https://orcid.org/0000-0002-7410-7201
ABSTRACT
Assessing how democracy varies within countries is paramount to the subnational turn in comparative politics. Despite recent contributions, we still lack a comparable measure of democracy for provinces inside countries. To overcome this limitation, I present the Index of Subnational Electoral Democracy (ISED), a measure that tracks the electoral dimension of democracy across the provinces of nine Latin American countries, the United States, Canada, and India for a period of roughly 40 years, making it the largest dataset on subnational regime outcomes to date. I then use the ISED to assess the democratic trajectories of Argentinian, Brazilian, Mexican, and Indian states, revealing that: 1) Indian provinces have been, on average, more democratic than their Latin American counterparts. 2) The relative position of provincial regimes within these countries has been remarkably stable over time. 3) Most subnational units in the Americas have had “low intensity” regimes. 4) Subnational regime hybridity has been the norm rather than the exception, and that 5) for the Latin American cases under consideration, democracy and development are positively connected at the local level. I conclude by outlining the ISED's research applications and reflecting on the implications of these five conclusions for future research on subnational democracy.
KEYWORDS:
Introduction
Scholars interested in exploring democracy and autocracy at the country-level have a diverse and well-known set of measures to capture regime variation across time and space.Footnote1 While the “subnational turn”Footnote2 has generated invaluable insights on the causes and consequences of the regime heterogeneity observed within countriesFootnote3, we lack a common framework and comparable data to assess democracy systematically and cost-effectively at the subnational level. In this regard, the contributions of the Varieties of Democracy (V-Dem) initiative have been invaluable.Footnote4 However, the set of variables introduced by the V-Dem project are still aggregate, country-level scores. In other words, the V-Dem project does not include data for subnational units such as states, provinces, or municipalities. This has limited our ability to evaluate the impact of local socio-economic factors on subnational democracy, and to trace the regime trajectory of individual subnational unitsFootnote5 through time.
In line with recent contributions to DemocratizationFootnote6, to help overcome this limitation, in this article I present the Index of Subnational Electoral Democracy (ISED) and then use it to comparatively assess provinces across Argentina, Brazil, Mexico and India, four paradigmatic cases of the literature on subnational regimes. To construct the ISED, I follow Munck's and Verkuilen'sFootnote7 classical approach of conceptualization, operationalization, and aggregation. In relation to the former, following Robert DahlFootnote8, and Coppedge, Álvarez, and MaldonadoFootnote9, I identify contestation and participation as the two constitutive dimensions of electoral democracy. In relation to both operationalization and aggregation, I adopt the framework articulated by Tatu Vanhanen.Footnote10 In sum, the ISED uses first round electoral data –from either local executives or local legislatures – to capture democracy in a total of 308 subnational units, covering the United States, Canada, nine Latin American countries and India for a period of roughly 40 years. With over 8,787 province-year observations, this is the most temporally and spatially ambitious project of its kind to date.
While the online material replicates the analysis for all the country-cases contained in the dataset, to keep the discussion focused – and contribute to the debate – in this article I zoom in on the provincial dynamics of Argentina, Brazil, Mexico, and India, four cases that have shaped our understanding of subnational regime heterogeneity.Footnote11 In doing so, I present four descriptive arguments: First, for the past four decades, Indian provinces have been, on average, more democratic than their Latin American counterparts. Second, the relative position of provincial regimes within countries has been remarkably stable over time. States that were more (less) democratic in the early 1980s continue to be so today. Third, most subnational units in the Americas have had “low intensity” regimes, in which levels of electoral participation sit systematically below those of electoral competition. Fourth, for at least the past forty years, subnational regime hybridity has been the norm rather than the exception. Fifth, the evidence suggests that democracy and development are connected at the local level. While more research is needed, there is enough evidence to contend – à la LipsetFootnote12 – that across Argentina, Brazil, and India, the more well-to-do a subnational unit, the more likely it is that it sustains a democratic regime.
The article is structured as follows: In the first section I review existing measures of subnational democracy, highlighting their strengths and weaknesses. I then discuss my conceptualization and operationalization strategies, case selection, along with the sources used to construct the new index. Afterwards, I address validity concerns. In the fourth section of the article, I pay particular attention to Argentinian, Brazilian, Mexican, and Indian provinces. Specifically, I look at the democratic trajectories of subnational units, I put forth a regime classification strategy, and I explore the association between democracy and development at the local level. I conclude by outlining some of the ISED's limitations, and by reflecting on the implications that the four substantive conclusions have for future research on national and subnational democracy.
Strength and weaknesses of existing measures
Scholars have put forth a variety of measures to explore the heterogeneity of subnational regimes. In , I summarily present ten examples relevant to our purpose. I highlight the democratic dimensions covered, the variables or indicators used, the procedures followed for aggregation, as well as the cases and periods covered. These measures represent valuable steps towards capturing subnational democracy as they: a) tackle several dimensions of democracy, b) use a rich selection of variables and indicators, and c) track a considerable number of years. All the indices reviewed here are treated as interval-level measures, suggesting that dichotomous conceptualizations and scores are ill-fitted to capture subnational regime variation. Similarly, the list includes both subjective – expert-based – and objective measures. Moreover, an overview of reveals that scholars have used both straightforward mechanisms for aggregation, such as addition and multiplication, and more complex tools like principal component analysis (PCA), structural equation modelling (SEM), and factor analysis (FA).
Table 1. Existing measures of subnational democracy.
The depth of these measures, however, comes with a trade-off vis-à-vis their breadth and usability. First, conceptually, although most scholars adopt a Dahlian understanding of democracy, they nonetheless deploy distinct conceptual dimensions. A glance at the second column of reveals that along contestation, and participation, scholars of subnational democracy have also conceptualized provincial (fiscal or political) autonomy, constitutional constraints, and liberal or civil rights as constitutive dimensions of subnational democracy.
Second, even in cases where the conceptual dimensions are similar, the variables and indicators used for operationalization tend to differ. Third, when dimensions and variables do coincide, either the aggregation mechanisms differ, or they are not strictly transparent. Fourth, on a more practical note, these indices are quite demanding in terms of data collectionFootnote13, a fact that diminishes their portability, and increases the costs of keeping them updated.
In addition, while these measures can be used for the analysis of specific countries, this state of affairs makes subnational, cross-country comparisons – what Jefferey Sellers called transnational comparisonsFootnote14 – difficult if not impossible. This in turn hinders our ability to accumulate knowledge on the causes, the consequences, and the trajectory of democracy across individual subnational units. Based on Dahl's and Vanhanen's framework, in the next section I present a strategy to overcome the limitations previously outlined.
