Linking electoral realignment to welfare politics: an assessment of partisan effects on active labour market policy in post-industrial democracies

This article examines the political determinants of the variations in active labour market policies across advanced democracies. Specifically, it investigates the conditions under which a welfare state accommodates rather than disregards the interests of labour market outsiders. Relying on the literature on post-industrial electoral realignment, this article argues that ideological orientations not only in socio-economic but also in sociocultural dimensions dictate the policy preferences of political parties for labour market programmes. This study then hypothesizes that libertarian governments are more likely than authoritarian governments to support human capital formation of labour market outsiders. An analysis of cabinet-based periodization data of 21 advanced industrialized countries from 1985 to 2017 shows that left- and right-libertarian governments favour public spending on active labour market programmes, thereby supporting this study’s hypothesis. Furthermore, it also reveals that while left-libertarian governments increase expenditures for direct job creation schemes, right-libertarian ones do so for employment assistance and training programmes.


Introduction
This study investigates the political determinants of variations in labour market policies across advanced democracies. During the past few decades, mature welfare states have been required to balance security for vulnerable workers and flexibility in the labour market. The approaches they use to attain this balance vary across governments and countries: some reinforce public investment in human capital formation through active measures, others provide both positive and negative incentives regarding employment for the jobless and inactive, while still others build a dualized labour market by protecting regular contract workers' tenure and making atypical employment flexible. How can we understand these diverse policy responses to shared labour market issues among post-industrial democracies?
Heretofore, research on labour market programmes has centred on political partisanship. Conventional partisan models commonly identify pro-welfare parties as a monolithic block: while the left defends passive and/or active labour market programmes (ALMPs), the right does not (Allan and Scruggs 2004;Boix 1998;Huo et al. 2008;Iversen and Stephens 2008). However, the theoretical and empirical foundations of this simple assertion are not necessarily solid. Assuming that ALMPs deliver net benefits only to labour market outsiders, the insider-outsider theory claims that labour market insiders and outsiders have distinct policy preferences for labour market programmes, and that while leftist governments promote employment protection for insiders, they rarely influence ALMPs (Rueda 2005(Rueda , 2006(Rueda , 2007. These seemingly contradictory arguments indicate that comparative political economists should update their unidimensional concept of political partisanship to incorporate the insider-outsider divide into their theoretical framework. Although the political coalitional approach recognizes the significance of the insider-outsider divide and probes the conditions under which policy demands from outsiders are accommodated (Huo 2009;Iversen and Soskice 2015;Martin and Thelen 2007;Thelen 2014), it either disregards partisan effects (Martin and Thelen 2007;Thelen 2014) or adheres to unidimensional partisan models (Iversen and Soskice 2015).
Reflecting the theoretical development of the post-industrial political realignment school (cf. Beramendi et al. 2015;Häusermann 2010a), this article introduces a model to explain the policy consequences of party realignment for active labour market policies in the post-industrialization era. It argues that political parties now compete with one another in a two-dimensional policy space composed of the first 'socio-economic' and second 'sociocultural' dimensions. Further, policy positions in both dimensions affect the preferences of political parties regarding human capital formation and labour market policies in post-industrial democracies. The model predicts that while economically left parties are more likely than economically right parties to promote the interests of labour as a whole, culturally libertarian parties have more incentives to cater to precarious and vulnerable workers than authoritarian parties. As a result, a libertarian government is more likely to support human capital formation of labour market outsiders.
