Existing Datasets

While previous data-collection efforts have much improved our understanding of how governments' ideologies differ across space and time, the scope of existing datasets on the ideological orientations of political leaders has remained limited. They often cover exclusively or mostly industrialized democracies in Europe and North America (see, for example, Armingeon et al. Reference Armingeon2019; Brambor, Lindvall and Stjernquist Reference Brambor, Lindvall and Stjernquist2017; Jolly et al. Reference Jolly2022; Volkens et al. Reference Volkens2021) or in Latin America and the Caribbean (see, for example, Huber and Stephens Reference Huber and Stephens2012), and exclude leaders and parties in Africa, Asia and the Pacific, and the Middle East.

Other datasets tell us more about the ideological orientation of governments in these regions but have an otherwise limited coverage. Manzano (Reference Manzano2017) covers the ideologies of chief executives globally, yet only for dictators. The Global Party Survey (Norris Reference Norris2020) provides economic ideology measures for the largest political parties in most countries, yet only for the year 2019. Moreover, the Varieties of Party Identity and Organization (V-Party, Lührmann et al. Reference Lührmann, Tannenberg and Lindberg2020) provides ideology information for political parties in many countries since 1970, but it only covers election years.

The dataset with the widest coverage across countries and time, and the one commonly used by researchers, has been the Database of Political Institutions (DPI) (Cruz, Keefer and Scartascini Reference Cruz, Keefer and Scartascini2021). For 180 countries from 1975 until 2020, the dataset annually codes the economic ideologies of a country's chief executive and the largest government and opposition parties.

While far-reaching and fine-grained, the DPI's ideology data have many missing values, non-transparent and contradictory sourcing, and possibly tautological reasoning.Footnote ² It has high shares of observations without a coded ideology (approximately 40 per cent for chief executives and the largest government party), especially in non-democracies and in sub-Saharan Africa and the Middle East. Its specific sources are unclear, and its primary source for parties' ideologies – the editions of the Political Handbook of the World (Lansford Reference Lansford2019) – does not corroborate or even contradicts their own coding in a fair number of cases. Furthermore, its coding rules allow for a chief executive's ideology to be inferred from the policies they implemented, which risks making any analysis of the effects of political leaders' ideology on policymaking that relies on these data tautological.

Taken together, although existing datasets cover a fair share of countries and years, and can tell us much about the ideological orientation of political leaders, their coverage ultimately has remained limited. The next section therefore presents a dataset on chief executives' ideologies with both comprehensive coverage across time and space and refined coding procedures.

Data Contents and Collection

This letter introduces the Global Leader Ideology dataset, which identifies the chief executives, their parties and their ideologies in 182 countries for each year from 1945 or independence to 2020. The dataset codes the economic ideology and party affiliation of two types of chief executives: the head of government and the leader, that is, the politically most powerful individual.Footnote ³ To identify the heads of government, a team of research assistants and I used data from the Varieties of Democracy (V-Dem) project (Coppedge et al. Reference Coppedge2021) and supplemented them with information from Cahoon (Reference Cahoon2021), Schemmel (Reference Schemmel2021) and Lentz (Reference Lentz1994). I identified leaders with data from the Archigos project (Goemans, Gleditsch and Chiozza Reference Goemans, Gleditsch and Chiozza2009), with data for the years 2016–2020 added from Bell, Besaw and Frank (Reference Bell, Besaw and Frank2021). In many cases, the head of government and the leader are the same person. When they differ, the head of government tends to be primarily responsible for domestic policymaking, while the leader is often a head of state, focusing also or more on foreign policy. I coded the head of government and leader in office on December 31 of each year. In total, the dataset covers the heads of government in 178 countries from 1945 or independence to 2020, and the leaders in a set of 178 slightly different countries from 1945 or independence to 2020. The dataset thereby both covers chief executives in more countries and – with V-Party and the DPI starting in 1970 and 1975, respectively – covers them decades before existing datasets.

For these heads of government and leaders, the dataset distinguishes between leftist, centrist, rightist and no economic ideology, understood as the preferences over how much the state should intervene in the economy. Leftist chief executives are those that want the state to take an active role in the economy, while rightist chief executives want the state to rarely intervene. Centrist chief executives hold a middling position on this question.

Chief executives may be genuine believers or express these preferences for strategic reasons. Leftists may believe in increasing social equality and rightists in increasing freedom, or they may, in part, hold their views because they want to attract social support, such as from workers or business owners. I only make a nominal distinction between leftist, centrist and rightist chief executives because finer-grained measures – such as distinguishing between centre- and far-left ideologies – may be incomparable across the wide country and year coverage (Brambor and Lindvall Reference Brambor and Lindvall2018). I generally assumed that a chief executive did not sufficiently change their ideology over time to warrant a different coding. Importantly, distinguishing chief executives based on their economic ideology does not mean that they did not have other views, such as on the role of religion or ethnicity.

