The database contains the following variables:
- Country (3 letters abbreviation)
- Year
- Total vote share for left-wing populist parties
The data are from Timbro
Data are for legislative elections.
When a country uses a two-round system, first-round results are
used.
Countries CAN have multiple elections in a year.
- Total vote share for right-wing populist parties
Same comment as for left-wing populists
- Election date
Variable stating the date of the election. If two-round elections, first-
round date is considered. If elections are held over two days, later
date is considered.
- Terrorist data
Data on all attacks are taken from the Global Terrorism Database
Data on Islamist attacks are taken from the Fondation Pour
l'Innovation Politique
Data on left-wing and right-wing terrorism have been hand-coded in
two steps: (1) When the terrorist group carrying the attack is known,
the group has been classified left or right (see file here where a
value of one designates a left-wing group and a value of zero a
right-wing group
(2) when the group is unknown, classification has been done by
looking at the motive of the attack
All data are aggregated yearly relative to the election date for each
country.
Attacks look at the number of attacks for all terrorist attacks
nkill look at the number of killed citizens for all terrorist attacks
nwound look at the number of wounded citizens for all terrorist
attacks
leftproxy rightproxy are the number of attacks by left-wing and
right-wing groups/individuals, respectively
nkill_left nwound_left nkill_right nwound_right are the number of
deaths and wounded linked to groups with left-wing/right-wing
ideology
IslClaim are all attacks claimed by Islamist movements as
categorized by the Fondation Pour l'Innovation Politique
IslPoss are all attacks which can be attributed to Islamist
movements as categorized by the Fondation Pour l'Innovation
Politique
nkill_Isl* and nwound_Isl* are the number of deaths and wounded
linked to the different type of Islamist attacks
Home* means attacks that happen in the country where the election
is held Y1 to Y4 indicate the number of years to election (e.g., Y1
means that the attacks/n death/ n wounded are in the year to the
election (in the 365 days prior to the election in X))
You also have attacks in other countries which took place in the
years prior to the election in country X.
The name of these variables have three components: variable type,
country attacked, year(s) prior to election (e.g., AttacksITAY1
captures the attacks who took place in Italy in the year prior to the
election in country X
nkillICEY3 captures the number of killed in Iceland in the third year
(1095 days to 730 days) prior to the election in country X)
- Language
The language similarity measures are taken from the dataset
prepared by Toubal and Melitz, introduced in their Journal of
International Economics paper (2014).
Three measures can be used: prox1, prox2, and cle which are
described on pages 354-355 of their article
Note that there is no data for Serbia and Montenegro. Belgium and
Luxembourg are also equivalent in the data, so they have similar
indices
A variable name has two components: type of variable (prox1,
prox2, cle) + comparison country
- Border
Data source
The dummy takes a value of one if country X borders the country
listed under the variable Country
Note that a country is not considered as bordering itself (e.g., the
variable BorderAUT has a value of 0 for country AUT, the variable
BorderUK has a value of 0 for country UK).
Only neighbors included in the dataset are considered (e.g., Ukraine
is not considered though it borders Poland for example).
BorderITA, for example, would take value 1 if a country borders Italy
and 0 otherwise
- Cultural similarity
The cultural similarity index is constructed using the data from
Michalopoulos and Xue's paper "Folklore" published in the QJE
The paper computes the number of tales containing certain motifs
(23 different motifs in total). We use the country level data
(MTurk_countries.dta)
Two indexes are computed:
o 1) normalised_culturalsimilarity uses the number of tales
using the motifs normalised
o 2) percent_culturalsimilarity use the percentage of tales using
a motif per country.
In both cases, a difference index is computed by taking the average
distance over each motif. A similarity index corresponds to one
minus the difference
A variable name, like above, has two components: the similarity
index used + the abbreviation of the comparison country
- Unemployment, etc. are a set of possible controls
Percentages are expressed in percent (so 9 would mean 9%)
Economic variables are expressed in millions euros
More details on the variables can be found below
Unemployment: Unemployment rate (Source: Eurostat)
Import: Yearly import (Source: Eurostat)
Export: Yearly export (Source: Eurostat)
Tertiary: % over 25 with tertiary education (Source: Worldbank)
population: population (Source: Eurostat)
marriage: Crude marriage rate (source:Eurostat)
immigration: immigration (Source: Eurostat)
gdp: gdp (Source: Eurostat)
emigration: emigration (Source: Eurostat)
Divorce: Crude divorce rate (Source: Eurostat)
Over65: % pop over 65 years old (Source: Eurostat)
Growth: GDP growth % on yearly basis (Source: Eurostat)
Export_penetration: Export as % of GDP (Source: Eurostat)
Import_penetration: Import as % of GDP (Source: Eurostat)
Immigration_percent: Immigration as percent of pop (Source:
Eurostat)
Emigration_percent: Emigration as percent of pop (Source: Eurostat)