed mostly linear in all three subsamples by the level of interest rates. LID results are similar for the other two policy variables (administrative burden on startups and antitrust exemptions).
Figure 2. Hainmueller et al.’s (Reference Hainmueller, Mummolo and Xu2019) linear interaction diagnostic plot.
In Model 4, the dependent variable is administrative burden on startups. Consistent with expectation, the interaction between left partisanship and long-term interest rates (instrumented with the proportion of workers aged 55 and over) is indeed positively signed: the lower interest rates, the more strongly left partisanship is associated with less administrative burdens on startups. This core finding of a positive interaction remains unchanged in Model 5, where I instrument interest rates alternatively with the ratio of older workers to younger/middle-aged workers following Acemoglu and Restrepo (Reference Acemoglu and Restrepo2022). Model 6 turns to regulatory protection of market incumbents as the outcome of interest, again yielding a positive interaction effect: the lower interest rates, the more strongly left partisanship is associated with less regulatory protection of incumbents. The core finding also remains unchanged in Model 7, which measures left partisanship alternatively as the center-left party share of parliamentary seats. Using Model 6 as an example, Figure 3 plots the marginal effect of one standard deviation increase in left partisanship on the extent of regulatory protection of incumbents, conditional on the level of long-term interest rates and overlaid with a histogram of the distribution of interest rates across their range.
Figure 3. Effect of left partisanship on regulatory protection of market incumbents.
As the figure shows, at relatively high interest rates the confidence interval straddles the zero line, implying that left partisanship has no distinctive association with the extent of regulatory protection of incumbents, consistent with the vanishing of partisanship effect on competition policy often highlighted in the existing literature. However, as interest rates fall, a novel partisanship effect starts to emerge: left partisanship becomes increasingly associated with a lessening of regulatory protection. For example, at an interest rate of .5%, each standard deviation increase in left partisanship is associated with a reduction of the regulatory protection index by .6, and when interest rates fall to −.5%, the same increment of increase in left partisanship will lower the index by .08, which is close to one third of standard deviation in the index.
In Table A1 of the online Appendix, I test the robustness of the findings to the control variables by starting with a barebones model with no controls (Model 1) and then adding groups of controls organized by theme: economic openness (trade and capital openness in Model 2), macroeconomic conditions (inflation, unemployment, and economic growth in Model 3), and institutional constraints (union density, wage bargaining coordination, and constitutional veto points in Model 4). Because four of the countries in Figure 1 (Ireland, Japan, Spain, and Switzerland) did not exhibit a pattern of consistently falling long-term rates, I also test the robustness of the findings with a leave-one-out analysis, where one of these four countries is dropped from the estimation at a time (Models 5 through 8 in Table A1). Across all eight models, the core finding of a positive interaction between left partisanship and interest rates remains robust.
Back to Table 1 in the main text, Model 8 turns to regulatory exemptions of large firms from antitrust investigation as the outcome of interest, and presents estimates from a mixed-effects model. On antitrust exemptions, there is again a positive interaction between left partisanship and interest rates. I plot this relationship in Figure 4, which displays a pattern similar to regulatory protection of incumbents in Figure 3: when interest rates are relatively high, left partisanship has no discernible effect on antitrust exemptions, but as interest rates fall, left partisanship becomes increasingly associated with a decline in regulatory exemptions from antitrust investigations.
Figure 4. Effect of left partisanship on regulatory exemptions from antitrust investigations.
Besides the three core policy mechanisms analyzed so far, the disaggregated PMR scores also cover various policies less likely to directly affect market concentration and thus should not exhibit the distinctive interaction effect for left partisanship seen above. I test these policies in Table A2 of the online Appendix. Model 1 of Table A2 estimates administrative burdens on corporations, which, unlike administrative burdens on startups, are less specifically targeted at either large or small firms. Models 2 through 4 estimate various policies that mainly reflect regulatory distortions generated by state intervention (direct state control over business enterprises in Model 2, state involvement in business operations in Model 3, and price controls by the government in Model 4). Models 5 through 7 estimate various policies that mainly restrict the international flow of finance and goods (barriers to foreign direct investment in Model 5, tariff barriers in Model 6, and differential treatment of foreign suppliers in Model 7). None of the seven models exhibit the distinctively positive interaction effect between left partisanship and interest rates seen in estimates of the three core policies that more directly affect market concentration.