To test the theoretical model, this study empirically explores the effects of a government policy position in a two-dimensional policy space on ALMPs. It then measures the policy position of each cabinet in this two-dimensional policy space by constructing an indicator based on the data of the Comparative Manifesto Project (CMP) (Budge et al. 2001;Klingemann et al. 2006;Volkens et al. 2015Volkens et al. , 2019. It then assesses the effects of government partisanship by analysing cabinet-periodized data from 21 advanced industrialized democracies from 1985 to 2017. This reveals that libertarian governments tend to enhance public expenditures on ALMPs, as well as that while left-libertarian governments enlarge their spending on direct job creation schemes, right-libertarian ones elevate their spending on employment assistance and training programmes. These empirical results suggest that the sociocultural dimension might be more influential for active labour market policies than the socio-economic one in post-industrial democracies. The latest research in the postindustrial electoral realignment approach rarely connects party positions with labour market policy outputs in its two-dimensional framework, although it identifies individual-level preferences for active measures, party-level preferences for welfare programmes, and the congruence between them (e.g., Garritzmann and Schwander, forthcoming;Garritzmann et al. 2018;Geering and Häusermann 2013;Häusermann 2018;Häusermann et al. 2016;Neimanns et al. 2018). Thus, the findings of this study make a meaningful contribution to the literature.
This article is structured as follows. The next section reviews the current literature and presents a two-dimensional model of partisan differences with respect to labour market programmes. Following that, the data for the quantitative analyses are introduced, and the analytical methods used in the regression analyses explained. The article then highlights the results of the study's quantitative analyses, and the final section summarizes and discusses the implications of this study for comparative welfare state research.

Post-industrial party competition and active labour market policy
The majority of previous studies on passive and active labour market programmes concur that the presence of social democratic governments augments public spending, the coverage, and/or the generosity of these programmes (Boix 1998;Huo et al. 2008;Iversen and Stephens 2008;Janoski 1994). Boix theorizes a partisanship approach to human capital investment policy, maintaining that left and right parties have different priorities and employ distinct supply-side economic policies to maximize growth and reduce unemployment in global competition. If this argument were valid, partisan differences along the economic left-right dimension would lead to distinct labour market policies. Rueda (2005Rueda ( , 2006Rueda ( , 2007 presents his insider-outsider approach as an alternative to the conventional left-right partisan model and claims that government partisanship has no effect on passive and active labour market programmes. He argues that while social democratic parties have incentives to protect industrial workers as their core constituents (insiders), these parties are less enthusiastic about promoting the interests of precarious-such as fixed-term contract, temporary agency, and parttime-workers (outsiders). Rueda states that the presence of leftist governments elevates the strictness of employment protection legislation but does not influence the generosity of unemployment benefits and ALMPs because they yield few benefits but impose greater tax burdens on labour market insiders with employment protection (see also Gaston and Rajaguru 2008).
Acknowledging that the insider-outsider divide exists in post-industrial societies, the political coalitional approach tries to specify the conditions under which governing coalitions avoid dualism and stand for the interests of actual and potential outsiders (Huo 2009;Iversen and Soskice 2015;Martin and Thelen 2007;Thelen 2014). Huo (2009) argues that inclusive industrial relations systems facilitate third-way reforms that activate unemployed and underemployed young people and women. Adding to this perspective, Martin and Thelen (2007) and Thelen (2014) maintain that a large and strong public sector also contributes to forging relatively inclusive coalitions, enabling social investment in labour market outsiders. Critiquing these corporatist arguments, Iversen and Soskice (2015) point out that Thelen and others hardly discuss party politics, which in their view dictates the formation of inclusive corporatist institutions and large public sectors. Instead, Iversen and Soskice (2015) claim that the combination of electoral and party systems determines whether skilled workers in manufacturing sectors (i.e., insiders) and low-skilled workers in service sectors (i.e., outsiders) are capable of forming centre-left coalitions, which is fundamental to maintaining inclusive governing coalitions between government, business, and unions in post-industrial societies.
This study does not deny the positive function of macro-corporatist institutions and large public sectors in catering to labour market outsiders, and it agrees with Iversen and Soskice (2015) that party politics conditions the survival of this kind of egalitarian industrial relations system. In that sense, the latter is an epiphenomenon of the former. However, Iversen and Soskice's (2015) approach is problematic in that when they discuss decoupling a class coalition between skilled and semi-skilled workers as the social foundation of egalitarian capitalism, they ignore the fact that the composition of 'low-skilled workers' has changed between the industrial and post-industrial era. That is, while both semi-skilled and skilled workers were mostly native, male citizens under the Fordist economy, low-skilled workers are over-represented by young, female, immigrant, and/or ethnic minority workers under the knowledge economy (cf. Chauvel 2009: for the young; Emmenegger and Careja 2012: for migrants; Esping-Andersen 2009: for women). The labour market policies for low-skilled workers are expected to reflect governing parties' policy orientation towards these disadvantaged groups in the post-industrial period. Nevertheless, Iversen and Soskice's (2015) unidimensional left-centre-right approach is incapable of capturing governing coalitions' attitudes towards these demographic groups.
