Political ideology and economic freedom across Canadian provinces

Transcription

1 Political ideology and economic freedom across Canadian provinces Christian Bjørnskov Aarhus School of Business, Aarhus University Niklas Potrafke 1 University of Konstanz October 11, 2009 Abstract: This paper examines how political ideology influenced economic freedom in the Canadian provinces. We analyze the dataset of economic freedom indicators compiled by the Fraser Institute in 10 Canadian provinces over the period and introduce two different indices of political ideology: government and parliament ideology. The results suggest that government ideology influenced labor market reforms: market-oriented governments promoted liberalization of the labor market. Parliamentary ideology did not influence economic liberalization at all. This finding (1) identifies differences between leftist and rightwing governments concerning the role of government in the economy and (2) indicates that ideological polarization concerns governments but less parliamentary fractions in the Canadian provinces. Keywords: economic freedom, taxation, regulations, ideology, panel data JEL Classification: O51, P16, R11, R50 1 Bjørnskov: Aarhus School of Business, Aarhus University, Department of Economics, Frichshuset, Hermodsvej 22, DK Åbyhøj, Denmark. chbj@asb.dk. Potrafke: University of Konstanz, Department of Economics, Box 138 D Berlin, Germany, niklas.potrafke@uni-konstanz.de. 1

2 1. Introduction Government ideology influenced economic policies in the Canadian provinces. Empirical studies suggest that, for example, leftwing governments pursued more expansionary fiscal policies by increasing public expenditures while public revenues appear to be less driven by government ideology. The fact that studies identify ideological effects in fiscal policy is meaningful because the Canadian constitutions and federal politics leave provincial governments with significant room to maneuver. 2 In a similar vein, labor market regulations and legislation are also primarily the abode of provincial governments instead of being governed from the federal level. Yet, the influence of government ideology on labor market regulation and economic liberalization in general has been largely ignored in the political economy literature. Against the background of the sustained interest in the role of political ideology in Canadian economic policy, this is a surprising omission, not least since the few crosscountry studies exploring government intervention and liberalization tend to find evidence of partisan effects (Imbeau et al. 2001). The ideological structure of provincial parliaments and the variations across different factions and parties plays a significant role in Canadian politics. However, contrary to US politics the Canadian political system is most often characterized as closely fitting a brokerage model. Only considering the ideology of the party in government, as previous studies have done, does not provide a comprehensive picture of political ideology and economic policies in the Canadian provinces. Neither does a static characterization of party positions, as political parties tend to shift positions on a left-to-right scale over time. Political economy studies on partisan politics in the Canadian provinces may hence benefit from addressing parliamentary cohesion and party changes. 2 The political and legal institutions are the same across the ten provinces as are, with Québéc as the exception, the broader cultural environment. The issue of Francophone Québéc and, in particular, its strong independence movement takes up a special place in Canadian politics (e.g. Duchesne, Eagles and Erfle, 2003, Hinich et al. 1998). Federalism is one of the dimensions determining the preferences of Canadian politicians (Cross and Young 2002). 2

3 In this paper, we consequently examine how political ideology has influenced economic freedom in the Canadian provinces, taking these challenges into consideration. We analyze the dataset of economic freedom indicators compiled by the Fraser Institute in ten Canadian provinces over the period and introduce two different indices of political ideology: government and parliament ideology. We allow for parties to drift on the left-to-right scale, taking party changes into account. The results suggest that government ideology influenced labor market reforms: market-oriented governments promoted liberalization of the labor market but parliamentary ideology did not influence economic liberalization at all. At a more disaggregated level, we find that the ideological effects on the labor market are due to more leftwing governments increasing public employment. These findings (1) identify differences between leftist and rightwing governments concerning the role of government in the economy and (2) indicate that ideological polarization concerns governments but less so parliamentary fractions in the Canadian provinces. The rest of the paper is organized as follows: Section 2 discusses the related literature on partisan politics in Canadian provinces and formulates the hypotheses to be tested. Section 3 presents our indices on political ideology in the Canadian provinces. Section 4 presents the data on economic freedom and specifies the empirical model. Section 5 reports and discusses the estimation results, and investigates their robustness. Section 6 concludes. 2. Partisan politics in Canadian provinces Partisan politics has been a heavily debated topic in empirical political economy. Scholars have extensively examined to what extent and in which areas government ideology influenced economic policy (e.g. Hibbs 1977, 1992, Alesina et al. 1997, Imbeau et al. 2001, Bjørnskov 2008). While most studies are conducted at either a cross-country level or as comparisons between US states (e.g., the companion paper by Bjørnskov and Potrafke 2009), a small literature deals explicitly with Canada. 3

