Cabinet Durability and Fiscal Discipline Data and Model Construction government can reasonably expect to remain in office For our main analyses,we gather data on several through the CIEP without holding elections.21 By estimating a distribution of expected durations decades of budgeting in 15 European democracies that for each of our sample cabinets,we have a straightfor- allow for parliamentary dissolutions-the same data ward way of accounting for the uncertainty of these analyzed by Bawn and Rosenbluth(2006)and Martin predictions.This is important for both empirical and and Vanberg(2013).17 Our dependent variables are the theoretical reasons.Empirically.this variable is,after OECD calendar-year estimates of central government all,an estimate with an associated error structure and spending as a percentage of GDP and this spending fig- ignoring this error may bias our estimates of the rela- ure minus central government receipts as a percentage tionships of interest,and therefore bias our substantive of GDP,where greater values indicate greater deficits. conclusions.Theoretically,expectations rarely take the Budgets in our sample are typically submitted in the form of a point when they are generated by individu- latter half of the year for spending in the following cal- als and are explicitly distributions when they are gen- endar year.Most submission dates fall between August erated by a collective,whether it is a system of firms and October of the calendar year preceding the budge a betting market,or the ministers composing a cabi- year,though dates as early as July and as late as april net.If we assume that rational expectations are dis- of the budget year appear in the data.8 The data also tributions that are,in the aggregate,centered on the include information on several political economic char most probable (or,most expected,given the state of the acteristics salient to budget-making,which we discuss world)outcome as Muth (1961)theorized and others below. have found empirically,then modeling these distribu- The explanatory variable of interest is the cabinet's tions,rather than merely their central tendency,is crit- predicted duration-the number of days it expects to ical to hypothesis testing. remain in office-at the time the budget was submit- We model these distributions by estimating our ted.We derive the measure by first reestimating Chiba spending and deficit models 1,000 times-once for 4号元 Martin,and Stevenson's (2015)model of government each prediction of cabinet survival.Thus,for each it- duration,which jointly estimates cabinet formation and survival.9 To account for uncertainty in this esti- eration,we impute an expected remaining duration for mate,we employ a nonparametric bootstrap.20 At each each cabinet-year in our data using a single set of boot- strapped survival predictions,estimate the spending of the 1.000 bootstrap iterations,we randomly resam- and deficit models,and record the results.This yields ple the data,with replacement,from our set of 432 cab 1,000 regression results for the main models which we inets and reestimate the model.We then use the model summarize and interpret below,but we first discuss the estimates to predict the duration for each cabinet in our construction and estimation of the spending models. data,record the predictions,and reiterate,generating a Alongside our focal variable (predicted duration), distribution of 1.000 predicted survival times for each we include a set of political economic control vari- cabinet. ables borrowed from Bawn and Rosenbluth(2006)and For each cabinet-budget year in our spending data Martin and Vanberg (2013)to account for potential we alter these distributions in two ways:(1)we subtract confounders to our relationship of interest while keep- the number of days the cabinet has served at the time ing our substantive results comparable to previous re- of budget submission and(2)we trim any expected du- search.The measurement of these covariates and the rations that exceed the CIEP back to the expiration of reasons they are included in the models are described the CIEP.Predicted durations exceeding the CIEP are in great detail by Bawn and Rosenbluth and Martin fairly rare,but must nonetheless be accounted for-no and Vanberg,so we do not reiterate that information here.We provide,instead,a more general discussion of 17 Sample countries include Austria(1971-2006).Belgium (1971- the rationale motivating inclusion of these covariates 2007.Denmark(1972-2009).Finland(1971-2007),France(1979- 2009),Germany(1971-2009),Greece(1979-2004),Ireland(1971- which break down into three groups:variables captur- 2009).Italy (1971-2008).Luxembourg (1991-2004).the Nether- ing the government's taste for public spending;vari- lands(1971-2006),Portugal(1978-2009),Spain(1980-2009),Sweden ables accounting for the state's revenue supply and en- (1971-2009),and the United Kingdom (1971-2009).Notably,Nor- titlement burden;and variables indicating institutional way.which is included in the Bawn and Rosenbluth(2006)sample.is constraints on spending depth and responsibility omitted here because its elections are fixed. 18 Budget dates were coded from OECD First,consider spending tastes,or the breadth of Journal on Bud- geting country issues (http://www.oecd.org/governance/budgeting spending demands within the cabinet.There is broad 四 oecdjournalonbudgeting.htm).Each reports the deadline by which theoretical consensus in the literature is that left- the government must present the annual budget to parliament,typi- leaning parties prefer to spend more than right-leaning cally between August and October of the preceding year.For coun parties and we account for this by including Pow- tries not covered by the Journal on Budgeting,we refer to the respec- tive constitution or applicable legal framework.In cases in which the ell's (2000)measure of the cabinet's ideological po- budget deadline was unclear or fell within two months of a change sitioning:the mean,seat-weighted,left-right stance of of government,the exact date on which the budget was presented to parliament was located in the respective parliamentary archives, ensuring that all budgets are attributed to the correct cabinets. The constitutional interelection period is included in our cabinet 19 This is a conditional logit model of the government selection stage. duration model and is a powerful duration predictor.We note,how- joined to a Weibull survival by means of a Gaussian copula function ever,that model estimates using durations that are not trimmed to See Chiba,Martin,and Stevenson(2015)for details. CIEP also support our central predictions. 20 Bootstrapped model estimates are available in the Appendix. 22 But see Clark(2009)for opposing evidence. 