COMPROMISE, DIFFERENTIATION, AND COLLECTIVE RESPONSIBILITY

Compromise and cooperation can foster several intuitive preferred outcomes for cabinet parties. Cabinet parties that “play nice” with their partners are likely to find that behavior reciprocated, leading to more efficient policymaking and division of office spoils, while simultaneously signaling to the chamber that they are desirable partners in governance. These benefits, however, are not without cost. In short, multiparty governance obscures the strategically selected ideological positions taken by member parties and makes credit claiming for various accomplishments more difficult (e.g., Fortunato and Stevenson Reference Fortunato and Stevenson2013; Martin and Vanberg Reference Martin and Vanberg2008). Indeed, Fortunato (Reference Fortunato2017) presents evidence that voters who perceive higher levels of compromise by a coalition party—manifest in perceptions of increasing ideological similarity to its partner(s)—discount both the policy statements of that party and its contribution to the government’s performance. This tendency provides powerful incentives for cabinet parties to differentiate from their partners to protect the integrity of their strategically selected policy positions and demonstrate competence to their supporters.

However, there are substantial obstacles to differentiation in multiparty governance. Member parties may only differentiate on issues that are salient and present, which effectively constrains public dissent to the contents of the legislative agenda. Furthermore, coalition members are bound by collective cabinet responsibility, a set of formal and informal rules that set the parameters of behavior for members of government (Laver and Shepsle Reference Laver and Shepsle1994). Collective responsibility inhibits coalition parties from, for example, voting against proposals offered by their partners or speaking out against the proposal once it has been passed into law. The penalty for violating these rules may be the loss of portfolio, dismissal from cabinet, or even dissolution of the government.

Taking agenda limitations and collective responsibility together, coalition parties have precious few opportunities to differentiate from their partners without risking their position in cabinet. Parties must have a proposal over which to squabble and they must be able to express themselves before the proposal’s final vote. Thus, cabinet parties are constrained to the legislative review process, the window of time between a bill’s initiation and passage (or death), during which the proposal may be scrutinized. In practice, this leaves three fora in which cabinet parties may demonstrate how they differ from their coalition partners: legislative amendments, parliamentary debate, and direct communication strategies (e.g., the issuing of party press releases).

Here, I examine legislative amendments, which are, by their very nature, expressions of dissent with a policy proposal. They may only be offered to an initiated proposal, and are therefore germane to the agenda, and they are costly, requiring expertise, time, and labor to draft and propose, and therefore cannot be dismissed as “cheap talk.” Thus, we may consider amendments credible signals of differentiation between coalition partners sent by the reviewing party in reference to the proposal’s authoring party.

Importantly, there is qualitative and quantitative evidence that proposing amendments sends differentiating signals that are likely to be absorbed by the electorate by conditioning the information environment surrounding the cabinet. For instance, most parliaments publish daily reports cataloging events in committee and plenary meetings, including ministerial questions, debate, and the proposal of amendments. These records are vital resources for understanding politics in general and the policymaking process in particular in both scholarly research (e.g., Martin and Vanberg Reference Martin and Vanberg2011) and popular press (e.g., Lund Reference Lund2013). Increasing conflict in the review process is quite likely to make for a more antagonistic tone in media reports of party interactions and new research cataloging the effects of this reporting on party interactions (such as the review process) suggests that these behaviors do, in fact, substantially shape media narrative on politics and that voters receive and assimilate these messages into their perceptions (Adams, Weschle, and Wlezien Reference Adams, Weschle and Wlezien2016). Taken together with Fortunato (Reference Fortunato2017), the research suggests that differentiation strategies may help mitigate electoral losses.

WHEN WILL PARTIES AMEND?