The ideology coding is based on many diverse sources. My research assistants and I drew on chief executives' own statements about their views and agendas, their personal background – such as membership in leftist student organizations – and descriptions in secondary sources of a chief executives’ ideology, including the datasets by Brambor, Lindvall and Stjernquist (Reference Brambor, Lindvall and Stjernquist2017) and Manzano (Reference Manzano2017). In many cases, we identified a chief executive's ideology by first identifying the party they were affiliated with and then which ideology the party had. Finally, in rare instances, we used chief executives' specific actions unrelated to policies, such as constitutional provisions for socialism, bans on leftist newspapers or close ties to other political leaders.

We also relied on numerous sources to code chief executives' parties. For chief executives' party affiliation, we used information by Mattes, Leeds and Matsumura (Reference Mattes, Leeds and Matsumura2016), Cahoon (Reference Cahoon2021), Schemmel (Reference Schemmel2021) and many other sources. To identify parties' ideologies, wherever possible, we used the datasets by Armingeon et al. (Reference Armingeon2019), Huber and Stephens (Reference Huber and Stephens2012), the Global Party Survey and V-Party, as well as information from Cahoon (Reference Cahoon2021), Döring and Manow (Reference Döring and Manow2020), the Perspective Monde project (Perspective Monde 2021), the DPI party codingFootnote ⁴ – which we checked against the Political Handbooks of the World – and many additional sources, including parties' memberships in international organizations, such as the Socialist International. If we found evidence that the chief executive's ideology deviated from their party's, we coded their personal ideology.

To ensure the validity of the coding, we went to great lengths to not infer a chief executive's ideology from their policies. We excluded any descriptions of implemented economic or social policies, such as the nationalization of companies or cuts to social services, and we disregarded sources that seemed to base their assessment on such policies. We also sought to distinguish rightist and centrist economic positions from other issue dimensions, such as disregarding sources that used ‘rightist’ or ‘centrist’ in terms of social matters, ethnicity or religion.

We further worked to make the coding consistent across sources and comparable across countries and years. We sought to base each coding on at least two sources, often having more. When sources disagreed, we worked to either bolster one of the views or to resolve the discrepancy, such as because the ideology of the chief executive and their party differed. We preferred sources covering several or many countries, sources describing a chief executive's views in detail, and academic and expert sources to newspaper reports. Finally, we used training vignettes, and two or more coders evaluated all the sources and the coding for each chief executive.

To make our coding transparent, we have written several hundred pages of country profiles. In addition to the information in the dataset – the annual head of government and leader, their parties, and their ideologies – the profiles list and often quote the sources used for each chief executive.

The dataset can easily be merged with other datasets on political parties and leaders by including a party's Party Facts ID (Döring and Regel Reference Döring and Regel2019) and indicators for whether the chief executives match those identified by Brambor, Lindvall and Stjernquist (Reference Brambor, Lindvall and Stjernquist2017), Manzano (Reference Manzano2017) and Mattes, Leeds and Matsumura (Reference Mattes, Leeds and Matsumura2016). Both the dataset and country profiles are available on my homepage (bastianherre.com/data) and GitHub (github.com/bastianherre/global-leader-ideologies).

Descriptive Statistics

The Global Leader Ideology dataset allows us to explore leaders' ideological orientations worldwide. It reveals that most chief executives have identifiable ideologies: Figure 1 shows that among the 10,708 country-year observations covered for heads of governments, we only could not identify an ideological orientation – or found sources explicitly stating that they had none – for 760 observations (7 per cent).Footnote ⁵ Examples are monarchs Khalifah ibn Hamad Al Thani of Qatar and Bhutan's Jigme Dorj Wangchuk. Among heads of governments with an identifiable ideology, leftist country-years constitute almost a majority (4,758 observations), with most other heads of government being rightist (4,206) and the remainder centrist (984). Examples of leftist chief executives include democrats Robert Fico of Slovakia and South Africa's Nelson Mandela, as well as dictators Gamal Abdel Nasser of Egypt and North Korea's Kim Il Sung. Rightist chief executives include democratically elected Japanese prime minister Shinzo Abe and Lithuanian President Dalia Grybauskaite, as well as autocratic President François Duvalier of Haiti and Ivory Coast's Felix Houphouet-Boigny. Centrist chief executives include President Bill Clinton of the United States, Bambang Yudhoyono of Indonesia and South Korea's Kim Dae-jung.

Fig. 1.

Heads of government's ideologies across regimes, 1945–2020.

The dataset also shows that ideologies differ across regimes. Figure 1, with regime data from Lührmann, Tannenberg and Lindberg (Reference Lührmann2018), demonstrates that heads of government with a missing ideology preside almost exclusively over non-democracies.Footnote ⁶ Whereas they constitute a sizeable minority (15 per cent) of democratic country-years, centrist governments are relatively rare in dictatorships. In democracies, leftist and rightist heads of government are about equally common, while in dictatorships, leftist heads of government are clearly more common than rightist ones. This questions the common assumption that dictatorships are political systems of and for the wealthy (see, for example, Acemoglu and Robinson Reference Acemoglu and Robinson2006; Acemoglu et al. Reference Acemoglu2015) and instead corroborates and expands research suggesting that autocracies differ not only in their political institutions, but also in the actors that run them (Manzano Reference Manzano2017).