The index of subnational electoral democracy (ISED)
In this section I discuss first the conceptualization of democracy behind the ISED. I then focus on its operationalization, addressing the caveats and critiques that are commonly raised when using Vahanen's framework, and discuss my country-level case selection rationale. I conclude this section by assessing the ISED's validity, contrasting it with other existing measures and the relevant country-level aggregate items of the Varieties of Democracy project.
Conceptualization
Democracy is one of the best examples of an essentially contested concept, for it invites disputes that, “although not resolvable by argument of any kind, are nevertheless sustained by perfectly respectable arguments and evidence”.Footnote15 Although there is indeed a plethora of respectable definitionsFootnote16, following Dahl, “we can think of [democracy] as made up of at least two dimensions: public contestation and the right to participate […] [consequently] a regime may be located, theoretically, anywhere in the space bounded by these two dimensions”.Footnote17
Contestation signals the extent to which citizens’ preferences are weighed in the conduct of government. Inclusiveness, for its part, signals the share of the population entitled to take part in the system of public contestation. In other words, it reflects the breadth to which citizens can control and contest the conduct of the government.Footnote18 Empirically, Coppedge, Álvarez and MaldonadoFootnote19 have shown that when examining several of the cross-national measures available, contestation and inclusiveness are indeed the two persistent dimensions of democracy. However, several approaches have been put forth to measure them.
Operationalization, aggregation and data sources
I operationalize Dahl's democratic pillars using the framework articulated by Tatu Vanhanen, in which contestation is captured by electoral competition and inclusiveness is proxied by electoral participation. Within this framework, then, “[t]he competition variable is calculated by subtracting the percentage of votes won by the largest party from 100 […] [and] The participation variable is calculated as [the share of voters] from the total population”.Footnote20
This is in line with a long scholarly tradition that looks at electoral outcomes as indirect proxies for a set of citizens’ rights and institutional guarantees.Footnote21 Rather than directly capturing the presence or absence of any given right, by focusing on electoral outcomes, one assumes that elections are underpinned by Dahl's liberal institutional scaffolding. If elections are contested and highly participatory, the logic goes, it must be because – to a considerable extent – citizens are free to form and join political organizations, and they have the right to vote and be voted. Contrarily, if elections are not contested and show low levels of participation, it must be because – to a considerable degree – the rights and freedoms of citizens have been significantly curtailed.
captures the two components of Vanhanen's index, their ontological link to the two dimensions of democracy, and their relation to Dahl's procedural-institutional underpinnings. The diagram shows that, while there might be additional elements to a democracy, electoral competition and participation are fundamental. For this reason, while the dashed rectangles do not fully overlap with contestation and inclusiveness, they occupy a sizeable amount of their area. The diagram also shows that, in terms of abstraction (left-downwards arrow), competition and participation are meso-level concepts, and that, within the framework of non-direct, representative rule (right upwards arrow), they constitute the output of a complex institutional scaffolding. The intersection of contestation and inclusiveness shows that together, they ontologically configure democracy as an outcome. Lastly, the dashed double-headed arrow at the bottom emphasizes that each individual component and dimension can be experienced and measured to different degrees.
Figure 1. Conceptualizing subnational democracy.
The electoral components are salient because they capture the effective political implications of changes in civil rights and freedoms. For example, a higher competition score, that is, a more contested election, reflects the absence of a dominant or hegemonic political force, thus signalling a more balanced weighing of individual preferences. Similarly, a higher participation score, that is, higher levels of electoral turnout (weighed by the population), reflects ample levels of political engagement and that a multiplicity of voices are able to signify and express their political preferences.
Addressing caveats and critiques
Scholars have criticized Vanhanen's framework for being too minimalist. To be clear, I am not claiming that the electoral components for contestation and inclusiveness are the sole constitutive elements of democracy. What I am asserting, however, is that jointly, Dahl's and Vanhanen's framework are, conceptually, empirically, and practically most useful when scaling down regime outcome analyses. In other words, what the ISED captures is only the electoral dimension of democracy and in doing so, it maximizes the overlap between the conceptual and empirical pillars of democracy, as well as our ability to capture regime variation across time and states.
Comparativists might readily accept the necessity of a competition component – it has, after all, a more straightforward interpretation and has been the focus of most measurements and indexes of democracy.Footnote22 However, sceptical readers might suggest adding an “alternation in power” proxy to the competition component, and they might additionally find it hard to believe that low levels of participation necessarily mean less democracy. After all, apathy, or lack of engagement with the political process is part and parcel of normal democratic politics.Footnote23 Moreover, given that provinces are embedded in national regimes, legal barriers to participation – such as suffrage exclusions – are unlikely if not implausible. These observations do have some basis. However, there are strong reasons to adopt, rather than adapt, Dahl's two dimensions and Vanhanen's electoral operationalization of the former.
First, while occasional apathy might not threaten democracy, and suffrage exclusion is indeed generally outside the purview of subnational jurisdictions, a temporally enduring or a systematically low participation score can signal informal barriers to political inclusion, such as voter intimidation, co-optation and other forms of soft repression.Footnote24 Second, as suggested earlier, operationalized electorally, these dimensions can be scaled down to capture territorially distinct regime outcomes within countries without sacrificing conceptual or empirical accuracy.
For instance, other factors or conditions sometimes included in subnational measures do not solely belong to the territorial boundaries of the units under scrutiny. For example, executive alternation or local executive constraints are important for subnational democracy, but it is unclear whether they always have a defining, ontological connection to it. After all, local alternations can be forced from above, and as the literature has shown, this does not always mean democratic change. Moreover, the validity of alternation as an indicator of democracy relies on the notions of popular self-determination and sovereignty, and subnational units, while increasingly politically autonomous, are not sovereign. Moreover, any assessment about governance, effective number of parties, or constraints on the executive, are bound to depend on country-level systemness. The functioning of the courts or the interaction between local and national party branches are factors that do not have clear spatial boundaries. This territorial conflation warns against including such factors when conceptualizing and measuring subnational regimes.
In this regard, some might rebut that local elections and campaigns are influenced by national events. However, the coattails effect would influence the competition value of the index during concurrent election years and consequently, any potential noise can be taken into account and controlled for during statistical analysis.Footnote25 Moreover, that local elections and campaigns are influenced by national ones does not invalidate the ontological association between the two dimensions of democracy and their corresponding electoral indicators.