How can we reconcile the contradiction between conventional partisan and insider-outsider approaches, and simultaneously comprehend the conditions under which governing coalitions include rather than exclude the interests of labour market outsiders? Rueda's (2005Rueda's ( , 2006Rueda's ( , 2007 critique of the conventional left-right approach is valid insofar as he states that social democratic parties care less about passive and active labour market programmes when they solely intend to extract electoral support from organized union workers who are protected by employment protection legislation. From the 'Hibbsian' point of view (cf. Hibbs 1977;Wenzelburger and Zohlnhöfer, forthcoming), which considers political parties to be entities transmitting the preferences of their voters into public policies, neither left nor right parties have incentives to promote ALMPs for atypical and precarious workers and the unemployed. However, from the perspective of the 'agency-based' approach (Wenzelburger and Zohlnhöfer, forthcoming), a political party is an ideologically motivated, policy-seeking camp as well as a vote-and office-seeking electoral vehicle catering to aspiring politicians (cf. Strom 1990). While labour market outsiders cannot be the power base because these workers are extremely diverse and less organized in politics, expanding ALMPs for them rarely hinders the electoral prospect of political parties. ALMPs are less conspicuous and, unlike public pension and health care, hardly grow tangible enough in the budget size to bother unionized workers and middle-class citizens with their tax burdens (Bonoli 2013: 170-172). For instance, while mature welfare states spent 8.8% and 6.5% of GDP on public pension and health care, respectively, they expended only 0.6% on ALMPs on average among 21 Organisation for Economic Cooperation and Development (OECD) countries in 2015 (Brady et al. 2020). Even though ALMPs are less likely to be pushed by the electoral mandate, these characteristics of ALMPs leave political parties the leeway to manoeuvre these programmes for their ideological goals. When the ideological orientations of governing parties have an affinity with an inclusive coalition incorporating the interests of labour market outsiders, these parties are more likely to expand ALMPs. In this sense, this article stresses the supply side rather than the demand side of politics.
Relying on the literature concerning post-industrial electoral realignment (cf. Beramendi et al. 2015;Häusermann 2010a;Kitschelt 1994Kitschelt , 2004Kriesi et al. 2008), this study argues that the ideological position of each political party in the second 'sociocultural dimension' is one significant determinant of its policy preferences regarding labour market programmes for outsiders. As Kitschelt (1994Kitschelt ( , 2004 suggests, the party competition space has been transformed from a unidimensional space into a two-dimensional one, in which the socio-economic left-right axis crosses over the sociocultural libertarian-authoritarian axis, since the emergence of post-materialism in post-industrial democracies. 1 As Häusermann's (2010a, b) works suggest, not only the socio-economic but also the sociocultural dimension influences the configuration of welfare state and government intervention. The socio-economic dimension dictates the extent to which each political party prefers the welfare state to intervene in and modify material distribution in the market economy. In other words, this dimension governs the size of the welfare state. In contrast, the sociocultural dimension defines each party's preferences regarding the appropriate unit of social protection. Whereas an authoritarian position assigns a higher priority to protecting the heads of a nuclear family against social risks because of its particularism and gender stereotypes, a libertarian position treats individuals as the fundamental unit of social protection because of its universalism and egalitarianism. The sociocultural dimension affects to whom the welfare state devotes its resources.
