State of the world 2023: democracy winning and losing at the ballot

ABSTRACT

This article presents trends in democracy, autocracy, and regime transformation using the Varieties of Democracy (V-Dem) dataset version 14. We offer five main findings. First, the average level of democracy as measured by the Liberal Democracy Index (LDI) continues to decline and has reached the 1985-level when looking at the population-weighted averages. Second, 42 countries are in ongoing episodes of autocratization, and 18 countries are in ongoing episodes of democratization in 2023. Third, we are the first to show that out of the 42 autocratizing countries, 19 are “Bell-turn” autocratizers, while nine out of the 18 democratizers are “U-turns.” Fourth, we unpack the familiar analysis of the three waves of autocratization and democratization by dividing these into the newly discovered stand-alone episodes of autocratization versus Bell-turn episodes, as well as stand-alone episodes of democratization versus U-turn episodes. This analysis suggests that regime volatility has increased in recent years, as a larger share of both autocratization and democratization are made up of Bell-turns and U-turns, respectively. Fifth, out of the 60 countries holding national elections this year, 31 countries are in some degree of decline in democracy, while only three are improving on democracy.

KEYWORDS:

• Democracy
• autocracy
• autocratization
• democratic resistance
• reversed democratization
• u-turn
• bell-turn
• waves of democratization
• waves of autocratization

Introduction

This article presents the trends in democracy*, autocracy, and regime transformation using the Varieties of Democracy (V-Dem) dataset version 14.¹ Democracy continues to decline in the world, and autocratization continues to be the dominant trend. Instances of democratization remain relatively few and mostly happen in small states. Yet, we are finding new complexities in these trends. Extending on the new Episodes of Regime Transformation (ERT) methodology,² we show that many autocratizers are Bell-shaped turnarounds – instances of failed democratization turned into autocratization – and many democratizers are U-turns, that is, “bright spots" where autocratization was halted and reversed.

There are five main findings. First, the average level of democracy as measured by the Liberal Democracy Index (LDI) continues to decline. However, the extent varies substantially depending on the perspective. Going by country averages where equal weight is given to any territory with a government, the level of democracy is back to levels last seen in 1998 and the point estimates are still within the confidence intervals.³ But democracy is rule by the people, and the number of people that enjoy democratic rights and freedom is arguably an important metric for how much democracy there is in the world. Thus, by population-weighted averages, the decline is much steeper – the point estimate for 2023 is equivalent to the average last seen in 1985 – and well outside the confidence intervals. Democracy levels weighted by territorial size show a similar decline. Meanwhile, the decline is the steepest when levels of democracy are weighted by the share of world gross domestic product (GDP). By that metric, the world average of democracy seen in 1973 was passed in 2019 and the downward trend continues, reflecting that the balance of economic power in the world is shifting towards non-democratic states.

Second, using the new ERT methodology, we identify 42 countries that are in ongoing episodes of autocratization, and 18 countries that are in ongoing episodes of democratization in 2023. For the past 15 years, the share of the world’s population living in autocratizing countries has overshadowed the share living in democratizing ones. In 2023, 35% of the world’s population reside in autocratizing countries – about half of which are found in India – and 5% of the world’s population (400 million) live in democratizing countries, with more than half of them accounted for by Brazil.

Third, out of the 42 autocratizing countries, 19 are Bell-turns – this demonstrates, among other things, how fragile many democratization processes are. The first-ever identification of Bell-turn episodes, or reversed democratization, presents our field with new research opportunities. For example, one could use comparative case studies and design-based causal inference approaches to answer new questions such as why some episodes of democratization succeed while others result in Bell-turns. Similarly, 9 of the 18 presently ongoing democratization episodes are U-turns. These are countries that successfully stopped and to some degree reversed autocratization. Again, the identification of this new type of regime transformation opens up new avenues for research. For example, what are the factors that distinguish U-turns from autocratization episodes that are not reversed? Identifying those would be of keen interest to the international democracy promotion community.

Fourth, we unpack the familiar analysis of the three waves of autocratization and democratization by dividing them into the newly discovered episodes of stand-alone autocratization versus Bell-turns, and episodes of stand-alone democratization versus U-turns, respectively. Bell-turns are episodes in which democratization is closely followed by and inherently linked to subsequent autocratization. Inversely, U-turns are episodes of change where autocratization is closely followed by and inherently linked to subsequent democratization. By contrast, stand-alone episodes of change are episodes that start after a period of relative stability in democracy levels. This approach offers a fresh perspective that puts well-known patterns in a somewhat different light. On the one hand and despite the sharp increase in the number of countries undergoing autocratization from around 2000, Bell-turns continue to make up on average about 45% of all autocratization episodes. Many democratization attempts still fail. On the other hand, U-turns are also becoming relatively more common suggesting that attempts at autocratization also fail and do so at an increasing rate. U-turns accounted for 18% of all democratization episodes in the 2000s, 34% in 2010s, and 46% in the first few years of the 2020s. We interpret these two trends as suggestive evidence of increasing regime volatility and contestation.