Vanhanen's operationalization framework has been additionally critiqued for penalizing majoritarian electoral systems.Footnote26 To guard against this, I specifically look at the first round of elections for the most salient subnational political office. That is, for the presidential systems of Latin America and the USA, I look at first-round elections of local executives. For the parliamentary systems of Canada and India, I use the electoral results for local legislatures. This maximizes the comparability across cases, as well as the comparability between this new data and the extant scholarly work on subnational democratic variation.Footnote27
Finally, measures focused on the electoral dimensions of democracy are usually accused of side-lining substantive and informal regime dynamics. Indeed, previous research has shown that at the national and the subnational level illiberal, informal practices hide behind the façade of formal democratic institutions.Footnote28 As stated previously, there are relevant aspects to democracy beyond its electoral dimension, and the ISED cannot capture them. However, given enough time, we can expect some informal practices to influence electoral participation and competition. For example, just as temporally enduring or systematically low levels of participation can signal soft repression, temporally enduring or systematically high values of participation could signal vote-buying or clientelistic “poll-packing”.Footnote29 In this regard, the longitudinal and cross-sectional aspect of the ISED – having repeated observations through time across individual subnational units – can help to identify these types of instances, even if indirectly. As such, rather than adjusting the measure a priori, this is something researchers can take up as a matter for empirical exploration.
Before introducing the sources and the panel structure of the data, it is also important to comment on my country-level case selection strategy. In addition to a substantive interest in the Americas, my decision to focus on the USA, Argentina, Bolivia, Brazil, Canada, Colombia, Ecuador, México, Peru, Uruguay, Venezuela, and India, for this first version of the ISED, follows a practical and a theoretical rationale. First, for more than half of these countries we previously did not have a measure of subnational regime outcomes. Second, second-tier governments in these countries have ample political and administrative autonomy.Footnote30 Third, these cases have occupied a predominant place in the literature on subnational regime variation. Fourth, by including countries like India, Uruguay, and Colombia, I aim to demonstrate that the ISED framework can be used in more than one region, and that it can be used to capture democracy in federal and decentralized unitary countries.
That is, it was important to present data for new and well-known country-cases in the Americas, while also keeping this first iteration of the ISED anchored in country-cases where second-level subnational units have historically and institutionally configured a political arena of their ownFootnote31, and for which we have accessible and reliable electoral data. This point on data accessibility and reliability, along with considerations of chronological comparability, were similarly important to establish the temporal scope of this first rendition of the ISED. As such, future research can then push the agenda further by 1) extending the ISED's temporal scope, 2) using the ISED's framework to measure subnational democracy in other countries, and by 3) scaling further down the territorial scale to capture electoral democracy across municipalities and other third-level units.
Overall, despite its limitations, my joint use of Dahl's and Vanhanen's framework offers five concrete advantages: 1) it captures two theoretically relevant dimensions of democracy, 2) it allows these dimensions and their measures to travel down the territorial scale, 3) it enables the cost-effective collection of cross-sectional and longitudinal observations, 4) it uses objective and accessible electoral data, and 5) it has a transparent mathematical aggregation procedure.Footnote32 The ISED is an interval-level measure bounded between zero and a hundred. Its interpretation is intuitive: higher scores correspond to open, democratic polities, and lower ones signal closed, autocratic regimes. As such, the ISED captures the levels of electoral democracy in 308 provinces across 12 countries for the past four decades, making it the largest dataset on subnational regime outcomes to date.
Panel A in summarizes the ISED's dimensions, their operationalization, as well as the indicators and the aggregation mechanism used to construct the index. Panel B lists the core critiques that are usually raised when using Vanhanen's framework, along with the specific rebuttals offered, and the overall strengths of the measure previously highlighted. Panel C then displays the main sources of electoral data.
Table 2. Measurement strategy: strength, weakness, and data sources.
Electoral data for Argentina come from Andy Tow's Electoral Atlas.Footnote33 For Brazilian and Bolivian units, electoral results come from their corresponding Superior Electoral Courts (TSE). I obtained Colombian, Ecuadorian, Peruvian, and Venezuelan electoral information from their corresponding National Electoral Councils (CNE). For Mexican states, I obtained the corresponding data from the Centre of Research for Development (CIDAC). To capture local legislative elections in Canada and gubernatorial ones in the USA I scraped the websites of local electoral departments, Wikipedia, and the Inter-university Consortium for Political and Social Research (ICPSR).Footnote34 I followed a similar procedure for India, for which I used data from the Electoral Commission (ECI) and the IndiaVotes website. Uruguayan electoral records were obtained from the National Electoral Court (CE).
I complemented and verified the information with national and local electoral institutes. Local population values to build the participation component were obtained from national statistics agencies such as the Census Bureau, INDEC, IBGE and INEGI.Footnote35 Overall, the resulting dataset is an unbalanced short panelFootnote36 with 308 groups, over 2200 elections, and a total of 8,787 province-year observations (after linearly interpolating values for the years between electoral cycles). summarily presents this information disaggregated by country.
Table 3. Panel data structure.
Assessing the ISED’s validity
Does the ISED manage to capture territorially distinct regime outcomes within countries? Answering this question requires tackling two fundamental measurement concerns: reliability and validity. The former alludes to consistency in measurement while the latter refers to the degree to which said measure accurately captures the phenomena under investigation. The ultimate test for the former is intercoder reliability: if several scholars code the same case in the same way, that measure is deemed to be reliable. Conversely, the conventional strategy to test for validity is through the correlation with another measure of the same or a similar concept. By using readily available, electoral data, the ISED maximizes replicability, which is why here I focus on validity.
To show that the ISED does capture how democratic (or not) provinces are, I contrast it with the measure put forth by FidalgoFootnote37, since her measure is designed to capture subnational electoral democracy in federal countries. Here, and throughout the rest of the article, I only focus on Argentina, Brazil, Mexico, and India – four large, developing economies and federal democracies that have been emblematic of the debate on subnational regime variation. However, in the online appendix I further assess the ISED's validity by contrasting it with the measures put forth by Caroline BeerFootnote38, MonteroFootnote39, GervasoniFootnote40, and GrumbachFootnote41, and I also replicate all the analyses that follow for the full set of country-cases contained in the dataset.