This study proposes a two-dimensional model of partisan differences with respect to labour market programmes. In the first dimension, a political party faces a choice concerning the extent to which the state intervenes in the labour market and then protects labour against various social risks. As the conventional partisan model suggests, while an economically right party trusts the market mechanism and prefers less intervention in the labour market, an economically left one puts more faith in the role of public sectors and aspires to safeguard workers against labour market risks.
In the second dimension, a political party chooses which social groups should be protected through labour market programmes. A libertarian party opts for a policy strategy to improve human capital through extensive training schemes, active support for job searches, and job experiences. This type of party supports policies protecting female, young, and/or ethnic minority workers against newly emerging social risks such as irreconcilability between paid work and care obligations and low-education/low-skill attainment. The goal of the libertarian party is to enhance individual citizens' employability so that they can become self-reliant as autonomous individuals. For libertarian parties, ALMPs are an important policy instrument in reconciling protection against job loss or precarious jobs and preventing inactivity under the rubric of 'social investment' (Bonoli 2013: 182-183;Gingrich and Häusermann 2015: 51). Indeed, policy evaluation literature reports that the young and immigrants are over-represented among the unemployed, which means that public spending on ALMPs are more devoted to these demographics than others, and these programmes actually improve the employment prospect of female, young, and minority workers (Bergemann and Van Den Berg 2008;Butschek and Walter 2014;White and Knight 2002).
In contrast, an authoritarian party prefers to assign to families the primary obligation to protect vulnerable citizens. This type of political party prioritizes public policies that safeguard permanent workers-among whom native male breadwinners are dominant-against traditional social risks such as sickness and old age, and indirectly secure a living wage for their wives and offspring (cf. Emmenegger 2011: 339; Esping-Andersen 1999: 23). For instance, during the 2012 French general election, Front National (FN), a typical authoritarian party, asserted re-industrialization to appeal to blue-collar workers, who often hold an open-ended job contract and the status of head of the household (Picot and Menéndez 2019: 910). It is true that blue-collar workers in declining manufacturing sectors might lose their jobs under post-industrialization, automation, and globalization, and ALMPs can help those unemployed workers transfer to a new job. However, authoritarian parties are less concerned with ALMPs because these programmes are universalistic in that they bring benefits to all job seekers regardless of their age, gender, and ethnicity, and do not contribute to preserving the dominant social order. 2 For authoritarian parties, social insurance schemes are at least better to address unemployment issues than ALMPs because the former reflect each recipient's contribution profile and lead to status maintenance (cf. Esping-Andersen 1990).
This study hypothesizes that government partisanship in both the socio-economic and sociocultural dimensions affects the development of ALMPs in post-industrial democracies. As discussed above, while the policy position of the left-right axis determines the size of government intervention in the labour market, that of the libertarian-authoritarian one prescribes the targets of labour market policies. First, the two-dimensional model suggests that left-and right-libertarian governments develop ALMPs in general. These governments aim to help individual citizens to be self-reliant through paid work. Second, it is inferred that left-libertarian governments are the most likely to expand ALMPs directed at investing in the human capital of precarious workers and the unemployed with extensive government programmes. Third, it is assumed that right-libertarian governments aspire to elevate the employability of outsiders by making full use of the labour market. The above argument is formulated into the following three hypotheses: Hypothesis 1 Left-and right-libertarian governments are positive towards ALMPs for labour market outsiders in general.
Hypothesis 2 Left-libertarian governments favour investing in the human capital of labour market outsiders.
Hypothesis 3 Right-libertarian governments favour elevating the employability of labour market outsiders by facilitating the functions of the labour market.

Explained variables
To assess the validity of the hypotheses discussed above, this study examined public spending on ALMPs. OECD publicizes its member countries' public expenditures for ALMPs and categorizes them as (1) public employment service and administration, (2) training, (3) job rotation and job sharing, (4) employment incentives, (5) supported employment and rehabilitation, (6) direct job creation, and (7) start-up incentives (OECD 2019a). The main dependent variable of this study is public spending for ALMPs as a percentage of GDP, which measures each country's annual budget dedicated to all these programmes across 21 advanced industrialized countries from 1985 to 2017. This variable is suitable for assessing the validity of Hypothesis 1.