To further increase the robustness of findings, I next turn to the Liberalization Database (Armingeon et al. Reference Armingeon, Baccaro, Fill, Galindo, Heeb and Labanino2019) as an alternative data source for competition policy. While this database helps extend the time period further back to 1973, its data is regrettably more dated because the coverage stopped at the year 2013. This database provides an index for the degree of overall restriction in product market competition as well as an index for overall liberalization. To maintain consistency with the interpretation for the OECD’s PMR index, I calculate a score for overall restriction on product market competition by subtracting the liberalization index from the restriction index, so that a lower score implies less regulatory restrictions to competition. Model 9 of Table 1 in the main text presents the mixed-effects model of overall restrictions on product market competition, and consistent with earlier findings, the interaction between left partisanship and interest rates continues to be positively signed. The finding remains unchanged in Model 10, where I instrument long-term rates with the proportion of workers 55 and over, and in Models 11 and 12, which respectively estimate two disaggregated policy mechanisms more directly related to market concentration: regulatory forbearance on abuse of dominant market positions, and entry barriers.
Based on findings in this section, policies embraced by left governments that reduce regulatory barriers to competition may involve either lighter regulation (such as less administrative burden on startups, less regulatory protection of incumbents, or less entry barriers) or tougher regulation (such as less exemptions from antitrust investigations or less regulatory forbearance on abuse of dominant market positions). This pattern again implies that the support by left governments for market competition is more focused on constraining market concentration than a general ideological shift to deregulation.
Empirical analysis: distributive outcomes
In Table 2, the main distributive outcome of interest is the labor share of income, measured as total labor compensation as percentage of gross corporate value-added in the economy, based on data from the EU KLEMS Growth and Productivity Accounts (September 2017 Release).Footnote 2 In estimating the labor share of corporate income, I control for macroeconomic conditions (trade and capital account openness, economic growth, unemployment, and inflation), labor market institutions (union density and wage bargaining coordination), as well as the amount of government spending and government budget deficit. All these control variables are drawn from the Comparative Welfare State Dataset (Brady et al. Reference Brady, Huber and Stephens2020). Before analyzing the labor share, I start by briefly examining market concentration. In Model 1 of Table 2, the dependent variable is the Herfindahl-Hirschman index of market concentration by firm value-added, available from CompNet.Footnote 3 Consistent with expectation, there is a positively signed interaction effect between left partisanship and interest rates: as interest rates become lower, left government partisanship indeed becomes increasingly associated with a reduction in market concentration. Next, I examine the effect of market concentration on the labor share. In Model 2, the dependent variable is the labor share of income measured as percentage of corporate gross value-added, and consistent with expectation, the coefficient for market concentration is negatively signed: the higher market concentration, the lower the labor share of corporate income. For ease of interpretation, Model 2 directly reports the effect of one standard deviation increase in market concentration. According to Model 2, each standard deviation increase in market concentration will reduce the labor share of income by .79 percentage points. The finding remains similar when I measure market concentration as the combined revenue share of the top ten firms in the economy.
Table 2. Left partisanship and the labor share of income
* p < 0.1; ** p < 0.05; *** p < 0.01.
According to hypothesis H2, as interest rates become lower, left government partisanship will become increasingly associated with a higher labor share of income. In other words, on the labor share, there should be a negatively signed interaction effect between left partisanship and interest rates. As before, I first check for potential violation of the LIE assumption with Hainmueller et al.’s (Reference Hainmueller, Mummolo and Xu2019) LDP plot of labor share against left government seat share in Figure 5. As the figure shows, the conditional expectation of labor share given left partisanship is indeed mostly linear in all three subsamples by the level of interest rates.
Figure 5. Hainmueller et al.’s (Reference Hainmueller, Mummolo and Xu2019) linear interaction diagnostic plot.