The correlations are illustrated in the different Panels of . The graph on the top-left displays the association between Fidalgo's score and my data for the Argentinian provinces. The top-right graph contrasts states in Brazil, the bottom-left corner plots the Mexican case, and the bottom-right graph shows the association for Indian units. In all instances, the correlations are positive and significant, ranging from 0.34 for the Mexican case to 0.76 for Brazilian units. In fact, except for Canada, contrasting the full set of cases with all available alternative measures yields positive, and significant correlation coefficients which roughly average 0.45. Contrasting the ISED's scores for the Canadian case with Fidalgo's measure yields a correlation coefficient indistinguishable from zero. However, unpacking this into the individual components of the ISED reveals that while the participation component has a negative, non-significant relation to Fidalgo's score, the competition component of the ISED is associated with Fidalgo's alternative measure, with a positive and significant Pearson correlation coefficient of 0.89.
Figure 2. The index of subnational electoral democracy: a validity assessment.
The index of subnational electoral democracy neatly identifies provinces long considered democratic over and under performers, which grants the measure face validity. For example, for Argentina, Jujuy, Salta, and Santiago del Estero are identified as under-performers, while Santa Fe, and Buenos Aires lie at the other end of the spectrum. For Brazil, there is a divide between southern and north-eastern states, with Bahia and Ceará systematically receiving low scores, while Rio de Janeiro and Rio Grande do Sul consistently achieve higher values. For India, Gujarat and Sikkim are identified as democratic under-performers, while Kerala and Puducherry are clear democratic achievers.
Lastly, looking at the USA, the ISED picks up the legacies of the “Solid South” or “Dixie's Land”. South Carolina, Tennessee, Georgia, and Alabama are identified as longstanding under-performers. In the online appendix I report tables with the descriptive statistics for each individual province by country. Scholars interested in the USA will note that against the narrative of American exceptionalism, USA states are not the top performers on this metric. This lends credence to Levistsky's and Ziblatts's characterization of US states as “laboratories of authoritarianism”.Footnote42 While moderate voices suggest that “the empirical pattern of subnational democracy in the United States is not one of recent erosion, but [one of enduring] unevenness”Footnote43, Grumbach's recent book Laboratories Against Democracy, and Mickey'sFootnote44 recent article on the persistent challenges of subnational democracy in the US, point towards the need to further explore the status of democracy within the North American giant.
As a final validity check, I contrast country-year aggregates of the ISED with items of the Varieties of Democracy Project (V-Dem v.12).Footnote45 Specifically, I compare both the level and the unevenness of subnational democracy. To examine how well the ISED captures the levels of local democracy, in the plots in I set it side-by-side with the V-Dem item v2elffelr, which asks country experts whether they consider subnational elections to be free and fair.Footnote46 As with earlier figures, in clockwise order, the plots track the Argentinian, Brazilian, Mexican, and Indian cases. In all instances the solid line is the country-year mean of the ISED, and the dashed line corresponds to the V-Dem values.
Figure 3. The ISED and V-Dem items: capturing the level of subnational democracy.
In terms of tracking the level of subnational democracy within a given country, reveals that for Brazil, Mexico, and India, the ISED is positively and significantly associated with the V-Dem item. In fact, this is the case for Colombia, Peru, and Venezuela also. The ISED and the v2elffelr V-Dem item are negatively associated for cases in Argentina, Bolivia, Canada, and the USA. The discrepancy highlights that the ISED captures the electoral component of democracy only, while the V-Dem item taps into the liberal dimension of subnational regimes. This suggests that while factors such as polarization might be driving electoral competition and participation up, polarized politics have also negatively impacted the liberal tenets of democracy in cases such as Argentina and the USA.
To assess how well the ISED captures the unevenness of a subnational democracy, that is, to determine if we can use the index to measure the heterogeneous territorial spread of democracy, I first obtained country-year estimates of the standard deviation of the ISED. Then I contrasted it with the V-Dem item v2elsnlsff, which asks country experts to evaluate whether the freeness and fairness of elections vary across different areas of a given country.Footnote47 Just like before, the solid line is the country-year estimate of the ISED (in this case the standard deviation), and the dashed line corresponds to the V-Dem values.
shows that – albeit weakly – for Argentina, Brazil, and India, the ISED coincides with the V-Dem score in signalling whether democracy within countries becomes more or less uneven through time. This is also true for Canada. For Mexico, Peru, Bolivia, and Venezuela, however, there is another interesting difference. On the one hand, the ISED points towards an increase in the territorial unevenness of electoral democracy. On the other hand, the V-Dem scores signal a tendency towards liberal democratic homogeneity. Here the differences between the liberal and the electoral dimensions could again be at the root of this discrepancy. In other words, while the liberal aspects of democracy – such as the rule of law and civil liberties – might in fact be increasingly homogeneous inside countries, the electoral dynamics across subnational units – that is, the patterns of electoral competition and participation in individual states – continue to diverge.
Figure 4. The ISED and V-Dem items: capturing the heterogeneity of subnational democracy.
Perhaps most surprisingly, for units in Colombia, Uruguay, and the United States, the V-Dem score registered no unevenness for the past forty years. In contrast, the yearly standard deviation of the ISED captures, for the first time, the territorial heterogeneity of democracy along these three countries. This previously untapped variation – which in the USA clearly reflects Rocco'sFootnote48 point regarding enduring unevenness – provides ample avenues for future research.
Having discussed the conceptualization and operationalization of the ISED, and having assessed its strengths, weaknesses, and its validity, in the next section of the article I further my assessment of Argentinian, Mexican, Brazilian, and Indian units to substantively contribute to the debate, while illustrating the different ways in which future scholars can use the index to explore and explain the heterogeneity of regimes within countries.
Different and enduring democratic trajectories across subnational units
The ISED dataset allows for a thorough descriptive assessment of the spatial and temporal patterns of subnational democracy across the Americas and India. For instance, the graphs on the left-hand side of track the country-year average of subnational democracy in colour and highlight in black the trajectories followed by individual provinces for the past decades. The top-left graph displays the case of Argentina, mapping Santa Fe as a well-known overperformer, and Santiago del Estero as a province with strong democratic deficits. The following graphs do the same for Brazil, Mexico, and India. For the latter case, in line with previous findings presented in DemocratizationFootnote49, the ISED maps Kerala as a democratic state, and shows that Arunachal Pradesh has remained below the national average for the past four decades. For Brazil and Mexico, Jalisco and Rio Grande do Sul receive systematically higher scores, while Guerrero and Bahia have scored significantly lower values for the past four decades.