4 In the Canadian provinces, government ideology has played an important role in fiscal policy. Consistent with standard partisan theory, leftwing governments pursued a more expansionary fiscal policy than rightwing governments. In particular, leftwing governments increased public expenditures (see, for example, Tellier 2006, Pétry et al. 1999) and government ideology influenced the budget composition (Kneebone and McKenzie 2001, Blais and Nadeau 1992). 3 The empirical studies also suggest that electoral motives manipulated fiscal policies in the Canadian provinces (see also Reid 1998, Imbeau and Tellier 2004, Tellier 2004). At the federal level, electoral motives influenced public expenditures, while ideology-induced effects were less significant in the period (Pétry and Harmatz 1995). In the long run, however, neither partisan nor electoral cycles influenced public expenditures in Canada (Ferris et al. 2008). 4 At a more disaggregated level, spending on public health in the Canadian provinces appears to be driven by government ideology. The results by di Matteo (2009), for example, suggest that center-left parties are associated with smaller public shares for spending on health professionals and larger shares for total health spending. 5 Ideology-induced responses to globalization tended to depend on the level of unionization: when unionization was high and globalization was proceeding rapidly, leftwing governments have reduced public spending in the Canadian provinces (Pickup 2006). When unionization was low and globalization was proceeding rapidly, leftist governments have implemented compensatory spending. Conversely, party affiliation and government ideology do not appear to have an influence on (air) pollution in Canada. The empirical results by McKitrick (2006) suggest that government s political stripe in 13 Canadian cities has not reliably been associated with positive or negative effects on air 3 Political ideology also influenced fiscal policies in the US states. See, for example, Berry et al. (1999). 4 In general, electoral and partisan cycles did not influence macroeconomic indicators such as GDP and unemployment in Canada (Serletis and Afxentiou 1998). 5 On the effects of government subsidies to employer-provided health insurance in the Canadian provinces see, for example, Stabile (2001). 4

5 pollution. McKitrick (2006: 616) comes to the conclusion that on air quality all the leading political parties seem to favour slow, incremental improvements, with only small differences among parties in the desired rates of change. The implementation of abortion policy a strongly value-ladden question in North America has not only appeared to be a controversial issue between leftist and rightwing parties in Canada but also between the federal government and the provinces. For example, the Liberal Party which was in power on the federal level in 1995 opted for abortion services as medically necessary service. The governments in the provinces, however, such as the conservative one in Alberta as well as the leftwing government in the prairie province Saskatchewan were opposed to more abortion services (for an encompassing discussion of Canadian abortion policy from a political economic perspective see, for example, Palley 2006). In some areas, parties evidently acted differently on their home turf than at the federal level. There is thus plenty of evidence for partisan ideological effects in Canada, although the evidence is not uncontested and varies across different purposes. Only few studies, however, take into account that the political system and political traditions in Canada are somewhat different from those of other federal nations, and in particular the US. 6 Virtually all studies consider the political ideology of the incumbent government, yet much of the complementary literature on party systems and political institutions has characterized Canada as a country in which politics mostly resembles the brokerage model (e.g. Stevenson 1987, Young and Cross 2002). Canadian party positions are known to be ideologically somewhat opaque. Malloy (2003: 121), for example, assesses that party platforms are generally vague and it is typical following elections for governments to embark on major new policies not mentioned in the general election and with little consultation of parliamentary caucuses. In addition, the ideological cohesion of Canadian parties seems relatively weak, as parties more function as 6 Economic liberalization is, however, a long-run process. For this reason, it is not likely that politicians implement short-run policy changes to become re-elected. That is why we do not focus on electoral motives in economic liberalization. 5

6 brokers between many different factions and individual views represented in the party. As already weakly cohesive parties try to attract voters, their ideological position becomes less certain, with the possible exception of the new parties in a system that used to be a two-party plus game (Stephenson and Bélanger 2007). What is the result of a perfect brokerage model is therefore a political game, which tends to resemble purely Downsian voting where official parties ideology would be little more than cheap talk (Downs 1957). However, contrary to what most people seem to believe, this does not mean that one cannot have ideological influences on policy choices. First, Cross and Young (2002) show that Canadian politics have an ideological component when examining the attitudes of party members. Results in Kam (2001: 115) also suggest that exploring members voting behavior in Westminster style parliamentary systems such as Canada and the United Kingdom, party affiliation contains vastly more information about his or her behavior than do his or her preferences. Canadian party members, despite the weak ideological cohesion within their parties, are substantially more disciplined than in particular their American counterparts, at least when voting in the Canadian House of Commons (Malloy 2003). Instead, Canadian party members behave more like their North European colleagues than politicians in the US House of Representatives, in which party dissent is more common (Poole and Rosenthal 2001). As such, simply defining a standard position of a party is sufficient to capture most of the structure of actual voting in the Canadian parliaments. Second, a perfect brokerage model instead means that the ideology of the governing party/parties only exerts an influence on policy choices through their relative bargaining power. If, as is the case in brokerage politics, parties agree across parliamentary divides what to do, outcomes would presumably be determined by the bargaining power of each single political position. In addition, bargains across political divides would presumably be easier given the intra-party diversity of ideological positions that necessarily follows from weak ideological cohesion, which would tend to create inter-party ideological overlaps. In a Canadian and similar contexts, this means that ideological effects on political outcomes would follow not from the ideological position of the incumbent government but the average ideological 6