In Model 3, the dependent variable is the labor share of income and consistent with expectation, the interaction between left partisanship and interest rates is indeed negatively signed. This finding continues to hold when I alternatively measure long-term rates following the approach of Mehrotra and Sergeyev (Reference Mehrotra and Sergeyev2021) by subtracting a three-year moving average of the inflation rate from Jordà et al.’s (Reference Jordà, Schularick, Taylor, Eichenbaum and Parker2017) nominal long-term interest rates data (Model 4), or alternatively measure left partisanship as the cumulative left share of government seats (Model 5).
While left government partisanship is treated as an independent variable in estimates of labor share, it is also possible for a higher labor share, as a distributive outcome favoring workers, to boost the popularity of left governments, thus making left partisanship endogenous to the labor share. In this context, the literature highlights two potential instruments for left government partisanship. The first is the proportionality of the electoral system. On one hand, this institutional feature of the electoral system is largely exogenous to the labor/capital division of corporate income in the economy; on the other hand, as Iversen and Soskice (Reference Iversen and Soskice2006) point out, more proportional electoral systems tend to result in more electorally successful center-left parties because such systems are more effective in protecting the pivotal median voter from unilateral attempts by the poor to impose higher taxes. A second instrument for left partisanship identified by the literature is the ‘skew’ of wage distribution (the top/median wage ratio divided by the median/bottom wage ratio). As Lupu and Pontusson (Reference Lupu and Pontusson2011) point out, the more skewed the structure of wage distribution, the closer the median voter feels toward the poor relative to the rich, thus increasing the likelihood of supporting center-left parties. Meanwhile, the labor/capital contest for income is a process largely orthogonal to how income is distributed within labor. For example, Lupu and Pontusson (Reference Lupu and Pontusson2011) show that a highly skewed wage distribution can be found both in (Scandinavian) countries with strong labor bargaining power versus capital and (Anglo-Saxon) countries with weak labor. Empirically, the largely exogenous variation of ‘skew’ relative to the ‘labor share’ is also reflected in the very weak correlation between these two measures in the data (r = .08).
Based on discussions above, Model 6 in Table 2 re-estimates Model 5 by instrumenting left government partisanship with the proportionality of the electoral system (which I measure using the mean electoral district magnitude) and the skew of wage distribution (which I follow Lupu and Pontusson (Reference Lupu and Pontusson2011) and measure as the 90th to 50th wage ratio divided by the 50th to 10th wage ratio). Wald F-test statistics in this 2SLS estimate is again above the 10 percent critical value threshold suggested by Stock and Yogo (Reference Stock, Yogo and Andrews2005), rejecting the null hypothesis of weak instruments. In Model 7, I use Gallagher’s (Reference Gallagher1991) disproportionality index as an alternative measure for the electoral system proportionality instrument. In both Models 6 and 7, the core finding of a negative interaction between left partisanship and interest rates continues to hold. Figure 6 plots the marginal effect of one standard deviation increase in left partisanship on the labor share of income, conditional on the level of interest rates. The downward interaction slope indicates that the lower the interest rates, the more strongly left partisanship is associated with a higher labor share of income. For example, when interest rates are elevated at 5%, left partisanship is unrelated to the size of the labor share. However, once interest rates drop below 1.5%, left partisanship starts to become increasingly associated with a higher labor share. For example, at 1% interest rate, each standard deviation increase in left partisanship will boost the labor share of income by .38 percentage points; when interest rates fall to −.5%, the same increment of left partisanship will increase the labor share by .66 percentage points of corporate gross value-added.
Figure 6. Effect of left partisanship on labor share of income.
In Table A3 of the Appendix, I again test the robustness of the findings, first by adding groups of controls stepwise to a barebones model, and then through a leave-one-out analysis, where one of the four countries with unusual interest rate dynamics (Ireland, Japan, Spain, and Switzerland) is dropped from the estimation at a time. Across the eight models of labor share in Table A3, the core finding of a negative interaction between left partisanship and interest rates remains consistent.
As today’s low-rate environment delivers an increasingly favorable left partisanship effect for labor, it should also deliver an increasingly adverse left partisan effect for capital. To assess this prediction, the next two models in Table 2 present mixed-effect models of corporate gross saving (i.e., undistributed profits). The data on corporate gross saving come from Chen et al. (Reference Chen, Karabarbounis and Neiman2017), which I convert to percentage of corporate gross value-added. In both models, the interaction effect between interest rates and left partisanship (measured, respectively, as left government seat share in Model 8 and left parliamentary seat share in Model 9) is indeed positively signed: as interest rates become lower, left partisanship is increasingly associated with lower corporate saving.