Figure 5. Different and enduring democratic trajectories across subnational units.
In Venezuela (shown in the online appendix), democracy has collapsed both at the national and the subnational level. Most if not all the provinces have some of the lowest scores of the overall sample, and, as Polga-HecimovichFootnote50 suggests, they have also continued to experience a significant downward trend in recent decades. Surprisingly, however, the figure in the online appendix also reveals that there are states such as Barinas – where the officialist PSUV recently faced electoral defeatFootnote51 – that have managed to remain somewhat competitive. With the re-democratization of Venezuela still on the balance, exploring subnational variation in the Southern Bolivarian Republic might offer hints as to where new democratic challengers may arise.
The fact that for the past four decades states like Arunachal Pradesh, Santiago del Estero, Guerrero, and Bahia have systematically scored low on the ISED scale, while Jalisco, Rio Grande do Sul, Santa Fe, and Kerala continuously attained high values, points to the salience of different and enduring democratic trajectories across subnational units. To explore this further, the graphs on the right-hand side of plot past and present ISED mean values. That is, I split the sample in half, with values on the X-axis corresponding to the ISED mean scores for years between 1980 and 2000.Footnote52 The Y-axis then corresponds to the mean ISED value for scores between the next two decades (i.e. 2001–2021). What these graphs reveal is that across these four country-cases – and in fact for most of the countries in the sample – the relative position of states and provinces has been remarkably stable over time.Footnote53
That this equilibrium has survived for roughly four decades casts doubt on the explanatory power of accounts emphasizing subnational rentierismFootnote54, as well as those theories highlighting circadian boundary managing strategies.Footnote55 These theories rely on the constant conjunction of cause and effects. Nonetheless, the evidence presented here suggests that future assessments would do well to consider the intricacies of time and history.Footnote56
Beyond tracking the trajectories of individual provinces, the ISED dataset also allows scholars to look at how the two components of the index have varied within each subnational unit through time. In , I break down the patterns of subnational competition and subnational participation for the over and under-performing cases identified earlier.Footnote57 The plots confirm that, over the last decades, subnational units across Argentina, Brazil, Mexico, and India have been “low-intensity” democraciesFootnote58, since levels of competition have been systematically higher than the levels of participation. Moreover, the findings of HuntingtonFootnote59, Boix, Miller, and RosatoFootnote60, and MillerFootnote61 suggest a post-World War II shift, one in which participation overcomes competition. That is, the general trend of contemporary national democracies is for participation scores to be higher than those measuring competition. The analysis reveals that this has not been the case for the majority of the subnational units in the whole sample.Footnote62
Figure 6. Tracking competition and participation within subnational units across Argentina, Brazil, Mexico, and India.
Subnational regime hybridity: the norm rather than the exception
The ISED allows scholars to classify subnational regimes. Following MillerFootnote63, in I plot the competition and participation scores for the subnational units of Argentina, Brazil, Mexico, and India.Footnote64 The graph is subdivided into six zones, with the dashed lines being one standard deviation above or below the pooled mean of the corresponding component. Each zone then identifies a distinct regime type. The interpretation is intuitive: democracies, located in the first zone, have high levels of both competition and participation. The second zone shows competitive oligarchies which have similar levels of competition but restrict participation. Electoral authoritarian regimes, displayed in the third zone, show mid-levels of competition and higher degrees of participation. Electoral oligarchies restrict participation even further (fourth zone). Cases of low competition but high participation are classified as plebiscitary authoritarianisms (fifth zone), and cases where both dimensions receive low scores are classified as closed or fully authoritarian regimes, which are located on the bottom left (sixth zone). Importantly, under this framework, cases outside zones one and six are considered hybrid regimes.
Figure 7. Subnational regime hybridity across Argentina, Brazil, Mexico, and India.
summarizes this information, showing that for the past four decades, rather than the exception, hybridity has been then norm at the subnational level. In fact, out of the 8,787 province-year spells of the full sample, over 85% of them can be classified as hybrids. For the specific cases of Argentina, Brazil, Mexico, and India, both and show that, on average, Indian provinces outperform their Latin American counterparts, and that 61.22% of the available observations fall within the range of electoral authoritarianisms, while 10.12% do so in the range of electoral oligarchies, with roughly 12.9% of cases classified as full democracies, and only 4.9% as closed authoritarianisms. This classification is of course subject to the established thresholds, which in this case are set using the average value and the standard deviation of the two components of the index. Future researchers can set different thresholds and try to classify subnational regimes using a completely different typology, or on the sole basis of the index, rather than its components.
Table 4. Subnational regime classification Argentina, Brazil, and Mexico.
Democracy and development at the subnational level
Researchers interested in exploring the drivers of subnational democracy can use the ISED data to explore classic hypotheses of the regime change literature. The graphs in look at the association between subnational democracy and economic development (measured as the natural logarithm of the provincial Gross Domestic Product per capitaFootnote65) across our four country-cases. This relationshipFootnote66 has been considered a long-standing regularity at the national levelFootnote67, which – based on the graphs shown in – seems to travel down to the subnational level, at least for Latin America and India.
Figure 8. Democracy and development at the subnational level.
Although the correlation for the Mexican case is positive and fails to achieve statistical significance, this preliminary assessment shows a link between economic development and democracy at the subnational level in Argentina, Brazil, and India, with a positive and significant Pearson correlation coefficients ranging between 0.19 and 0.68. A preliminary analysis of the USA reveals that its overall relationship between subnational democracy and development is positive, and significant but rather weak.Footnote68 However, when breaking down this association between “Deep South” and “Other” statesFootnote69, the former shows a marginally stronger and significant correlation between state-level democracy and state-level development, with a coefficient of 0.23. Looking at all other states, the relationship is significant and negative, although only slightly smaller than zero.
Canadian units similarly display a small, negative but significant association between electoral democracy and economic development. These graphs beg future exploration of both the robustness and causal nature of this relationship, as well as the mechanisms behind it. For now, to put it provocatively, this evidence would suggest that across India and the largest federations in Latin America – à la Lipset – the more well-to-do a subnational unit, the more likely it is that it sustains a democratic regime.