7 position of the provincial parliament. Hence, even though we expect, on the basis of the existing literature, to find evidence of partisan political effects across the Canadian provinces, we a priori cannot hypothesize about which field of politics the governing party or the entire party system is relevant. Following the implications of the related literature on partisan politics in the Canadian provinces and government ideology and economic liberalization the two competing hypotheses to be investigated in the following are: 1. Market-oriented governments have been more active in economic liberalization in Canadian provinces than leftwing governments. 2. Government ideology does not matter for policy-making in Canada, but the relative bargaining power of ideological positions does, given that Canadian politics are characterized by the brokerage model. To test these hypotheses, we first need a measure of government and parliamentary ideology consistent across time and provinces. 3. An ideology index for Canadian provinces Since Castles and Mair s (1984) pioneering efforts, many attempts have been made at coding ideology consistently across countries and periods (e.g. Huber and Inglehart 1995, Mair and Castles 1997, Woldendorp, Keman and Budge 1998, 2000, Bjørnskov 2008, Potrafke 2009a). The literature reflects three overarching problems when placing parties on a left-right scale: 1) the potential multidimensionality of political positions; 2) the comparability of scales across countries; and 3) the stability of scales across time. In the following, we base our argument on the premise that by exploring the potentially ideological political behavior across Canada, we address the specific problems associated with whether ideology-induced effects are driven by the position of the government or the ideological make-up of the provincial parliament. Common to all ideology indices is a fourth underlying problem, the question what ideology is. Ideology is difficult to define and conceptualize but most people recognize it when they see it. When 7

8 treating ideology with a modicum of scientific precision, a definition is necessary (cf. Gerring 1997). Most recent studies follow Gerring s (1997) approach, concluding his survey of the post-wwii literature by stressing the need for internal coherence in any core definition of ideology. He tries to differentiate ideology from, e.g., culture or belief-systems by arguing that a set of values becomes ideological only insofar as it specifies a concrete program, a set of issue-positions (Gerring 1997: 975). Any measure of political ideology thus has to rest on a coherent and consistent program, associating issue-positions with preferred outcomes or actions, although this program may only be implicit. Our political ideology coding does not suffer from comparability problems across objects as we only deal with the Canadian party system and do not make any cross-country comparisons. The analysis therefore takes place within the same institutional and political setup and broadly similar political tradition. Second, the coding explicitly addresses the first problem by focusing only on economic issues, which takes up an important part of Canadian politics (Kam, 2001; Cross and Young, 2002; Blais et al., 2004). Third, as we outline below, we consider a specific source of movements in the political scale to address the potential problem of temporal comparisons. We eventually employ the index illustrated in Figure 1, which allows the three major parties to take up more than a singular point on the left-to-right line. For example, we distinguish between the social democrat part of the New Democratic Party, which we ordinarily place at -2/3, and its more outright socialist faction placed at -1. Likewise, we distinguish between the social liberal faction of the Liberal party, placed at -1/3, and the business liberal faction, which we place at 0 and consider the standard position of the party. Finally, the Progressive Conservative Party is spread between Red Tories at 1/3, the standard position of the party at 2/3, and a clear right wing at 1, which would be comparable to the Republican Party in the US during the Reagan years. This implies that, accepting our coding, we are also able to distinguish between provincial governments lead by different factions of each party. For example, at the federal level we would place the liberal governments under Jean Paul Chrétien as social liberal but those of his successor Paul Martin 8

9 as business liberal. 7 At the provincial level, all three major parties have also moved their public profiles among positions. This happened perhaps most clearly when the Alberta conservatives with Ralph Klein taking over in 1993 moved to the right and away from a Red Tory position. Among Canadian liberals, examples include the New Brunswick party moving to the right when Frank McKenna took over the leadership in 1985 and the Newfoundland Liberal Party that moved to a similar business liberal position during the leadership of Clyde Wells. Even the New Democratic Party experienced ideological moves, although it otherwise seems more cohesive than the two larger counterparts (cf. Young and Cross 2002). Apart from the remnants of the ideological battle between the standard position and The Waffle, an outright socialist faction in the late 1970s, a move to the right was for example clear as the Saskatchewan branch moved to a more fiscally austere position when Roy Romanow took over the leadership from Allan Blakeney in Across the provinces, relatively stable differences also exist with the Ontario conservatives being strongly associated with a Red Tory position and the Québéc liberals placing themselves to the left of most other Canadian liberals and closer to the Parti Québécois. Finally, the Social Credit Party and the Reform Party / Canadian Alliance are placed at 1/3 on the scale while the two parties particular to Québéc, the Parti Québécoise and Action Démocratique, are placed at -2/3 and -1/3, respectively. A few other small parties that obtained representation are place as follows: the Saskatchewan Party (populist rightwing) at.5; the Confederation of Regions (in New Brunswick) at.5; the Equality Party in Québéc at 1/3; the Representative Party (in Alberta) at.5; and the Alberta Alliance at 1. We do not consider various minor communist and libertarian parties as they never attained representation within the period and code all independent members of parliament at the middle of the scale, 0. While this coding clearly has an inevitable subjective element, it retains what is a standard ranking of Canadian parties and replicates the increasing political span of Canadian politics (e.g. Huber and 7 We should stress that the ideology of the federal governments is not included in our econometric model, as it is picked up by year dummies. 9