In addition, this study analyses three auxiliary dependent variables. Previous studies claimed that ALMPs consist of diverse programmes serving different policy purposes; thus, for analysis, the OECD's indicator of ALMPs should be decomposed into subcategories (Bonoli 2013;Vlandas 2013). For example, Bonoli (2013: 109-110) divides the OECD's expenditure data on ALMPs into three subcategories: (1) employment assistance, which contains public employment services and administration, employment incentives, and start-up incentives; (2) training; and (3) direct job creation. Among these, while employment assistance leans towards a pro-market employment orientation, training and direct job creation are balanced between a pro-market employment orientation and investment in human capital (Ibid.: 24). To gauge the impacts of government partisanship in the two-dimensional space on distinctive aspects of ALMPs, this study separately analyses public spending for employment assistance as a percentage of GDP, public spending for training as a percentage of GDP, and public spending for direct job creation as a percentage of GDP. While the former variable is employed for evaluating Hypothesis 3, the latter two are used for appraising Hypothesis 2. All data are derived from the OECD (2019a, b).

Explanatory variables
To test the influence of government partisanship on labour market programmes along the socio-economic and sociocultural dimensions, this study constructed a new dataset using data on government composition and party manifestos. It covers 21 OECD countries from 1960 to 2017. (For the detailed data construction, see Appendix A in the supplementary material.) First, based on the data from the CMP (Budge et al. 2001;Klingemann et al. 2006;McDonald and Mendes 2001: 108-111;Volkens et al. 2019), this dataset generates two indicators-an economically left-right policy position and a culturally libertarian-authoritarian policy positionto estimate each political party's policy position in the two-dimensional party competition space. Second, the dataset estimates the policy positions of each one-party or coalition cabinet in the left-right and libertarian-authoritarian dimensions after each political party is placed in these two dimensions in each election. The measurements of each cabinet's policy position are the average of the coalition party's policy position weighted by its share of seats in the lower house among governing parties.
Following the above-mentioned protocol, this study generated two indicators: the CMP government left-right policy position and CMP government libertarian-authoritarian policy position. To appraise the robustness of the findings attained through the CMP data, this study also generated the Chapel Hill Expert Survey (CHES) government left-right policy position and CHES government libertarian-authoritarian policy position. This was based on the Chapel Hill Expert Survey data (Bakker et al. 2015;Polk et al. 2017), although it covers only 14 European countries from 1999 to 2017. While the former measure averaged the coalition parties' ideological stance on economic issues, the latter averaged their views on democratic freedoms and rights. 3 Both are weighted by each coalition party's share of seats in the lower house among governing parties. The data on each government's partisan composition, seat ratio, and period of duration come from Armingeon et al. (2019b). Since both CMP and CHES data lack the concrete measure of immigration, which has been a dominant issue of the sociocultural dimension in recent decades, these variables concerning government libertarian-authoritarian policy positions might underestimate the variance of government positions in this dimension. Figure 1 plots the values of the variables based on the CMP and CHES data in the two-dimensional policy space, respectively, and its marker symbols indicate the party family type of each cabinet. 4 Although some might suspect that the left-right and libertarian-authoritarian axes are perfectly correlated, as Fig. 1 shows, the CMP and CHES government libertarian-authoritarian policy positions show substantive variations across their left-right policy positions, respectively. The figure also demonstrates that cabinet policy positions vary in each party family. While those experts surveyed by the CHES put cabinets into the preconceived policy positions (e.g., social democratic governments into left-libertarian and conservative governments into right-authoritarian positions) in orderly fashion, actual policy positions presented by party manifestos are scattered across the two-dimensional policy space even in the same party family. These data justify this study's use of positional rather than party family variables for analysis, and simultaneously putting the socio-economic and sociocultural government policy position variables into its regression models.