Conclusion
The ISED is a valid instrument that captures the electoral dimension of democracy at the subnational level. It has a positive and significant correlation with similar existing measures and shows interesting associations with the relevant V-Dem items which are worth exploring further. By relying on publicly available electoral data, the ISED can be readily updated. As I have argued throughout this article, its validity and cost-effectiveness make it a valuable asset for current and future national and subnational research.
In , I summarize the distinct potential research applications of the ISED. As discussed earlier, the index captures both the level and the heterogeneity of electoral democracy at the provincial level and allows calculating aggregate, country-level estimates. However, the ISED is most useful to explore the longitudinal, cross-sectional, and cross-national variation of subnational democracy. Importantly, the ISED data track the distinct temporal and spatial unfolding of not only the index, but also of its individual components (competition and participation). Lastly, it provides the basis for classification, hypothesis testing, and it can also be used as a control variable for studies looking into different outcomes at various territorial scales.
Table 5. The road taken.
Nonetheless, it is important to underscore that the ISED captures the electoral dimension of democracy only. Put differently, it's necessary to stress that it does not intend to tap into the liberal, or more substantive aspects of subnational regimes. In addition, while this first iteration covers a significant number of country cases, and offers observations of over three hundred subnational units, is still limited in its scope. As mentioned previously, future, collaborative research should aim to extend the data collection efforts to other countries and regions across the globe.
Subsequent projects can use the ISED to formulate new questions and grapple with the substantive conclusions of this article, summarized here in . For example, future scholars can look at the institutional factors which might explain why Indian states democratically outperform Argentinian, Brazilian, and Mexican provinces. Similarly, future inquiries could explore why the relative position of subnational units vis-à-vis their level of democracy has been considerably stable over time. While current theories have been focused on the endurance of subnational regimesFootnote70, we currently do not have a theory fully explaining the origins of this lasting trajectories. Finally, there is a promising avenue to further explore the classic modernization puzzle at the subnational level, as well as to examine the top-down or bottom-up dynamics that account for the trends of competition and participation observed.
Table 6. Conclusions on the state of democracy in Argentinian, Brazilian, Mexican and Indian Provinces.
Acknowledgements
The author thanks Timothy J. Power, David Doyle, Ezequiel González Ocantos, Raquel Chanto, Elle Pfeffer, Daniel Barker, Wil Buitrago, Marina Erikson, Gonzalo Contreras, as well as Miguel, Javier, and Lisbeth for their help and their continued support.
The author would also like to thank the two anonymous reviewers for their helpful comments on the manuscript.
Disclosure statement
No potential conflict of interest was reported by the author(s).
Data availability statement
Replication data is available at https://doi.org/10.7910/DVN/KO9C2M.
Additional information
Funding
Notes on contributors
Javier Pérez Sandoval
Javier Pérez Sandoval is a Departmental Lecturer at the School of Global and Area Studies (OSGA) at the University of Oxford. His research focuses on the political economy of subnational regimes across Latin America.
Notes
1 Skaaning, “Different Types of Data.”
2 Giraudy, Moncada, and Snyder, Inside Countries.
3 Giraudy, Democrats and Autocrats; Gervasoni, Hybrid Regimes within Democracies.
4 McMann, “Measuring Subnational Democracy.”
5 I use the term subnational to refer to the level or tier immediately below the federal or central government. This pertains, for example, to Argentinian provincias, Brazilian unidades federativas and Mexican entidades. Throughout the paper I use subnational, local and provincial interchangeably. In a similar vein, to refer to countries as a whole, I use nation and country-level. Lastly, when necessary, I capitalize State to distinguish the abstract-Weberian notion from state, which in this text alludes specifically to the subnational units.
6 Harbers, Bartman, and van Wingerden, “Conceptualizing and Measuring Subnational Democracy.”
7 Munck and Verkuilen, “Conceptualizing and Measuring Democracy.”
8 Dahl, Polyarchy: Participation and Opposition.
9 Coppedge, Álvarez, and Maldonado, “Two Persistent Dimensions of Democracy.”
10 Vanhanen, “A New Dataset for Measuring Democracy: 1810-1998.”
11 See for example Giraudy, Democrats and Autocrats, and Bohlken, Democratization from Above.
12 Lipset, “Some Social Requisites of Democracy.”
13 Gervasoni, “Subnational Democracy in (Cross-National) Comparative Perspective.”
14 Sellers, “From Within to Between Nations.”
15 Gallie, “Essentially Contested Concepts,” 169.
16 Abellán, Democracia; Held, Models of Democracy.
17 Dahl, Polyarchy: Participation and Opposition, 5–6.
18 Dahl, 4.
19 Coppedge, Álvarez, and Maldonado, “Two Persistent Dimensions of Democracy.”
20 Vanhanen, “A New Dataset for Measuring Democracy: 1810–1998,” 253.
21 Munck, Measuring Democracy: A Bridge between Scholarship and Politics.
22 Coppedge, Álvarez, and Maldonado, “Two Persistent Dimensions of Democracy.”
23 Almond and Verba, The Civic Culture.
24 Jämte and Ellefsen, “The Consequences of Soft Repression.” Thanks to the anonymous reviewers for highlighting that, in addition to politically motivated soft repression, there is also research on how violence negatively impacts democracy at the local level by diminishing turnout, see for example Trelles and Carrera, Bullets and Votes.
25 National-level events could also influence non-concurrent subnational elections. Research on second-order electoral dynamics (see for example Schmitt and Teperoglou, Second-Order Elections and Electoral Democracy) has identified that voters sometimes punish or reward local incumbents for the performance of their national-level partisans. Nonetheless, to address this concern scholars could follow a similar approach to the one suggested previously, identifying the year in which a national scandal or a performance blunder occurs and controlling for it or adjusting the estimation accordingly. I thank the anonymous reviewers for highlighting this other potential source of influence and for suggesting a possible way around it.
26 Majoritarian electoral systems maximize the share of votes received by the winning party. As a consequence, in countries with this system the competition component is systematically lower. Given the multiplicative aggregation mechanism, when Vanhanen's framework is used to measure majoritarian democracies, the index is naturally biased downwards.
27 The salience of executives in presidential systems in the Americas, and in Latin America specially, has been highlighted by classic work such as Cox and Morgenstern's “Latin America's Reactive Assemblies and Proactive Presidents”. Canada and India are the only two countries with a parliamentary system in the sample. For these cases, given the dependence of the executive on the legislative branch, I focused on the elections of local legislatures.
28 See for example, Geddes, Wright and Frantz, Autocratic Breakdown and Regime Transitions, as well as Behrend and Whitehead, Illiberal Practices.