Assume that the cabinet parties have arrayed themselves along the left–right spectrum such that they have each maximized their expected electoral returns. Now assume that voters perceive the parties’ strategic self-placement, but will update these perceptions in response to signals from the policymaking process. Parties that appear to be cooperative or compromising will be updated as more similar, converging upon each other and shrinking the distance between them. Parties that squabble and antagonize one another will be updated as more distinct, growing the perceived distance between them. These assumptions are supported by the extant literature (e.g., Adams, Ezrow, and Wlezien Reference Adams, Ezrow and Wlezien2016; Fortunato Reference Fortunato2017; Fortunato and Stevenson Reference Fortunato and Stevenson2013). We can think of the utility a cabinet party derives from amending the legislative proposals of their partners in government as a function of three parameters: the cost of amending, the policy payoff, and the electoral payoff, which is a function of the benefit of differentiation from the bill’s authoring party. ^{Footnote 1} As the perceived distance between the proposing party and the reviewing party closes (relative to their strategically selected positions), the benefit of differentiation via amendment increases.

• • The more similar a cabinet party pair is perceived, the more they will amend one another’s legislative proposals.

This hypothesis is the focus of the manuscript; however, considering it in a vacuum discounts both the policy benefits of amending and the costs, which previous research shows can be significant. As such, a discussion of both factors is warranted and, while this may not yield an exhaustive accounting of the predictors of review, it will yield a listing of measurable covariates that will help recover clean estimates of the focal relationship.

The canonical work on legislative review and coalition policymaking argues that ministers often propose bills at their own ideal point, rather than at the coalition compromise, in order to reap position-taking benefits—another form of differentiation (Martin and Vanberg Reference Martin and Vanberg2011). This “ministerial drift” must be mitigated by the minister’s coalition partners during legislative review to prevent agency loss. As such, one of the most important covariates in the model below is the ideological distance of the proposing party from the coalition compromise (“compromise distance”), which is the mean, seat-weighted ideological position of all cabinet parties. I also include the dyadic ideological distance between the proposing party and the reviewing party (“CMP distance”). In addition to providing a control required to recover the relationship of interest, this allows us to begin to understand whether the observed patterns of amending are more consistent with self-interest or coalition policing behaviors. As Martin and Vanberg’s (Reference Martin and Vanberg2011) amendment data are collected at the bill-level (they sum the total number of article changes submitted by all parties), amendments motivated by private policy concerns are empirically inseparable from amendments motivated by coalition policing. The data I have gathered count article changes submitted to a bill by each individual party and are therefore better able parse these behaviors.

Whether or not the cabinet controls a majority of seats in parliament is also critical. Minority cabinets will not only be compelled to monitor ideologically dissimilar coalition partners, but they must also appease some pivotal opposition party. In expectation, this need to maintain opposition support should reduce the number of amendments we observe. This is because every position of the opposition party relative to the ministerial and reviewing parties must accomplish one of the following: increase the minister’s incentive to make an offer about the coalition compromise; increase the minister’s incentive to make an offer about the reviewing party’s ideal point; or decrease the reviewing party’s incentive to amend. A formal explanation is given in the appendix.

As for the cost of amending, extant literature suggests that holding a committee chair empowers reviewing parties, substantially reducing the costs of legislative review, and Fortunato, Martin, and Vanberg (Reference Fortunato, Martin and Vanberg2017) provide empirical evidence to this effect. The empirical model therefore includes indicators for the identity of the committee chairperson: the party of proposing minister, reviewing party, another cabinet member, or an opposition party. As my data are coded at the bill (rather than party) level, they allow more leverage in discovering whether committee chairs’ primary influence is positive (increasing their party’s ability to scrutinize), negative (inhibiting other parties from scrutinizing), or both. Relatedly, Thies (Reference Thies2001) and others have argued that junior ministers are the executive complement to committee chairs and the model controls for their presence as well.

Finally, amendments require expertise and labor to draft and submit and having more members in parliament to share the burden of scrutiny should enable the reviewing party to amend more freely. I therefore include the reviewing party’s seat share to account for differences in the ability to amend as a function of information and labor resources across parties. ^{Footnote 2}

DATA AND MODEL SPECIFICATION

The dependent variable is constructed from original data on amendments offered to government proposals in committee. I gather information on over 2,200 draft bills introduced by cabinet ministers in three parliamentary democracies with long histories of coalition governance: Belgium (1992–2010), Denmark (1991–2004), and the Netherlands (1995–2013). ^{Footnote 3} This is the largest collection of such data assembled.