Control variables
This study incorporates several control variables to properly assess the effects of government ideological positions in the two-dimensional policy space. (For detailed variable definitions and sources, see Appendix B in the supplementary material.) It ian, liberal, religious, conservative, regional and ethnic, populist radical right, and protest. Then, it calculated the relative ratio of each party family type in each cabinet by summing the seat ratio of governing parties in the lower house by party family. The 11 party family variables generated through these protocols are also used for the later analysis in this paper.

Footnote 4 (continued)
is maintained that the European Union (EU) has made ALMPs a central part of its European Employment Strategy since 1997 (Armingeon 2007: 909), and that European countries had the 'activation turn' in the mid-1990s and early 2000s (Bonoli 2013: 167). To address the effects of EU's recommendations to its member countries, this study includes EU Employment Strategy as a dummy variable, which takes 1 for the EU member states from 1997 onwards and 0 otherwise. It also includes three period dummies-the 1990s, 2000s, and 2010s-to control for the effects of time trends.
This study also controls for economic institutional, socio-structural, and business cyclical variables in the regression analyses. Because explained variables concern the intensity of public involvement in ALMPs, following the power resources theory, the study examines the relationship between labour market programmes, on the one hand, and union density and wage bargaining centralization, on the other (cf. Gordon 2015). To assess the influence of each country's exposure to globalization, the KOF Globalization Index is included in the regression equations. To control for the effects of business cycles, the consumer price index, unemployment rate, and real GDP growth rate are added to the models. In addition, this study uses GDP per capita (purchasing power parity) in the regression models to control for the effects of economic affluence. Furthermore, since each government's fiscal conditions constrain the budget devoted to ALMPs (cf. Gaston and Rajaguru 2008), the cumulative debt ratio to GDP is employed in the regression equations.

Research design
To analyse cross-national and temporal variations in government programmes, researchers must design their quantitative analyses with caution because the selection of analytical approaches can drastically change results. Although panel data analyses of time-series and cross-sectional (TSCS) data based on country-year observations are popular in the comparative political economy literature, the structure of annualized TSCS data is problematic for detecting partisan effects (Garritzmann and Seng 2016;Plümper et al. 2005;Schmitt 2016). First, the prevailing procedures of panel data analyses only address short-term changes within countries. The standard approaches in comparative political economy, which analyse annual TSCS data with a lagged dependent variable, panel-corrected standard errors, and/or unit-fixed effects, relate the yearly changes of a dependent variable with independent variables (cf. Beck and Katz 1995). However, since the partisan composition of governments rarely alters annually but usually changes at elections, these approaches are unsuitable for detecting the partisan effects of a government during its tenure.
Second, the standard approaches to annual TSCS data have trouble in discerning the heterogeneous dynamic structures of partisan variables. Since it is implausible that political and macroeconomic variables exert their full impacts on public spending and those effects disappear immediately in any year, previous research has contrived ways to model the lag structure (cf. Beck and Katz 2011;De Boef and Keele 2008;Kittel and Winner 2005). However, these works still assume that the lagged effects of independent variables, which are produced through a lagged dependent variable, are homogenous across macroeconomic, socio-demographic, and political variables and across time. Although this assumption is probably tenable for economic and demographic variables, it is not plausible for partisan variables. While government partisanship might exercise its immediate effects on public spending through the annual budgeting process in some countries and cabinets, it might in others enforce its influences through reform legislation, which usually kicks in several years later. Although capturing the proper impacts of partisan variables requires modelling their unit-specific lag structure across countries and time in the standard setting of annual TSCS data, the same setting does not allow researchers to extract information sufficient to model the unit-specific lag structure appropriately (Schmitt 2016(Schmitt : 1450. To overcome these methodological problems, relying on previous works (Garritzmann and Seng 2016; Schmitt 2016; Vis 2012), this study adopts cabinets rather than country-years as a unit of observation. As mentioned, it is presumed that the lag structure of government partisanship varies across countries and time. While some cabinets might embark on a major reform of ALMPs immediately after the turn of government, others might gradually expand those programmes over several years. Although it is difficult to model the heterogeneous dynamic structure of partisan variables in the country-year settings, using a cabinet-based periodization enables researchers to ignore the heterogeneity because it puts together the effects of those variables by each cabinet's tenure. Furthermore, although we do not exactly know when government partisanship enforces its effects on ALMPs, it is theoretically assumed that each cabinet reflects its policy preferences on these programmes before its term ends, as any-either single-party or coalition-government must yield results to appeal to the constituents before the next election. (For a detailed description of this study's methodological approaches, see Appendix C in the supplementary material.) Table 1 shows the results of the regression models of the four dependent variables on political and socio-economic variables. This study conducted Robust Hausman specification tests (Kaiser 2014), and no models rejected the null hypotheses. Hence, it reports the results of random-effect models, which are consistent with those of fixed-effect models. Readers can find the results of the latter models in Appendix D.