29 See for example Bruhn, Too Much Democracy?
30 Falleti, Decentralization and Subnational Politics; Montero and Samuels, Decentralization and Democracy in Latin America; Baicker, Clemens, and Singhal, “The Rise of the States”; Jayal, Prakash, and Sharma, Local Governance in India.
31 By this I mean subnational units where local executives and local legislatures are directly elected by citizens, and where the interaction of parties, citizens and other political organizations conditions the access and the exercise of political power within the own territorial and legal boundaries of second-level units.
32 By using multiplication as an aggregation mechanism, the index ensures that both participation and competition are equally weighed. This implies, for example, that high levels in one component do not make up for low values in the other.
33 Tow, Andy Tow's Electoral Atlas.
34 It's important to note that Nunavut and the Northwest Territories in Canada are subnational units where elections are conducted without political parties. As such, I was only able to estimate the electoral participation component. Consequently, for these units the ISED is based solely on this one component.
35 In Argentina the INDEC is the National Institute of Statistics and Censuses. IBGE is the Brazilian Institute of Geography and Statistics, and INEGI is Mexico's National Institute of Statistics and Geography.
36 A balanced panel is one in which all observations have the same time frame. The unbalanced nature of the data stems from the fact that subnational units have distinct electoral cycles. A short panel is one in which the cross-sectional (n) dimension – i.e. subnational units – is larger than the temporal one (t). My dataset has n = 237 and t = 42. In a long panel, n<t.
37 Fidalgo, “How Democratic Is Government Really?”
38 Beer and Mitchell, “Democracy and Human Rights.”
39 Montero, “Uneven Democracy?”
40 Gervasoni, “A Rentier Theory of Subnational Regimes.”
41 Grumbach, Laboratories against Democracy. The author would like to thank Grumbach, Fidalgo, Beer, Montero, and Gervasoni for kindly sharing their data.
42 Levitsky and Ziblatt, How Democracies Die.
43 Rocco, “Laboratories of What?,” 299.
44 Mickey, “Challenges to Subnational Democracy.”
45 Coppedge et al., “V-Dem [Country–Year/Country–Date] Dataset V12.”
46 Specifically, the V-Dem question reads: “Taking all aspects of the pre-election period, election day, and the post-election process into account, would you consider subnational elections (regional and local, as previously identified) to be free and fair on average?” (V-Dem 12)
47 For this item, the V-Dem questions reads: “Does the freeness and fairness of subnational elections vary across different areas of the country?” (V-Dem 12). It is important to note that V-Dem codes this variable in terms of homogeneity, that is, higher values indicate that elections are equally free and fair. To make it comparable to the ISED standard deviation estimate, I need to express it in terms of heterogeneity, which is why I estimate the inverse of the v2elsnlsff V-Dem item, formally 1/ v2elsnlsff.
48 Rocco, “Laboratories of What?.”
49 Harbers, Bartman, and van Wingerden, “Conceptualizing and Measuring Subnational Democracy.”
50 Polga-Hecimovich, “Venezuelan Regional Elections.”
51 In the elections of early 2022 Sergio Ramón Garrido beat Nicolas Maduro's United Socialist Party of Venezuela (PSUV) under the banner of “Plataforma Unitaria” (Unity Platform), a political alliance among the different parties and civil organizations opposing the regime.
52 Given that I split each country's sample in half, the starting and ending points are different in each case.
53 The only exception to this is Ecuador where the coefficient is indeed positive but fails to achieve statistical significance at conventional levels. This, however, is likely driven by a reduced number of observations.
54 Gervasoni, Hybrid Regimes within Democracies.
55 Gibson, Boundary Control.
56 Pierson, Politics in Time.
57 Here I present smoothed trend graphs based on the linearly interpolated data. Using a three-year average or graphing only the observed data for electoral years (like I did in ) does not change the substantive conclusion regarding “low intensity” regimes. In the online material I further break down the participation and competition components for the full set of subnational units for all the country-cases available in the dataset.
58 O’Donnell, “Delegative Democracy.”
59 Huntington, The Third Wave.
60 Boix, Miller, and Rosato, “A Complete Data Set of Political Regimes, 1800–2007.”
61 Miller, “Democratic Pieces.”
62 Exception for this “low intensity” pattern – that is, cases in which participation scores indeed surpass competition ones – include all the provinces of Uruguay, Formosa and San Juan in Argentina, Cochabamba and Oruro in Bolivia, Carchi, Bolivar, and Imbabura in Ecuador, Montana, and North Dakota in the USA, along with Mexico City, Zacatecas, and Yucatán in Mexico.
63 Miller, “Democratic Pieces.”
64 The online material reproduces this exercise for each individual country-case in the dataset, reporting a classification the uses the country-specific mean – rather than the pooled or combined one – for both the competition and participation component.
65 Local GDP estimates for US states were obtained from Bureau of Economic Analysis (BEA). GDP estimates for India were obtained from the Reserve Bank of India (RBI). For Mexico and Argentina, I use estimates from the National Statistics Institutes (INEGI and INDEC). Brazilian GDP estimates were obtained from IPEA Data. To calculate the GDP per capita, subnational GDPs were first obtained as shares of the national one. Then, I multiplied these shares by World Bank GDP purchasing power (PPP) estimates. Finally, those results were divided by the total population of each state to obtain the per capita figures.
66 Lipset, “Some Social Requisites of Democracy.”
67 Geddes, “What Do We Know About Democratization after Twenty Years?”
68 For the US, the overall or pooled Pearson correlation coefficient is positive and significant and has a value of 0.11.
69 US states included in the “Deep South” category are: Alabama, Arkansas, Florida, Georgia, Kentucky, Louisiana, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia, and West Virginia.
70 See the PhD theses by Griffis “Democratic Blind Spots” and Mera “National Democracies, Local Autocracies”.
Bibliography
- Abellán, Joaquín. Democracia. Madrid: Alianza, 2011.
- Almond, Gabriel A, and Sidney Verba. The Civic Culture: Political Attitudes and Democracy in Five Nations. Newbury Park, CA: Sage, 1963.
- Baicker, Katherine, Jeffrey Clemens, and Monica Singhal. “The Rise of the States: U.S. Fiscal Decentralization in the Postwar Period.” Public Economics 96 (2011): 1–13.
- Beer, Caroline, and Neil J. Mitchell. “Democracy and Human Rights in the Mexican States: Elections or Social Capital?” International Studies Quarterly 48, no. 2 (2004): 293–312.