10 Inglehart 1995, Mair and Castles 1997). The results in Cross and Young (2002) reconfirm this ranking by showing that members of the five parties place themselves accordingly on most issues, and in particular clearly so on economic issues. We therefore believe that although the validity of the implicit assumptions of unidimensionality and equidistance may be uncertain, as is the case of all ideology indices, our coding captures entirely standard features of the basic structure of the Canadian political system. Our possibility to assign provincial governments to factions within parties allows us to some extent to take party changes into account. This additional flexibility solves at least some of the problem in the literature associated with shifting party positions and accompanying scale instability of ideology indices (Knight 2006, Berry et al. 2007, Potrafke 2008). 8 To illustrate the structure of our index, Figure 2 shows the average ideology of the provincial parliaments across the period, measured as either the average ideology of votes, seats or governments. Interestingly, the figure not only depicts the substantial differences across the ten provinces, but also the discrepancy between the average ideology of voters and that of the resulting parliaments and governments. This discrepancy, which is partially a result of the Westminster style, first-past-the-post institution of Canadian democracy, means that the distribution of seats is somewhat more variable than that of votes, and the distribution of governments even more so; we therefore employ the distribution of seats instead of votes to calculate the average parliamentary position. Yet, the data also emphasize the substantial difference in what may be termed ideological traditions across the ten provinces with Québéc substantially more leftwing than other provinces and Alberta the most conservative province. While Figure 2 shows the long-run spatial ideological differences across Canada, Figure 3 illustrates the development of average ideology over time across the ten provinces since Again, 8 It may be important to emphasize that we coded all Canadian parties, and thereby all provincial governments and parliaments, before exploring the data on economic freedom. As far as possible, we have therefore tried to prevent any bias due to knowledge of our dependent variables entering our ideological coding choices. 10

11 Figure 3 illustrates how a first-past-the-post system tends to scale up the parliamentary consequences of ideological movements among voters and the massive loss of voters suffered by the Progressive Conservative Party after its heyday when it colloquially was referred to as The Big Blue Machine. While average ideology had trended slightly to the left since 1981, the almost violent demise of the Tories came about after a series of scandals of high-level corruption and nepotism (e.g. Kaplan 2000). From the late 1990s onwards, the conservative party has regained its status as several provincial parties distanced themselves from the federal Progressive Conservative Party and also moved away from Red Tory positions to more clearly ideological platforms. Averaging economic freedom (as defined below) across the ten provinces reveals an ideological pattern, as the correlation over time between overall economic freedom and the average ideology of the provincial parliaments is.39, using our ideology index based on the distribution of seats,.37 using the ideology index based on the distribution of votes, and.71 using the average ideology of governments. In the following, we explore this association in substantially more detail across both time and provinces during the period. 4. Data and estimation strategy 4.1 Data on economic freedom in the Canadian Provinces We use the dataset on economic freedom in the Canadian provinces first introduced by Karabegovic et al. (2003). This dataset, which is available for the period, contains yearly data for the ten Canadian provinces: Alberta, British Columbia, Manitoba, New Brunswick, Newfoundland, Nova Scotia, Ontario, Prince Edward Island, Quebec and Saskatchewan (balanced panel). The index of economic freedom includes three components: 1) the size of government, capturing general government consumption expenditure (percent of GDP), transfers and subsidies (percent of GDP), and social security payments (percent of GDP); 2) the tax structure, measured as an index equally weighing tax revenue (percent of GDP), the top marginal tax rate and its applied income 11

12 threshold, indirect tax revenue (percent of GDP) and sales taxes (percent of GDP); and 3) labor market freedom, measured as the extent of minimum wage legislation, government employment (percent of total provincial employment) and union density. The overall indicators take on values between 0 (minimum of economic freedom) and 10 (maximum of economic freedom). When we in the following split one of the three components into its constituent parts, we use the original, non-indexed data. Figure 4 illustrates that average economic freedom in the analyzed sectors was quite moderate in the 1980s, but remarkably decreased in the beginnings of the 1990s in Canadian provinces. Since 1992, economic liberalization has proceeded steadily. Over the period, changes have been made in most of the areas covered by the indicators, but most spectacularly in the size of government, and to a lesser extent, in labor market freedom. Figure 5 illustrates that overall economic freedom was most pronounced in Alberta (overall index on average 6.8) and less pronounced in Newfoundland (overall index on average 3.9) and Quebec (overall index on average 4.0). In Newfoundland, however, overall economic liberalization was proceeding rapidly till the middle of the 1990s. In Ontario, the overall index dropped from 5.9 in 1989 to 4.6 in 1993, an effect mostly driven by state intervention in the government sector (Figure 6). In a similar vein, Figure 6 also shows that the evident economic liberalization process in Newfoundland mainly reflected a dramatically decreasing size of government while the component of economic freedom concerning Takings and discriminatory taxation (Figure 7) changed comparatively less. In Prince Edward Island, for example, the index on Takings and discriminatory taxation even decreased from 6.2 in 1981 to 4.4 in Finally, Canadian labor markets became more flexible over time (Figure 8). Labor market deregulation was proceeding fast in provinces such as New Brunswick, Newfoundland and Nova Scotia. We will use these four indicators on Overall economic freedom, Size of government, Takings and discriminatory taxation and Labor market freedom as dependent variables in the econometric model. 12