945Cabinet Durability and Fiscal Discipline Data and Model Construction For our main analyses, we gather data on several decades of budgeting in 15 European democracies that allow for parliamentary dissolutions—the same data analyzed by Bawn and Rosenbluth (2006) and Martin and Vanberg (2013).17 Our dependent variables are the OECD calendar-year estimates of central government spending as a percentage of GDP and this spending fig￾ure minus central government receipts as a percentage of GDP, where greater values indicate greater deficits. Budgets in our sample are typically submitted in the latter half of the year for spending in the following cal￾endar year.Most submission dates fall between August and October of the calendar year preceding the budget year, though dates as early as July and as late as April of the budget year appear in the data.18 The data also include information on several political economic char￾acteristics salient to budget-making, which we discuss below. The explanatory variable of interest is the cabinet’s predicted duration—the number of days it expects to remain in office—at the time the budget was submit￾ted.We derive the measure by first reestimating Chiba, Martin, and Stevenson’s (2015) model of government duration, which jointly estimates cabinet formation and survival.19 To account for uncertainty in this esti￾mate, we employ a nonparametric bootstrap.20 At each of the 1,000 bootstrap iterations, we randomly resam￾ple the data, with replacement, from our set of 432 cab￾inets and reestimate the model.We then use the model estimates to predict the duration for each cabinet in our data, record the predictions, and reiterate, generating a distribution of 1,000 predicted survival times for each cabinet. For each cabinet-budget year in our spending data, we alter these distributions in two ways: (1) we subtract the number of days the cabinet has served at the time of budget submission and (2) we trim any expected du￾rations that exceed the CIEP back to the expiration of the CIEP. Predicted durations exceeding the CIEP are fairly rare, but must nonetheless be accounted for—no 17 Sample countries include Austria (1971–2006), Belgium (1971– 2007), Denmark (1972–2009), Finland (1971–2007), France (1979– 2009), Germany (1971–2009), Greece (1979–2004), Ireland (1971– 2009), Italy (1971–2008), Luxembourg (1991–2004), the Nether￾lands (1971–2006),Portugal (1978–2009), Spain (1980–2009), Sweden (1971–2009), and the United Kingdom (1971–2009). Notably, Nor￾way, which is included in the Bawn and Rosenbluth (2006) sample, is omitted here because its elections are fixed. 18 Budget dates were coded from OECD Journal on Bud￾geting country issues (http://www.oecd.org/governance/budgeting/ oecdjournalonbudgeting.htm). Each reports the deadline by which the government must present the annual budget to parliament, typi￾cally between August and October of the preceding year. For coun￾tries not covered by the Journal on Budgeting, we refer to the respec￾tive constitution or applicable legal framework. In cases in which the budget deadline was unclear or fell within two months of a change of government, the exact date on which the budget was presented to parliament was located in the respective parliamentary archives, ensuring that all budgets are attributed to the correct cabinets. 19 This is a conditional logit model of the government selection stage, joined to a Weibull survival by means of a Gaussian copula function. See Chiba, Martin, and Stevenson (2015) for details. 20 Bootstrapped model estimates are available in the Appendix. government can reasonably expect to remain in office through the CIEP without holding elections.21 By estimating a distribution of expected durations for each of our sample cabinets, we have a straightfor￾ward way of accounting for the uncertainty of these predictions. This is important for both empirical and theoretical reasons. Empirically, this variable is, after all, an estimate with an associated error structure and ignoring this error may bias our estimates of the rela￾tionships of interest, and therefore bias our substantive conclusions. Theoretically, expectations rarely take the form of a point when they are generated by individu￾als and are explicitly distributions when they are gen￾erated by a collective, whether it is a system of firms, a betting market, or the ministers composing a cabi￾net. If we assume that rational expectations are dis￾tributions that are, in the aggregate, centered on the most probable (or,most expected, given the state of the world) outcome as Muth (1961) theorized and others have found empirically, then modeling these distribu￾tions, rather than merely their central tendency, is crit￾ical to hypothesis testing. We model these distributions by estimating our spending and deficit models 1,000 times—once for each prediction of cabinet survival. Thus, for each it￾eration, we impute an expected remaining duration for each cabinet-year in our data using a single set of boot￾strapped survival predictions, estimate the spending and deficit models, and record the results. This yields 1,000 regression results for the main models which we summarize and interpret below, but we first discuss the construction and estimation of the spending models. Alongside our focal variable (predicted duration), we include a set of political economic control vari￾ables borrowed from Bawn and Rosenbluth (2006) and Martin and Vanberg (2013) to account for potential confounders to our relationship of interest while keep￾ing our substantive results comparable to previous re￾search. The measurement of these covariates and the reasons they are included in the models are described in great detail by Bawn and Rosenbluth and Martin and Vanberg, so we do not reiterate that information here. We provide, instead, a more general discussion of the rationale motivating inclusion of these covariates which break down into three groups: variables captur￾ing the government’s taste for public spending; vari￾ables accounting for the state’s revenue supply and en￾titlement burden; and variables indicating institutional constraints on spending depth and responsibility. First, consider spending tastes, or the breadth of spending demands within the cabinet. There is broad theoretical consensus in the literature is that left￾leaning parties prefer to spend more than right-leaning parties22 and we account for this by including Pow￾ell’s (2000) measure of the cabinet’s ideological po￾sitioning: the mean, seat-weighted, left-right stance of 21 The constitutional interelection period is included in our cabinet duration model and is a powerful duration predictor. We note, how￾ever, that model estimates using durations that are not trimmed to CIEP also support our central predictions. 22 But see Clark (2009) for opposing evidence. 945 Downloaded from https://www.cambridge.org/core. Shanghai JiaoTong University, on 26 Oct 2018 at 03:53:05, subject to the Cambridge Core terms of use, available at https://www.cambridge.org/core/terms. https://doi.org/10.1017/S0003055418000436