- Boix, Carles, Michael Miller, and Sebastian Rosato. “A Complete Data Set of Political Regimes, 1800–2007.” Comparative Political Studies 46, no. 12 (2013): 1523–54.
- Coppedge, Michael, Angel Álvarez, and Claudia Maldonado. “Two Persistent Dimensions of Democracy: Contestation and Inclusiveness.” The Journal of Politics 70, no. 3 (2008): 632–47.
- Coppedge, Michael, John Gerring, Carl Henrik Knutsen, Staffan I. Lindberg, Jan Teorell, Nazifa Alizada, David Altman, et al. “VDem [Country–Year/Country–Date] Dataset V12,” n.d.
- Cox, Gary, and Scott Morgenstern. “Latin America’s Reactive Assemblies and Proactive Presidents.” Comparative Politics 33, no. 2 (2001): 171–89.
- Dahl, Robert. Polyarchy: Participation and Opposition. New Haven: Yale University Press, 1971.
- Falleti, Tulia G. Decentralization and Subnational Politics in Latin America. Cambridge: Cambridge University Press, 2010.
- Fidalgo, Amanda. “How Democratic Is Government Really? The Impact of Subnational Regime Variation on Evaluations of Democracy.” Governance (2021): 1–18.
- Gallie, W. B. “Esssentially Contested Concepts.” Proceedings of the Aristotelian Society 1955-1956 56 35 (1956): 167–98.
- Geddes, Barbara. “What Do Know About Democratization After Twenty Years?” Annual Review of Political Science 2 (1999): 115–44.
- Gervasoni, Carlos. “A Rentier Theory of Subnational Regimes: Fiscal Federalism,: Democracy, and Authoritarianism in the Argentine Provinces.” World Politics 62, no. 2 (2010): 302–40.
- Gervasoni, Carlos. Hybrid Regimes within Democracies. Cambridge: Cambridge University Press, 2018.
- Gervasoni, Carlos. “Subnational Democracy in (Cross-National) Comparative Perspective: Objective Measures with Application to Argentina, Brazil, Canada, Mexico, Uruguay and the United States.” APSA 2012 Annual Meeting Paper (2012): 1–33.
- Giraudy, Agustina. Democrats and Autocrats. Oxford: Oxford University Press, 2015.
- Giraudy, Agustina, Eduardo Moncada, and Richard Snyder. Inside Countries. Cambridge: Cambridge University Press, 2019.
- Griffis, John G. “Democratic Blind Spots: Organized Labor and the Persistence of Subnational Authoritarianism in Mexico.” PhD diss., Florida State University, 2019.
- Grumbach, Jacob. Laboratories Against Democracy: How National Parties Transformed State Politics. Princeton, NJ: Princeton University Press, n.d.
- Harbers, Imke, Jos Bartman, and Enrike van Wingerden. “Conceptualizing and Measuring Subnational Democracy Across Indian States.” Democratization 26, no. 7 (2019): 1154–75.
- Held, David. Models of Democracy. 3rd ed. Cambridge: Polity, 2006.
- Huntington, Samuel P. The Third Wave: Democratization in the Late 20th Century. Norman, OK: University of Oklahoma Press, 1991.
- Jämte, Jan, and Rune Ellefsen. “The Consequences of Soft Repression.” Mobilization:An International Quarterly 25, no. 3 (2020): 383–404.
- Jayal, N. G., Amit Prakash, and P. K. Sharma. Local Governance in India: Decentralization and Beyond. Oxford: Oxford University Press, 2006.
- Levitsky, Steven, and D. Ziblatt. How Democracies Die: What History Reveals About Our Future. London: Penguin Random House, 2018.
- Lipset, Seymour M. “Some Social Requisites of Democracy: Economic Development and Political Legitimacy.” American Political Science Review 53, no. 1 (1959): 69–105.
- McMann, Kelly M. “Measuring Subnational Democracy: Toward Improved Regime Typologies and Theories of Regime Change.” Democratization 25, no. 1 (2018): 19–37.
- Mera, Manuel. “National Democracies, Local Autocracies: The Uneven Democratization of Subnational Governments in Argentina and Brazil”. PhD diss., Georgetown University, 2016.
- Mickey, Robert. “Challenges to Subnational Democracy in the United States,: Past and Present.” ANNALS of the American Academy of Political and Social Science 699 (2022): 118–129.
- Miller, Michael. “Democratic Pieces: Autocratic Elections and Democratic Development Since 1815.” British Journal of Political Science 45, no. 3 (2015): 501–30.
- Montero, Alfred. “Uneven Democracy? Subnational Authoritarianism in Democratic Brazil.” In Latin American Studies Association Conference. Montreal, Canada, 2007.
- Montero, Alfred, and David Samuels. Decentralization and Democracy in Latin America. Notre Dame, IN: University of Notre Dame Press, 2004.
- Munck, Gerardo L. Measuring Democracy: A Bridge Between Scholarship and Politics. Baltimore: John Hopkins University Press, 2009.
- Munck, Gerardo L., and Jay Verkuilen. “Conceptualizing and Measuring Democracy:Evaluating Alternative Indices.” Comparative Political Studies 35, no. 1 (2002): 5–34.
- O’Donnell, Guillermo. “Delegative Democracy.” Journal of Democracy 5, no. 1 (1994): 55–69.
- Polga-Hecimovich, John. “Venezuelan Regional Elections,: Democratic Backsliding, and Autocratization in the 2010s.” Regional and Federal Studies 32 (2022): 1–22.
- Rocco, Philip. “Laboratories of What? American Federalism and the Politics of Democratic Subversion.” In Democratic Resilience, edited by Robert C. Lieberman, Suzanne Mettler, and Kenneth M. Roberts, 297–319. Cambridge: Cambridge University Press, 2021.
- Sellers, Jefferey M. “From Within to Between Nations: Subnational Comparison Across Borders.” Perspectives on Politics 17, no. 1 (2019): 85–105.
- Skaaning, S.-E. “Different Types of Data and the Validity of Democracy Measures.” Politics and Governance 6, no. 1 (2018): 105–16.
- Tow, Andy. Andy Tow’s Electoral Atlas. Available Online., n.d.
- Vanhanen, Tatu. “A New Dataset for Measuring Democracy: 1810–1998.” Journal of Peace Research 37, no. 2 (2000): 251–65.