13 4.2 Empirical model The basic panel data model has the following form: ln Economic Freedom Index ijt = α Ideology it + Σ k β k ln X ikt + β l Z ilt + η i + ε t + u ijt with i=1,,10; j = 1,, 4; k=1,,3; l=1,,2; t=1,...,24 (1) where the dependent variable ln Economic Freedom Index ijt denotes the growth rate of economic freedom index j. We distinguish between the four indicators of Overall economic freedom, Size of government, Takings and discriminatory taxation and Labor market freedom. Panel data unit root tests show that these series are stationary in growth rates. Ideology it describes the ideological orientation of the respective government or parliament as discussed in the previous section. Σ k ln X ikt contains two exogenous economic control variables. We follow the related studies on economic reforms to address potential economic crises and therefore include the five-year average of the unemployment rate, and international trade (outside Canada) as an approximation of globalization to address an overall trend in economic liberalization. Following Dickson (2009) we also control for the level of political competition by including the Herfindahl-Hirschman index of parliamentary competition because economic reforms are expected to proceed rapidly when political competition is strong (Dickson 2009 has identified political competition to foster public spending in Canada). The appendix provides descriptive statistics of all variables included. Lastly, η i represents a fixed province effect, ε t is a fixed period effect and u ijt describes an error term. We now turn to our choice of estimation procedure. We estimate the model with feasible generalized least squares and implement heteroscedastic and autocorrelation consistent (HAC) Newey- West type (Newey and West 1987) standard errors and variance-covariance estimates, since a Wooldridge test (Wooldridge 2002: ) for serial correlation in the idiosyncratic errors of a linear 13

14 static panel-data model implies the existence of unrestricted serial correlation when the indicators for Takings and discriminatory taxation and Labor market freedom are used. 5. Results 5.1 Basic Results Table 1 illustrates the regression results when the government ideology index is used and reports the coefficients and t-statistics (in absolute terms) for every single equation. The two economic control variables five-year average of the unemployment rate and the external trade variable mostly turn out to be statistically significant although the latter has a positive influence on labor market freedom (column 4) which is statistically significant at the 1% level. An F-test of the joint inclusion of the fixed period effects is strongly significant, which indicates time dependence of the dependent variables, i.e., a joint national trend. On the other hand, economic liberalization did not depend on political competition as the Herfindahl-Hirschman index does not turn out to be statistically significant across all equations. The main results in Table 1 show that government ideology did not influence overall economic liberalization (column 1), the sub-indicators on the size of government (column 2) or takings and discriminatory taxation (column 3). However, it has had a strong influence on labor market freedom: more market-oriented and rightwing provincial governments have been more active in promoting labor market freedom (column 4), an effect that is statistically significant at the 5% level. The effect is also economically significant as the numerical meaning is that a corresponding increase of the ideology variable by one point say from -1/3 (a left-leaning Liberal government) to 2/3 (a standard conservative government) would increase the growth rate of the labor market freedom indicator by about 1.5%. Table 2 provides the results when the parliament ideology index is used. This provides our test of the competing hypothesis that only the ideological make-up of the entire parliament ought to matter, 14

15 given that Canadian provincial policy-making is indeed characterized by the brokerage model. As expected, the influence of the control variables does not change compared to the results in Table 1. The parliament ideology index (rightwing), nevertheless, lacks statistical significance across all specifications. The results thus indicate that parliament ideology did not influence economic liberalization in the Canadian provinces at all. In other words, we do not find evidence of the type of ideological effects that would be consistent with a perfect brokerage type of politics. 5.2 Robustness of the results We checked the robustness of the results in several ways. First, we discuss an alternative econometric specification using a dynamic estimation procedure as it is important to note that the common fixed-effect estimator is biased when including a lagged dependent variable. The estimators taking into account the resulting bias can be broadly grouped into a class of instrumental estimators and a class of direct bias corrected estimators (see Behr 2003 for a discussion). In accordance with large sample properties of GMM methods, e.g., the estimator proposed by Arellano and Bond (1991) will be biased in an econometric model with N=10, such as in this paper. For this reason, bias corrected estimators are more appropriate. We apply Bruno s (2005a, 2005b) bias corrected least squares dummy variable estimator for dynamic panel data models with small N. 9 The regression results reported in Table 3 illustrate that the lagged dependent variable does not turn out to be statistically significant in all specifications. The inferences with respect to the government ideology variables on economic liberalization do not change at all. Market-oriented 9 We use the Blundell-Bond (1998) estimator as the initial estimator with which the instruments are collapsed as suggested by Roodman (2006). Applying this procedure ensures not to use invalid and too many instruments (see Roodman 2006 and 2009 for further details). Following Bloom et al. (2007), we undertake 50 repetitions of the procedure to bootstrap the estimated standard errors. Bootstrapping the standard errors is common practice applying this estimator, as Monte Carlo simulations demonstrate that the analytical variance estimator performs poorly for large coefficients of the lagged dependent variable (see Bruno 2005b for further details). The results do not qualitatively change with more repetitions such as 100, 200 or 500 as well as when the Arellano-Bond (1991) estimator is chosen as initial estimator. 15