Empirical findings
Models 1 and 2 regress the first differences in public spending for ALMPs as a percentage of GDP on government partisanship and control variables without and with the interaction term, respectively. The coefficients of partisan variables indicate that while government left-right policy positions have no influence on public expenditures for ALMPs, government libertarian-authoritarian policy positions clearly impact them. In both models, the latter have a statistically significant negative coefficient, which means that libertarian governments increase their budget devoted to ALMPs.
Since this study's hypothesis contends that the combination of left-right and libertarian-authoritarian policy positions affects public spending for ALMPs, it  calculated the interaction effects between the government left-right policy position and government libertarian-authoritarian policy position variables using model 2's variance-covariance matrix and presented the results in Fig. 2a (cf. Kam and Franzese 2007), demonstrating that a government's sociocultural libertarian-authoritarian policy position certainly influences the total expenditures for ALMPs. The marginal effects of government libertarian-authoritarian policy position remain negative across the economically left-right policy spectrum, which means that a libertarian government increases its spending on ALMPs,. These effects are statistically significant among moderate governments in the socio-economic dimension. These results confirm Hypothesis 1, which contends that left-and right-libertarian policy positions are positive for ALMPs. Models 3 to 8 analyse the political determinants of the subcategories of ALMPs. These models regress the first differences in public spending for employment assistance, training programmes, and direct job creation schemes as a percentage of GDP. As Table 1 shows, while government left-right policy positions indicate no statistically significant impacts, government libertarian-authoritarian policy positions denote significant coefficients for employment assistance, training, and direct job creation. Figures 2b-2d shows that a moderate-to right-libertarian government raises its expenditures for employment assistance and training programmes, while a moderate-to left-libertarian government enlarges its spending on direct job creation. The latter result agrees with Hypothesis 2. As Bonoli (2013: 22-26) suggests that job creation schemes have a weak pro-market orientation, economically left governments are more inclined to offer a sheltered job in the public sector and less likely to force the jobless to take a job in the free market. In addition, Fig. 2b also validates Hypothesis 3, indicating that moderate-to right-libertarian governments facilitate job search efforts and job experiences in the labour market with employment assistance programmes. By contrast, Fig. 2c partially rejects Hypothesis 2 because it indicates that economically right rather than left governments increase public spending for training programmes. Although training programmes are thought of as encouraging investment in human capital (ibid.: 22-26), these results suggest that moderate-to right-libertarian governments are willing to integrate the jobless into the labour market with these programmes despite their burdens on the public purse. It seems that right-libertarian governments value the pro-market orientation of training programmes more than left-libertarian ones do. Since sociocultural libertarian positions are a necessary condition for expanding any of the subcategories of ALMPs, these results still support this study's main claim that the second, 'sociocultural' dimension is an important determinant of ALMPs.
To check the robustness of the findings, this study also conducted regression analyses by replacing the government policy position variables based on CMP with those based on CHES. (See Appendix E in the supplementary material.) Because the CHES covers only 14 West European countries from 1999 onwards, the number of country-cabinet observations is limited to 60 or 66 in the analyses. This seems to enlarge the standard errors of government policy position variables compared to their coefficients, and makes it difficult for these to show statistical significance. However, as is the case with the above analyses, the government libertarian-authoritarian policy position variable consistently indicates negative coefficients-except for training programmes-across models without an interaction term. These results correspond to the findings of the CMP data analyses.
