DOC PREVIEW
UI STAT 4520 - Bayesian Regression Model for Predicting Cabinet Duration

This preview shows page 1-2-3-4-5-6 out of 18 pages.

Save
View full document
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
View full document
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience
Premium Document
Do you want full access? Go Premium and unlock all 18 pages.
Access to all documents
Download any document
Ad free experience

Unformatted text preview:

Bayesian Regression Model for Predicting Cabinet DurationBayesian Statistics Project Final ReportGail Buttorff and Jessica DayDecember 10, 2007I. IntroductionConsiderable variation cabinet duration exists across (and even within) parliamentary democracies. Cabinets (or governments) can last a few days or last twenty years (King et al. 1990: 846). This variation has engendered prolific research which has sought to uncover the key covariates influencing cabinet duration. However, the literature remains divided. There are two schools of thought associated with this cabinet duration literature. The first school of thought views cabinet duration as a function of various attributes: country, cabinet and party-system attributes. Such attributes include the level of polarization (measured by support for extremist parties) and the number of parties holding seats in parliament (i.e. party fractionalization). Moreover, whether there is an investiture requirement, whether the government is a majority or minority government and the amount of time (number of days) the government took to form have all been theorized to influence cabinet duration. The beginning of this school of thought is attributed to Riker’s 1962 “seminal” paper (Strom 1988: 923). The second school of thought is often referred as the “stochastic” model. This literature proposes cabinet duration to be a function of “critical” or “terminal” events such as the death of the prime minister or various scandals involving the government. According to Diermeier and Stevenson, “since Browne et. al. [1984, 1986], this second aspect has been pursued through the concept of ‘critical events,’ exogenous shocks that destabilize an existing government” (Diermeier and Stevenson 1999:1051). In addition to the debate between the two schools of thought, there also exists disagreement concerning which variables are the most influential and the direction of that influence. This project will focus on the first school of thought using a Bayesian framework to test four hypothesized covariates of cabinet duration: investiture requirement (investiture), whether the government is a majority or minority government (numerical status), the level of polarization (polarization) and the number of days taken to form the government (crisis). In particular, we are interested in the following research question: What is the probability that a government lasts until within 12 months of the constitutional interelection period (CIEP)? The CIEP refers to the maximum time period that can occur between elections as mandated by the constitution. In mostcountries this maximum time period is four years. Our data come from a replication dataset compiled by King et al (1990). There are 313 observations (governments) in fifteen parliamentary democracies for the years 1945 until the endof 1987. The countries included fourteen Western European parliamentary democracies as well as Israel. Average cabinet duration amongst these countries ranges from 4.7 months in Finland to 30.8 months in Ireland. 1II. Model SpecificationAs stated previously, we are interested in the affect of investiture, polarization, numerical status and crisis on the probability that a cabinet “survives” within 12 months of the CIEP. That is, we are interested in the factors that cause a governing cabinet to dissolve or face an election before itlegally has to according to the constitution. Descriptive statistics for all variables are given in Appendix A. Our response variable is a Bernoulli random variable measuring whether or not a cabinet lasted until within 12 months of the CIEP. A success outcome (1) occurs when the cabinet lasts until the CIEP. A failure (0) occurs when the cabinet dissolves or calls an election prior to the CIEP. Our four predictor variables are defined as follows:Investiture This variable measures whether or not the country has investiture, which requires the parliament to cast a formal vote approving (or not) of the newly formed government. We treat investiture asa binary response variable: if a county has this requirement, it is coded 1 and 0, otherwise. King et al. argue that investiture should have a negative impact on average duration “by causing some governments to fail very quickly” (King et al. 1990: 857). Party Polarization Polarization measures support for extremist parties within the electorate. This variable can take on any positive real values. In this dataset, the values range from 0 to 43. Polarization is considered an important measure of the complexity of the bargaining situation in the legislature (King et al. 1990: 858). Thus, greater polarization should lead to shorter durations. Numerical Status Numerical status measures whether or not the cabinet has a majority or minority coalition. Numerical status is again coded as a binary response variable: a government is coded 1 if it is a majority government and 0 if the government is a minority government. A majority governmentoccurs when the governing party or coalition of parties holds at least fifty percent of the seats plus one seat in the legislature. A minority government occurs when the governing party or coalition of parties holds less than a majority of seats in the parliament. Majority governments are expected to last longer given their numerical advantage. Crisis Duration Crisis duration measures the number of days after the election it takes a government to form. This variable ranges from 0 to 274 days. King et al. argue that, “a longer crisis indicates a more difficult bargaining situation in which it is easier for cabinets to collapse, suggesting diminished cabinet duration” (King et al. 1990: 858-59). However, Strom “argues and finds” the opposite. Since our response variable is a binary outcome variable (0-1), we use WinBUGS to run a logit regression model. Our model has the following form: Logit(CIEP12) = α+β1*Investiture + β2*Polarization + β3*Numerical Status + β4*Crisis Duration2III. ResultsFirst, we use Stata 9 to run a frequentist logit regression with our four predictor variables for comparison. We expect to find the similar mean values for our parameter estimates as we would find using non-informative priors in a Bayesian model. The results are shown in Table 1. For our Bayesian model, we use two sets of priors. First, we ran the model using vague, non-informative priors to compare to our frequentist model. For our


View Full Document

UI STAT 4520 - Bayesian Regression Model for Predicting Cabinet Duration

Documents in this Course
Load more
Download Bayesian Regression Model for Predicting Cabinet Duration
Our administrator received your request to download this document. We will send you the file to your email shortly.
Loading Unlocking...
Login

Join to view Bayesian Regression Model for Predicting Cabinet Duration and access 3M+ class-specific study document.

or
We will never post anything without your permission.
Don't have an account?
Sign Up

Join to view Bayesian Regression Model for Predicting Cabinet Duration 2 2 and access 3M+ class-specific study document.

or

By creating an account you agree to our Privacy Policy and Terms Of Use

Already a member?