16 governments have been more active in promoting labor market freedom. We will focus on this effect and stress the robustness of this finding. Table 4 shows results illustrating the robustness of the government ideology-induced effect on labor market freedom. To avoid any concerns about potential endogeneity problems of, for example, the trade variable, columns (1) and (2) show parsimonious specifications in which solely the government ideology variable (column 1) and the lagged dependent variable (column 2) are included. We have included further economic control variables such as trade between the Canadian provinces (trade inside Canada), GDP per capita in period t-1 and the size of government (total government final consumption expenditure as a share of GDP). It is important to note that purchasing-power parity data on GDP per capita at the provincial level are only available since 1981, so that we could not employ a five-year average without significantly reducing the sample. Yet GDP per capita in period t-1 does not turn out to be statistically significant and its inclusion does not change the inferences at all. In any event, the government ideology-induced effect on labor market freedom turns out to be robust and negative throughout. Political decisions to liberalize the economy may well need time to be implemented, and the political color of the government may change during this period. For this reason one may want to include lagged government ideology variables. When a single party has been in power for a fairly long time, however, lagged government ideology is less convincing because this particular party has been in office sufficiently long to implement its ideology-induced policy. In any event, to address potential time-lagged ideology effects on labor market freedom, we have replaced the government ideology variable in period t by its lagged values in the periods t-1, t-2 and t-3. The results show that influence (point estimate and statistical significance) of the lagged government ideology in period t-1 on labor market reform strongly increases. The influence of lagged government ideology in period t-2, however, is weaker than in period t, while the influence of lagged government ideology in period t-3 is similar to the influence of government ideology in period t. 16

17 A caveat applying to all panel data models concerns potential endogeneity of the explanatory variables. It is, however, not reasonable to expect that government ideology is influenced by changes in labor market freedom in the Canadian Provinces. Good instrumental variables for government ideology are simply not available, consistent with the prevalent view that voters decide based on predominantly non-economic information (e.g., Nannestad and Paldam 1997). Moreover, instrumenting ideology with the help of lagged government ideology would not be reasonable because ideology is highly persistent. The reported effects could depend on idiosyncratic circumstances in the individual provinces. We therefore have tested whether the results are sensitive to the inclusion/exclusion of particular provinces. We note that the influence of the ideology variable declines when Manitoba is excluded but increases when Saskatchewan is excluded. However, our main findings with respect to labor market freedom remain robust Decomposing labor market freedom While the results above show that political ideology influenced labor market deregulation in the Canadian provinces, they do not inform about which aspects of the labor market were affected. We therefore split the labor market freedom indicator into its three components: the extent of minimum wage legislation, the share government employment in total employment, and union density. The results in Table 5 show that the political ideology of incumbent governments did not influence minimum wage legislation. Conversely, the results show that the finding pertaining to overall labor market freedom is driven by the other two components. Decomposing the index thus reveals that leftwing provincial governments in general expanded government employment, and that union density also rose under leftwing governments. The coefficients of the government ideology variables on government employment and union density are statistically significant at the 5% level and imply that when the government ideology variable increased by 1 point (say from a center to a strong rightwing 17

18 government) the growth rate of the government employment share decreased by about 1% and the growth rate of the union density decreased by about 0.6%. Government employment is clearly susceptible to political changes, and in particular to changes in fiscal policy. The influence of government ideology on union density may seem puzzling. However, unionization may, for example, be higher in the government sector than in the private sector, which would indicate that the unionization result simply followed from ideological increases in government employment. We have therefore tested the statistical interaction between the changes in union density and the changes government employment but do not find any association between changes in union density and changes in government employment. Consequently, decreasing government employment and deunionization are statistically separate. The result that government ideology is significantly associated with changes in union density is thus most likely due to union density reflecting overall ideological changes in society, and not due to a causal effect. 6. Discussion and conclusions Whether and to what extent the political ideology of government affects policy-making and institutional choices has been a major question in political economy for decades. In this paper, we revisit the question by exploring ideological differences across the Canadian provinces over a 25-year period. Due to the special nature of the Canadian political system, we analyze the potential effects of both the ideological position of the government and the average position within each provincial parliament. We thus allow for the possibility that ideologically overlapping brokerage parties make political deals which do not depend directly on government ideology but on the relative bargaining power of ideological positions of all parties represented in parliament. We find that market-oriented governments promoted labor market deregulation in the Canadian provinces in the period. Government ideology, however, did not influence overall economic 18