Finally, this study also included party family variables in the regression models. 5 (See Appendix F in the supplementary material.) On the one hand, those models excluding party position variables indicate that no party family variable-except that for protest parties-has statistically significant effects on ALMPs. 6 On the other hand, those models incorporating both party family and policy position variables demonstrate that a cabinet policy position in the sociocultural dimension still has significant impacts on ALMPs-except for training programmes-even after controlling for the effects of party family types. As Fig. 1 shows, those cabinets categorized in the same party family (e.g., social democratic, conservative, and liberal) have remarkably heterogenous policy positions in the two-dimensional ideological space. These results suggest that the actual policy position rather than the party label of each government determines its preference for ALMPs.

Conclusion
This article explored the conditions under which governing parties promote ALMPs. It argued that political parties compete with each other regarding unemployment issues and human capital formation policy in the sociocultural as well as socio-economic dimensions, and that each political party maintains different policy preferences regarding labour market policies according to its ideological position in the two-dimensional space. The empirical evidence of the analyses of the cabinet-periodized TSCS data from 21 OECD countries in this study demonstrates, contrary to the conventional partisan model, that the promotion of active labour market programmes requires a libertarian government on the sociocultural dimension. On one hand, the results showed that left-and right-libertarian governments raise the budgets for ALMPs in general, which supports this study's hypothesis. On the other, the results also revealed that while right-libertarian governments increase public spending for employment assistance and training programmes, left-libertarian governments do so for direct job creation. These results indicate that the sociocultural dimension is more influential in determining each government's policy preferences for labour market programmes than the socioeconomic dimension. This empirical evidence may reflect the fact that some social democratic parties have adopted the so-called third-way approach, moderating their socio-economic policy positions and reinforcing ALMPs to include the inactive and underemployed in the mainstream labour market (cf. Green-Pedersen et al. 2001;Huo 2009;Keman 2011).
This study has important implications for the literature on comparative political economy. Its two-dimensional approach can enrich our understanding of politics concerning insider-outsider cleavage. Rueda (2005Rueda ( , 2006Rueda ( , 2007 argues that contemporary social democratic parties are severely constrained by the interests of labour market insiders, and then less inclined to corroborate public programmes for outsiders. This claim amounts to virtually no supporters for labour market outsiders in welfare politics. Hence, as Gordon (2015: 84) points out, Rueda's theory leaves an empirical conundrum since these programmes persist even after employment protection legislations insulated insiders from labour market risks. In contrast, this study's empirical findings suggest that a libertarian party caters to outsiders with various policy measures, relying on its ideological position in the socio-economic dimension. These empirical results highlight the possibility of resolving the above puzzle by directing our attention towards the political dynamics in the sociocultural dimension.
Furthermore, this study's two-dimensional approach can advance our understanding of multidimensional aspects of contemporary welfare politics in general (Häusermann 2010a; Häusermann et al. 2013;Iversen and Goplerud 2018). Since postindustrialization decoupled skilled workers-the then median voter in the Fordist economy-from semi-and unskilled labour, the unidimensional party competition model hardly expects that a redistributive coalition be forged and the welfare state expanded in advanced industrialized economies (Iversen and Soskice 2015). However, the two-dimensional model allows us to conceptualize a new welfare coalition forming that cuts across the conventional left-right dimension. The model suggests that a libertarian coalition favours an inclusive welfare state activating and enabling citizens through paid work, while an authoritarian coalition promotes a familialistic and/or chauvinistic one. The two-dimensional approach directs our attention to the types of welfare state beyond its size.
Finally, we note that this study has several limitations. While it established the correlation between government policy positions in the two-dimensional space and ALMPs, it did not directly observe a causal mechanism between them through a case study or other methods. In addition, this study did not theorize and elaborate how a party's ideological positions are aggregated into a government's policy positions. These issues should be addressed in future research.