For each bill, I record the proposing minister’s party and department, ^{Footnote 4} the number of articles in the original proposal, and the number of article changes offered to the bill by all cabinet parties save the authoring party. ^{Footnote 5} These article changes, which serve as the dependent variable in the coming analysis, are counted in the same manner as Martin and Vanberg (Reference Martin and Vanberg2011) with two exceptions. First, I count all submitted article changes, while Martin and Vanberg count all accepted changes; this is because I am interested in the use of review for signaling purposes while Martin and Vanberg are interested in its policy implications. Second, I attribute all amendments to a single party, whereas Martin and Vanberg sum all amendments proposed by all parties and use the bill as the unit of observation. Recording the data in this manner is critical. Only by coding changes at the bill-party level am I able to test the above hypothesis. Importantly, summing the amendments in my data to the bill level and replicating the Martin and Vanberg model produces very similar results, which, in addition to providing out-of-sample support for their argument, implies that the mechanics of review across the two samples are quite similar.

The independent variable of interest, the distance perceived by voters between the proposing party and the reviewing party, is calculated with survey data from Eurobarometer, European Electoral Survey, and the Comparative Study of Electoral Systems surveys conducted between 1989 and 2014. In each module, respondents were asked to place their country’s political parties on an ideological scale. These placements are rescaled to a common 0–10 range and a placement is estimated for each party in each survey via linear regression. ^{Footnote 6} These individual party placements are then used to create a voter perceived distance between each ministerial party and each reviewing party (“voter distance”). The expectation is that the parameter estimate on voter distance will be negative; the more similar voters believe the cabinet parties are, relative to their strategically selected positions, the more they should amend one another’s proposals in order to differentiate. The model employs distance estimates derived from surveys administered about the time the cabinet was formed. This is because the theoretical process—that today’s voter perceptions impact amendment activity to shape tomorrow’s voter perceptions, which in turn impact tomorrow’s amending—implies endogeneity among the outcome and covariate of interest. The distance measured about the cabinet’s formation should be the most insulated from these effects. ^{Footnote 7}

Also included are the controls discussed above—indicators for minority cabinets, junior ministers, and different committee chair types, as well as the manifesto distance between the ministerial and reviewing parties and the minister’s distance from the coalition compromise. These are calculated with left–right preference estimates derived from the Comparative Manifestos Project data following Lowe et al. (Reference Lowe, Benoit, Mikhaylov and Laver2011). CMP was chosen over alternative measures, such as the Chapel Hill expert survey measures for two reasons: first, the manifesto estimates are explicitly estimates of parties’ campaign platform, their strategically chosen positions, and therefore a better fit to theory; second, CMP has substantially better coverage for the sample period. Descriptive statistics are shown in Table 1.

TABLE 1. Descriptive Statistics of Key Variables

The correlation of the distance between coalition party pairs as perceived by voters and as selected by the parties is shown in Figure 1, where the overall correlation is plotted with both linear and LOWESS lines and the estimates of a linear model regressing CMP distance on voter distance (which are measured on different scales) are given in the lower right-hand corner. The figure shows that the relationship is positive, but weakly so. Given that the sample only includes cabinet dyads—parties sufficiently compatible to coalesce in the first place—this weakly positive relationship makes sense (including opposition parties would make the relationship appear much stronger). The figure also reveals considerable variation between the similarity perceived by voters and the similarity manifested in the selected policy positions of cabinet partners. This variation is the focus of the manuscript.

FIGURE 1. Comparison of Voter–Perceived and Manifesto Dissimilarity.

The data used to construct the variables were all collected from the respective countries’ legislative databases, the Constitutional Change and Parliamentary Democracies (CCPD) project data (Strøm, Müller, and Bergman Reference Strøm, Müller and Bergman2008), or the Parliament and Government Composition Database (Döring and Manow Reference Döring and Manow2011). In addition to the variables discussed, the model includes the cabinet’s size in number of parties (logged) and three exposure terms: the (logged) number of articles in the original proposal, the (logged) number of days the bill spent under review, and a dummy variable indicating that the plenary session expired before the conclusion of the bill’s scrutiny.