19 liberalization; nor did parliamentary ideology influence labor market freedom or any other type of economic liberalization. These results are robust to including a set of potentially important control variables and are not driven by any single province but appear general to Canadian politics. Our findings have two potentially important implications. First, our overall results correspond with empirical evidence on government ideology and economic liberalization in OECD countries where empirical studies suggest that market-oriented governments have promoted overall economic liberalization (e.g. Pitlik 2007). For example, government ideology influenced product market deregulation and early-transition privatization in postcommunist countries (e.g., Potrafke, in press; Bjørnskov and Potrafke, 2008). In a similar vein, political polarization between leftwing and rightwing governments appears to have shifted from budgetary to non-budgetary affairs (Potrafke 2009b). The empirical evidence on how government ideology influenced labor market deregulation is nevertheless somewhat mixed. Botero et al. (2004), for example, examine labor market deregulation in 85 countries and find that leftwing governments have been associated with more stringent labor regulations than rightwing governments. Across OECD countries only, government ideology merely influenced reforms of unemployment benefit systems, labor taxes, employment protection legislation and retirement schemes (Duval 2008). The frequency of EU labor market directives, however, can neither be explained by the partisan composition of the Council, nor by the national level of labor market regulation (see the case study evidence by Vaubel 2008: 462). Second, our findings are pertinent since the Canadian political system has been characterized by political brokerage for several decades. Intra-party ideological cohesion in Canada is known to be relatively weak although party loyalty is strong and members of parliament are more likely to vote in line with their party than what is the case in, e.g., the US. The official electoral platforms of parties have also traditionally been vague and non-committal. Most decisions have therefore been negotiated in the corridor between the leaders of parties covering disparate ideological points of view, which a priori 19

20 would lead one to expect that government ideology should not matter for actual policy-making. Instead, political deals ought logically to be affected only by the relative bargaining strength of parties represented in parliament. Yet, we find no evidence that any of our measures have been influenced by the average ideological position of parliament but evidence to the contrary, that the ideological position of incumbent governments has been important in a particular policy area. Our results are consequently not in line with the standard view of how Canadian politics work and also appear to imply that the concerns that Canada may be moving towards a one-party-plus situation are less significant (cf. Scotto, Stephenson and Kornberg 2004). Given that political ideology actually works despite party bargaining in a brokerage model, our results confirm that parties continue to matter. This is consistent with the increasing clarity of ideological positions in more recent years in which new parties have come to play an active role in some Canadian provinces. In summary, our study indicates that political ideology can play an important role, even in an apparently non-ideological political system as the Canadian. Yet, they also show that the policy area subject to partisan effects across the provinces does not correspond exactly with what is typically found in cross-country studies. The overall results therefore question whether one should expect similar effects across democratic countries. We thus conclude by calling for more research in the way that different institutional and political systems affect the importance of political ideology for policy-making as well as in which areas ideology comes to be important. References Alesina, Alberto, Nouriel Roubini and Gerald D. Cohen (1997) Political Cycles and the Macroeconomy. (Cambridge, MA: MIT Press) Arellano, Manuel and Stephen Bond (1991) Some tests of specification for panel data: Monte Carlo evidence and an application to employment equations, Review of Economic Studies 58,

21 Behr, Andreas (2003) A comparison of dynamic panel data estimators: Monte Carlo evidence and an application to the investment function, discussion paper 05/03, Economic Research Centre of the Deutsche Bundesbank. Berry, William, Evan Ringquist, Richard Fording and Russell Hanson (2007) The measurement and stability of state citizen ideology, State Politics and Policy Quarterly 7, Bjørnskov, Christian (2008) The growth-inequality association: government ideology matters, Journal of Development Economics 87, Bjørnskov, Christian and Niklas Potrafke (2008) Politics and privatization in Central and Eastern Europe: a panel analysis, working paper, University of Konstanz. Bjørnskov, Christian and Niklas Potrafke (2009), Does government ideology influence economic freedom in the US states?, working paper, University of Konstanz. Blais, André and Richard Nadeau (1992) The electoral budget cycles, Public Choice 74, Blais, André, Mathieu Turgeon, Elisabeth Gidengil, Neil Nevitte and Richard Nadeau (2004) Which matters most? Comparing the impact of issues and the economy in American, British and Canadian elections, British Journal of Political Science 34, Bloom, David, David Canning, Richard K. Mansfield and Michael J. Moore (2007) Demographic change, social security systems, and savings, Journal of Monetary Economics 54, Blundell, Richard W. and Stephen Bond (1998) Initial conditions and moment restrictions in dynamic panel data models, Journal of Econometrics 87, Botero, Juan, Simeon Djankov, Rafael la Porta, Florencio C. Lopez-de-Silanes and Andrei Shleifer (2004) The regulation of labor, Quarterly Journal of Economics 119, Bruno, Giovanni S.F. (2005a) Approximating the bias of the LSDV estimator for dynamic unbalanced panel data models, Economics Letters 87, Bruno, Giovanni S.F. (2005b) Estimation and inference in dynamic unbalanced panel data models with a small number of individuals, Stata Journal 5, Castles, Francis G. and Peter Mair (1984) Left-right political scales: some expert judgement, European Journal of Political Research 12, Cross, William and Lisa Young (2002) Policy attitudes of party members in Canada: evidence of ideological politics, Canadian Journal of Political Science 35, Dickson, Vaughan (2009) Seat-vote curves, loyalty effects and the provincial distribution of Canadian government spending, Public Choice 139, Di Matteo, Livio (2009) Policy choice or economic fundamentals: what drives the public-private health expenditure balance in Canada? Health Economics, Policy and Law 4,