Figure 7 plots the marginal effect of one standard deviation increase in left partisanship on gross corporate saving, conditional on interest rates. As the figure shows, at elevated levels of interest rates, left partisanship is unrelated to the size of corporate gross saving. However, when interest rates fall to 1%, each standard deviation increase in left partisanship will reduce corporate gross saving by 1.47 percentage points; at zero interest rate, the same increment of left partisanship will reduce corporate saving by 1.81 percentage points of corporate gross value-added. In other words, falling rates will indeed turn left partisanship into an increasingly adverse distributive force against capital.
Figure 7. Effect of left partisanship on corporate saving.
Conclusion
Scholars argue that while left partisan governments traditionally support stronger market regulation, this partisanship effect has started to vanish as left governments converge with the right in embracing deregulation, potentially resulting in higher inequality (Keller and Kelly Reference Keller and Kelly2015; Mudge Reference Mudge2018; Polacko Reference Polacko2021; Witko Reference Witko2014). Against this backdrop the current paper argues that, instead of vanishing, the partisanship effect on competition policy has intensified, but in a novel direction: left partisan governments have become stronger defenders of product market competition than other partisan governments. Furthermore, instead of retreating from redistribution, this new association between left partisanship and market competition has delivered new distributive gains for labor.
In understanding this outcome, I have highlighted the role played by the depressed long-term interest rates across rich countries today: low interest rates spark a rise in market concentration, which puts downward pressure on the labor share of income. Because stronger product market competition can counter this force, by boosting competition left partisan governments can defend the labor share of income, and thus revitalize their traditional redistributive agenda in today’s more challenging economic environment.
Consistent with this argument, I have shown empirically that, as long-term interest rates fall, center-left parties become more willing to voice opposition to market concentration and support for redistribution in their manifestos, but no more willing to express support for a laissez-faire economy. Furthermore, left government partisanship also becomes increasingly associated with policies that boost product market competition, which involve lighter regulation in some areas but tougher regulation in others. In terms of distributive outcome, as interest rates fall, left partisanship has become increasingly associated with less market concentration, a higher labor share, and lower corporate saving.
There are various directions for future research. First, by countering the rise of market concentration, left governments may potentially tap into a new redistributive coalition ‘against corporate behemoths’. It will be interesting to more systematically study where this coalition differs from the traditional constituency of left governments. For example, scholars find that product market competition tends to reduce job security but boost aggregate employment, hence benefiting labor market outsiders relative to insiders (Amable and Gatti Reference Amable and Gatti2004; Nicoletti and Scarpetta Reference Nicoletti and Scarpetta2005). In this light, it is important to examine if left governments’ pursuit of product market competition can address the ‘insider/outsider divide’ that traditionally characterized social democratic electoral constituencies (Rueda Reference Rueda2005, Reference Rueda2014).
Furthermore, besides boosting the labor share, a lowering of market concentration may also enhance the position of small/median firms and put downward pressure on consumer prices. From this perspective, it is important to explore distributive implications beyond the income dimension, and see if the current era of low interest rates has helped left governments reach out to small firms and consumers as additional recruits to their redistributive constituency. Finally, as various scholars point out, center-left parties are also trying to offset the decline of their traditional electoral constituency by reaching out to middle-class sociocultural professionals with progressive social cultural policy positions (Abou-Chadi and Wagner Reference Abou-Chadi and Wagner2020; Gethin et al. Reference Gethin, Martínez-Toledano and Piketty2022). However, as Engler and Zohlnhöfer (Reference Engler and Zohlnhöfer2019) find, culturally progressive professionals tend to favor regulation and oppose market competition. As a result, it will also be important to examine whether left governments’ pursuit of market competition as a redistributive strategy may complicate their pursuit of culturally progressive voters.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1475676525100406
Data availability statement
Data and replication files are available here: https://uwaterloo.ca/political-science/profiles/jingjing-huo
Funding statement
There was no external funding for this research.
Competing interests
I do not have competing interests to declare.
Open access