Finally, we show that out of the 60 countries holding national elections this year, 31 countries are witnessing to some degree a decline in democracy, while only three are democratically improving. The literature indicates that elections can be critical both for turning an autocratization process on its head and for legitimizing further autocratization. Twenty twenty-four thus looks to be a consequential year for democracy, not least with upcoming elections in recently or currently democratically declining countries, including India, Mexico, and the United States.

This article contributes to an on-going scholarly debate about the state of democracy in the world. Many observers have argued that democracy is in decline or even crisis.⁴ Some recent contributions argue, in contrast, that these claims are exaggerated or even alarmist.⁵ While we don’t elaborate here on the different approaches to measuring democracy (see e.g. Knutsen et al. for a recent discussion⁶), we want to emphasize that our approach is to engage with a series of different lenses through which the same underlying data is analysed. First, we discuss developments based on the Liberal Democracy Index, its sub-indices, and the constituent indicators thus leveraging analyses from the very aggregate level down to the fine-grained. This allows us to speak both of general and specific trends, as well as where the broader changes in democracy originate from (such as, media freedom, elections, executive constraints, or the like). Second, we also discuss the state of the world from the perspective of regime types, using the Regimes of the World⁷ measure, both in terms of number of countries and shares of the world population. Third, we analyse the number and extent to which countries are in episodes of change, either improving or declining in terms of democracy levels, thus giving account of countries democratizing and autocratizing across the regime spectrum today as well as historically. Finally, ours is the first attempt to break down autocratization and democratization into two types – stand-alone autocratization and Bell-turns, and stand-alone democratization and U-turns, respectively, to show how the world is changing democratically. We believe that this multifaceted approach is informative for assessing the state of democracy in the world and will further the debate on how democracy is developing or not developing.

Liberal democracy in the world, 1973–2023

How does one best measure the state of democracy in the world? Figure 1 shows the development of the Liberal Democracy Index (LDI) over the last 50 years, using four different metrics. The conventional approach is using country-averages (Panel A, Figure 1), where each country’s level of democracy is weighted equally. From the perspective that democracy exists at the country level, it is reasonable to suggest that the aggregate state of democracy is the average of each country’s level of democracy. Viewed from this perspective, democracy is in a downward trend – 2023 has gone back to a level last registered in 1998 – yet this decline is within the confidence intervals. The country-average metric, however, could be accused of being misleading in that it does not reflect the level of democracy enjoyed by the average person in the world. “How much democracy” there is in the world is arguably not equally influenced by a country like The Seychelles, with some 90,000 inhabitants, as it is by a country like India, with 1.4 billion people. An increase of democracy in The Seychelles by say 50% does not really compensate for a decline in India by the same percentage in terms of how democratic the world is.

Figure 1. Liberal Democracy Index by Country Averages, Population-, Territory-, and GDP-Weights, 1973–2023.

Note: The black lines represent global averages on the LDI with the grey area marking the confidence intervals. Panel A is based on conventional country averages. Panels B, C, and D show levels of democracy weighted by population, territory, and GDP, respectively. The data for the latter three figures are drawn from the World Bank and Fariss et al. (2022), both included in the v14 of the V-Dem dataset.

An alternative approach, which has been featured for some years now in V-Dem Institute’s annual Democracy Reports, is to look at population-weighted averages (Panel B, Figure 1). The rationale here is that the world is only as democratic as the number of people who enjoy democratic rights and freedoms. If democracy is rule by the people, we should let the level of democracy experienced by people be the determining factor for speaking about how democratic the world is. We believe that this approach holds merit, but an immediate consequence is of course that changes in a very populous country a have substantial impact on the world average. Yet, countries with large populations arguably also mean more for “how much democracy” there is in the world. Hence, substantial declines in democracy levels in several populous countries make up a substantial share of the decline observed in Panel B, Figure 1. By population-weighted averages, the level of democracy in 2023 is back to levels last seen in 1985. Importantly, this does not mean that the world as a whole, or any country for that matter, is back to 1985. All countries are in different situations today than they were then. Some countries are much better in terms of democracy, while others are worse off. But this comparison puts things into perspective of how far the decline has gone, in terms of numbers of people affected by democratic decline.

A third approach is looking at territory-weighted averages (Panel C, Figure 1). This approach is far less conventional but builds on the notion that one can speak of the state of democracy in the world in terms of how much of the territory geographically is governed democratically. Somewhat more akin to the country-averages in that it is concerned with borders of countries. The data here show a starker decline on the LDI during the last decade compared to the country-averages. By territory-weighted averages, the level of democracy is now back to the levels last seen in 1987 ­– very similar to the population-weighted metric.