22 Downs, Anthony (1957) An economic theory of democracy. (New York: Harper and Row) Duchesne, Erick, Munroe Eagles and Stephen Erfle (2003) Constituency homogeneity, economic risk and support for Quebec sovereignty: a research note, Canadian Journal of Political Science 36, Duval, Romain (2008) Is there a role for macroeconomic policy in fostering structural reforms? Panel evidence from OECD countries over the past two decades, European Journal of Political Economy 24, Ferris, J. Stephen, Soo-Bin Park and Stanley L. Winer (2008) Studying the role of political competition in the evolution of government size over long horizons, Public Choice 137, Gerring, John (1997) Ideology: A definitional analysis, Political Research Quarterly 50, Hibbs, Douglas A.J. (1977) Political parties and macroeconomic policy, American Political Science Review 71, Hibbs, Douglas A.J. (1992) Partisan theory after fifteen years, European Journal of Political Economy 8, Hinich, Melvin J., Michael C. Munger and Scott de Marchi (1998) Ideology and the construction of nationality: the Canadian elections of 1993, Public Choice 97, Huber, Peter and Ronald Inglehart (1995) Expert interpretations of party space and party locations in 42 societies, Party Politics 1, Imbeau, Louis M., Francois Pétry and Moktar Lamari (2001) Left-right party ideology and government policies: A meta-analysis, European Journal of Political Research 40, Imbeau. Louis M. and Genevieve Tellier (2004) The political economy of budget deficits in the Canadian provinces, , in: Politics, institutions and fiscal policy: deficits and surpluses in federated states ed. Louis M. Imbeau and Francois Pétry (Toronto: Lexington Books) Kam, Christopher (2001) Do ideological preferences explain parliamentary behaviour? Evidence from Great Britain and Canada, Journal of Legislative Studies 7, Kaplan, William (2000) Presumed Guilty: Brian Mulroney, the Airbus Affair, and the Government of Canada (Toronto: McClelland & Stewart) Karabegovic, Amela, Samida, Dexter, Schlegl, Chris M and FredMcMahon (2003) North American economic freedom, European Journal of Political Economy 19, Kneebone, Ronald D. and Kenneth J. McKenzie (2001) Electoral and partisan cycles in fiscal policy: an examination of Canadian provinces, International Tax and Public Finance 8, Knight, Kathleen (2006) Transformations of the concept of ideology in the twentieth century, American Political Science Review 100,

23 Mair, Peter and Francis G. Castles (1997) Revisiting expert judgements, European Journal of Political Research 31, Malloy, Jonathan (2003) High discipline, low cohesion? The uncertain patterns of Canadian parliamentary party groups, Journal of Legislative Studies 9, McKitrick, Ross (2006) The politics of pollution: party regimes and air quality in Canada, Canadian Journal of Economics 39, Nannestad, Peter and Martin Paldam (1997) From the pocketbook of the welfare man: a pooled crosssection study of economic voting in Denmark, , British Journal of Political Science 27, Palley, Howard A. (2006) Canadian abortion policy: national policy and the impact of federalism and political implementation access to services, Journal of Federalism 36, Pétry, Francois and Howard R. Harmatz (1995) Politico-economic interactions in Canada: an empirical assessment, Public Finance Review 23, Pétry, Francois, Louis M. Imbeau, Jean Crête and Michel Clavet. (1999) Electoral and partisan cycles in the Canadian provinces, Canadian Journal of Political Science 32, Pickup, Mark (2006) Globalisation, politics and provincial government spending in Canada, Canadian Journal of Political Science 39, Pitlik, Hans (2007) A race to liberalization? Diffusion of economic policy reform among OECDeconomies, Public Choice 132, Poole, Keith T. and Howard Rosenthal (2001) D-NOMINATE after 10 years: a Comparative update to Congress: a political-economic history of roll-call voting, Legislative Studies Quarterly 26, Potrafke, Niklas (2008) Parties change! Introducing a dynamic index on voter polarization, working paper, University of Konstanz. Potrafke, Niklas (2009a) Did globalization restrict partisan politics? An empirical evaluation of social expenditures in a panel of OECD countries, Public Choice 140, Potrafke, Niklas (2009b) Does government ideology influence political alignment with the U.S.? An empirical analysis of voting in the UN General Assembly, Review of International Organizations 4, Potrafke, Niklas (in press) Does government ideology influence deregulation of product markets? Empirical evidence from OECD countries, Public Choice, forthcoming. Reid, Bradford G. (1998) Endogenous elections, electoral budget cycles and Canadian provincial Governments, Public Choice 97, Roodman, David (2006) How to do xtabond2: an introduction to Difference and System GMM in Stata, working paper 103, Center for Global Development. 23