Click here for Data Set

Sample
We use IFS data for all the countries with available central government data on the Deficit, Total Expenditure and Total Revenue (including Grants). Where IFS data are missing we tried to complement them by using GFS data or alternative sources. A detailed list of all the adjustments made to the data appears in Table A1.
To restrict our sample only to democracies, we include only the observations with a non-negative score in the POLITY IV Level of Democracy index, which is produced by the University of Maryland.  Hence, only data points with a score of 0 and above are left in the sample.
In the former socialist economies in Eastern Europe and the former Soviet Union we exclude the observations for the first two years after transition, as they may represent the simultaneous effect of the shift to democracy and the collapse of central planning, rather than political manipulation of fiscal variables. The elections in these countries which are included in the sample are listed in Table A2 of the paper..

Fiscal policy variables
The dependent variables are the following: Balance- calculated as the difference between Total Revenue & Grants and Total Expenditure. Total Expenditure- taken from the IFS dataset. Total Revenue & Grants- calculated as Revenue plus Grants from the IFS dataset.  All these variables are presented as a percentage of GDP, the latter also taken from the IFS dataset.

Election variables
The data on election years and dates, are mainly retrieved from the Institute for Democracy and Electoral Assistance (IDEA), "Voter Turnout Since 1945 to Date" (http://www.idea.int/vt/index.cfm). Additional sources are: The International Foundation for Electoral Systems (IFES- http://www.electionguide.org), The Database of Political Institutions (DPI) Version 2000, (a project conducted by the World Bank) and are complemented by other political data sources.
Our election year variable Elect- is a dummy variable that receives the value 1 in the election year and 0 otherwise.
All our estimations contain fixed country effects, as well as one lag of the dependent variable.  Fixed year effects were tested and removed since they were not statistically significant and have not affected the main results.

Economic control variables
Trade- the share of international trade, as a percentage of GDP, taken from the World Development Indicators (WDI) 2002 publication of the World Bank.
Lgdp_pc - the log of real per-capita income. The data for 1975-2001 are taken directly from the WDI dataset (mentioned above). The data for the years 1960-1975 are computed using the WDI "GDP per capita in constant 1995 US$" series.
Pop1564, Pop65 - Two demographic variables measuring the fraction of a country's population, ranging between 15 through 64, and above 65, respectively.
Gdp_rhp - A measure of the output gap, calculated as the difference between real GDP and its (country specific) trend.  The trend was computed using the Hodrick-Prescott filter on the change in real GDP.  Real GDP data were extracted from the WDI dataset in constant 1995 US$.

Presidential vs. Parliamentary constitutional rules
The DPI database provides information whether the chief executive responsible for economic policy, in each country and in each election year, is elected directly by the public or by parliament.  In the former case we define the electoral rule as Presidential and in the latter as Parliamentary, as in Persson and Tabellini (2002).  Based on this distinction between the electoral rules we computed the following variables:
Pres - receives the value 1 in a Presidential electoral system, and 0 otherwise.
Parl - receives the value 1 in a Parliamentary electoral system, and 0 otherwise.
Elect_pres - an interaction between Pres and Elect= (Pres)*(Elect).
Elect_parl - an interaction between Parl and Elect= (Parl)*(Elect).
When estimating the Presidential vs. Parliamentary equation, we use both Elect_pres and Elect_parl variables, together with the economic control variables, and one lag of the dependent variable.

Proportional vs. Majoritarian electoral rules
The DPI database provides information, in each country and in each election year, whether candidates are elected based on the percent of votes received by their party. In this case they define the electoral system as Proportional representation and in the other case we define the electoral system as Majoritarian representation. Based on the distinction between the electoral systems we computed the following variables:
Prop – receives the value 1 in a Proportional electoral system and 0 otherwise.
Maj – receives the value 1 in a Majoritarian electoral system, and 0 otherwise.
Elect-prop – an interaction between Prop and Elect = (prop)*(Elect)
Elect-maj – an interaction between Maj and Elect = (Maj)*(Elect).

Level of democracy
The analysis regarding the level of democracy was based on the score of each country in the POLITY IV dataset. We split the sample between these countries with a score of 0 to 9 and those with a score of 10, because more than 50 percent of the data points represent countries with a score of 10. Where the score changed during the covered period, we split the data points for that country according to the score in each year.

Predetermined vs. Endogenous elections
Based on www. electionworld.com data, that indicate the frequency of elections in each country, we determined when the next elections should have been held. If the election were held in the expected year we classified them as predetermined; otherwise they were classified as endogenous.