ESTIMATION AND ANALYSIS

The dependent variable is the number of article changes offered by a cabinet party to the proposal of one of its coalition partners. A Poisson model is appropriate, but the data structure presents a small hurdle, grouping bills across different countries where amendment propensities vary systematically; Belgians amending more than Danes and Dutch amending more than Belgians on average. ^{Footnote 8} The model therefore includes country fixed effects, but it is also possible that certain bills are offered more amendments for reasons not captured by the measured covariates, and, because multiple parties may amend the same bill, there is potential for correlations across rows of data within bills. The model therefore allows random intercepts at the bill level. ^{Footnote 9} The results are presented in Table 2 alongside results from a stripped-down model containing only the covariate of interest, exposure terms, and fixed and random effects. The comparison shows that, while voter distance is robust (p < 0.05, single-tailed) in the simple model, its magnitude is suppressed. This reinforces how important it is to consider the costs as well as the policy benefits of amending when examining review behavior.

TABLE 2. Main Model Results

The full model bears strong support for the central hypothesis. Voter distance has a robust negative effect on amending—the more voters perceive the reviewing party as similar to the ministerial party, the more the reviewing party will amend. Holding all other covariates constant at their mean, the effect of a one standard deviation decrease in voter distance between a pair of cabinet partners, is a 9% (4%, 13% CI) increase in the predicted number of amendments. To put this in context, this is roughly 1/3 of the magnitude of the effect of a similar change in compromise distance (a 29% decrease in amendments). This is strong evidence for the hypothesis and supports the central argument that cabinet parties, limited by collective responsibility in their ability to highlight policy differences with their partners in government, use the legislative review process to communicate their ideological distinctiveness by voicing dissent with their partners’ policy proposals. These findings are supported by country-by-country estimates that can be viewed in the appendix. ^{Footnote 10}

The remaining variables in the model conform to expectations, which suggests proper specification. The number of articles in the original draft bill and the length of the review period have estimates near 1 (typical for exposure variables) and the more substantively interesting variables also have estimates in the “right direction.” The more MPs a party has to share the burden of amending, the more amendments it submits. If the reviewing party’s seat share increases from, say, 0.1 to 0.2, its predicted number of amendments increases by about $24{\rm{\% }}$ . The cabinets’ minority status also yields the predicted relationship, reducing the number of proposed article changes by 40%, though this fails to reach traditional levels of significance.

Also interesting among the control variables are the estimates on dyadic CMP distance and compromise distance. The estimate on compromise distance is positive, substantively large, and statistically robust. The farther away the proposing minister from the coalition compromise, the more the reviewing party tends to amend. It would appear that CMP distance exerts a negative effect, which no theory would predict. I note that the estimate on this covariate is among the more unstable across specifications and discuss this more in the appendix. Nonetheless, taken together, the patterns of observed amendment behavior are more consistent with a coalition-policing model of review behavior than with a self-interested model of review behavior, implying that conflict amelioration is paramount, as has been argued elsewhere (e.g., Bowler et al. Reference Bowler, Bräuninger, Debus and Indridason2016).

The final substantively interesting set of results regard the presence of a junior minister and identity of the committee chairman. Having a junior minister in the department of the proposing minister has no discernible effect on amendment behavior in this sample, though it should be noted that the reviewing party only has a junior minister in the proposing minister’s department in about 7% of the observations here and these are heavily clustered in the Netherlands. The model also suggests that the identity of the committee chair is largely irrelevant to legislative scrutiny amongst cabinet partners. This comports with previous research finding no difference in scrutiny between committees chaired by the party of the proposing minister and committees chaired by her partner in government, but substantial increase in scrutiny when the committee is chaired by a member of the opposition (Fortunato, Martin, and Vanberg Reference Fortunato, Martin and Vanberg2017). These results support the interpretation that the uptick in scrutiny under opposition chaired committees is evidence of opposition influence and that the chair’s power may be exercised positively when held by opposition but is almost certainly exercised negatively when held by cabinet, protecting bills from opposition influence. This implies that committee chairpersons, in practice, act as more of a “backstop” than active policeman in reference to the coalition bargain and that their “watchdog” function, theorized by Carroll and Cox (Reference Carroll and Cox2012), is a passive one.

SUPPLEMENTARY MATERIAL

To view supplementary material for this article, please visit https://doi.org/10.1017/S000305541800062X.

Replication materials can be found on Dataverse at: https://doi.org/10.7910/DVN/UVANPR.