When comparing the data on heads of government and leaders, I find that ideological cohabitation of chief executives is rare. Among the 26 per cent of country-years for which the head of government and leader differ, their ideological orientations differ in only 20 per cent of cases. The data thereby also speak to debates about the commonality of cohabitation and the pitfalls of systems of government (see, for example, Samuels and Shugart Reference Samuels and Shugart2010; Sedelius and Linde Reference Sedelius and Linde2018).

Validation

To externally validate the data, Table 1 compares the Global Leader Ideology dataset to datasets by Brambor, Lindvall and Stjernquist (Reference Brambor, Lindvall and Stjernquist2017) by Manzano (Reference Manzano2017), to the DPI, and V-Party. I find overlaps in a large majority of cases. Table 1 shows that when they identify the same chief executive, the ideologies match for 96 per cent of observations compared to Brambor, Lindvall and Stjernquist (Reference Brambor, Lindvall and Stjernquist2017), and for 94 per cent compared to Manzano (Reference Manzano2017). The data are more difficult to compare to the DPI and V-Party because the DPI does not identify chief executives by name and V-Party identifies the ideologies of parties. For the DPI, I therefore used its information on the system of government and assumed that their chief executive matches my leader in presidential systems and my head of government in parliamentary and semi-presidential systems. Under this assumption, the ideology matches about 81 per cent of the time.Footnote ⁷ To compare my data and V-Party, I collapse its data into leftist, centrist and rightist parties, and use Party Facts IDs to link them to my heads of government, thereby assuming that party and head-of-government ideologies match. This yields about 77 per cent matching observations. The main difference seems to be that Lührmann et al. (Reference Lührmann, Tannenberg and Lindberg2020) assign an outright centrist ideology to more parties in their dataset than does my dataset to its heads of government. Overall, my dataset yields very similar results for the chief executives also included in other datasets and similar results to datasets that are not as readily comparable.

Table 1.

Chief executives' ideologies across datasets

Illustrations

I further highlight the Global Leader Ideology dataset's uses by exploring how chief executives' ideologies differ across countries and over time. I find that leaders' ideological orientations have not been equally common around the globe. Figure 2 graphs the share of leftist and rightist heads of government for each country between 1945 or independence and 2020. The graphs show that leftist heads of government have been especially prevalent in South and East Asia, as well as Southern Africa, whereas rightist heads of government have been predominant in Western Europe and on the Arabian Peninsula. This again stresses that many governments beyond the commonly studied Organisation for Economic Co-operation and Development (OECD) have identifiable and diverse ideologies. The graph further shows that despite cross-regional and cross-national differences, many countries around the globe have experienced ideological changes and have been headed by both leftist and rightist heads of government.

Fig. 2.

Heads of government's ideologies per country, 1945–2020.

I further find trends across regimes and different dynamics within them. Figure 3 gives the share of political leaders of each ideology per year.Footnote ⁸ Across regimes, the graph shows that leftist political leaders became increasingly common in the first few decades after the Second World War, increasing from a low of about 30 per cent of all countries in 1950 to a high of more than 50 per cent in the mid-1980s. This contrasts with Brambor and Lindvall's (Reference Brambor and Lindvall2018) finding for advanced industrialized democracies that the political Left was relatively weak in the often-supposed ‘Golden Age’ of social democracy in the 1950s and 1960s. The global data instead suggest that leftist political leaders around the world were gaining ground during the Cold War. In the wake of its end, leftist leaders then became less and rightist ones more common, though political leaders until today have been most commonly leftist in almost all years. Since the mid-2000s, rightist leaders have indeed slightly lost ground to centrist leaders.

Fig. 3.

Heads of government's economic ideology per year, 1945–2020.

I moreover find that trends over time have differed between democracies and dictatorships. Again using regime data from Lührmann, Tannenberg and Lindberg (Reference Lührmann2018), I find that the trends for dictatorships and all regimes have been similar, as most countries have had non-democratic governments. Beyond these general similarities, however, leftist political leaders were even more predominant in dictatorships in the 1970s and 1980s, making up almost 60 per cent of all non-democratic leaders. Leftist leaders in democracies were only predominant in democracies in the early 1970s, fluctuated in the 1980s and 1990s, and were less common than rightist leaders in a fair share of years. Meanwhile, despite temporary gains in the 1990s and early 2000s, rightist non-democratic leaders have become less and less common, making up only about 20 per cent of all non-democratic leaders in 2020. This has gone along with more non-democratic leaders with no or a non-identifiable ideology. Finally, centrist non-democratic leaders have recently become more common, while centrist democratic leaders were prevalent (at times, around 30 per cent of all democratic leaders) in the 1950s and 1960s but have since been relegated to a stable, but ultimately limited, share of around 15 per cent of democratic leaders.