A fourth metric is GDP-weighted averages (Panel D, Figure 1), which captures how democratic the world is in terms of economic power. Using this approach, the level of democracy is weighted based on the economic size of the country, as measured by its share of global GDP. The higher the GDP-weighted democracy is, the more economic power rests in more democratic states, and vice versa. The GDP-weighted measure stands out among the four metrics used in Figure 1. It has declined substantially since around the millennium and is now down to levels below even that of 1973. By this metric, the world is now worse off than before the third wave of democratization started with Portugal’s Carnation Revolution.

Why are there such stark differences between the country-based and other weighted averages? Put simply, numerous large countries with big populations and large economies such as India, Mexico, The Philippines, Pakistan, Poland, Russia, Türkiye, and the United States, have declined significantly on democracy over the past decades. Countries that are advancing are typically small countries with smaller populations such as Fiji, The Gambia, Kosovo, Lesotho, The Seychelles, Timor-Leste, and recently Zambia. Brazil joined the advancers group this year that led to a slight change in the overall picture. It is also important to recognize that the patterns vary across regions. For example, going by population-weighted averages, democracy increased in Latin America and the Caribbean in the last years. On the other hand, the decline has been especially steep in Eastern Europe, and South and Central Asia (see Figure 2A, Online Appendix).

Regimes of the world, 1973–2023

By the best estimate using the Regimes of the World (RoW) measure,⁸ the world is almost evenly divided between 91 democracies (liberal and electoral) and 88 autocracies (electoral and closed) at the end of 2023. Figure 2 shows the most comprehensive overview of each country’s trajectory on RoW measure during the last fifty years. This division does not seem to have changed markedly in the past 15 years or so.

Figure 2. History of Regimes of the World (RoW) by Country-Year, 1973–2023.

Note: Each tile corresponds to one country-year. We use the PanelView R package developed by Mou, Liu, and Xu (2022) to visualize the history of Regimes of the World (RoW).

Figure 3 (left panel) also provides the exact count by regime category over the past 50 years using the RoW measure. In recent times, two trends are visible. On the one hand, the number of liberal democracies in the world has declined since around 2010, with many turning into electoral democracies. At the same time, the number of electoral autocracies has declined, with a number of those turning into closed autocracies. This shows the two sides of the third wave of autocratization: on the one hand, democratic backsliding and, on the other hand, autocracies becoming more repressive.

Figure 3. Regime Types by Number of Countries, 1973–2023.

Note: Figure 3 plots the number of countries (left panel) by regime type. The right panel takes into account measurement uncertainty, highlighting the number of “grey zone” countries above or below the line, distinguishing democracies and autocracies. The “grey zone” estimates are based on the version of the RoW indicator that includes uncertainty (v2x_regime_amb in the V-Dem dataset).

Naturally, categorical classifications like the RoW measure sacrifice some nuances and risk misclassifying countries when the underlying data puts them at, or very close to, thresholds. Figure 3 (right panel) shows that there are 18 countries that could be misclassified due to such measurement uncertainty as they are either in the “lower bound” of electoral democracies (N = 13) or in the “upper bound” of electoral autocracies (N = 5).⁹ We treat them as “grey zone” regimes.

Closed autocracies are increasing, from a low of 22 in 2012 to 33 in 2022-2023. There is quite some variation in the ways countries have descended into closed autocracies in recent years. For example, Afghanistan became a closed autocracy following the Taliban insurrection, Chad because of a military coup, and countries like Haiti, Mali, Myanmar, and Sudan, as a result of growing anarchy, breakdown of formal institutions, and military interventions. These examples remind us that the current wave of autocratization is not only, or perhaps not even mainly, about democratic backsliding. A significant portion of changes for the worse in the world comes from autocracies becoming more autocratic.

The number of electoral autocracies has been growing markedly in numbers over the past 50 years, from 36 in 1973 to peak at 65 in 2012, down to 55 in 2023. Much of the upward trend is explained by many closed autocracies liberalizing in the 1980s and 1990s and starting to hold multiparty elections. Some became democracies, but many stalled as electoral autocracies. The current decline is partially due to countries dropping down into the category of closed autocracy.

Electoral democracies amount to 59 in 2023, a four-fold increase from 15 in 1973. The gradual increase in electoral democracies in the last decade is in part the result of the decline of liberal democracies. Liberal democracies declined from a peak of 43 in 2007–2012 to 32 in 2023.

Figure 4 uses the RoW to assess these developments weighted by population. It shows that 71% of the world’s population – 5.7 billion people – live in electoral or closed autocracies in 2023.¹⁰ That is an increase from 48% of the world’s population ten years ago. The plurality – 44% of the world’s population, or 3.5 billion people – reside in electoral autocracies. Closed autocracies account for 27% of the world’s population, or 2.2 billion people. Electoral democracies, despite numbering 59, host only 16% of the world’s population, while the 32 liberal democracies are home to the smallest share, or 13% of the world’s population.

Figure 4. Regime Types by Number of Countries and World Population, 1973–2023.

Note: Figure 4 plots the number of countries (left panel) and the world population (right panel) by regime type. The grey area in the right panel indicate uncertainty about the exact regime classification of some countries.

Figure 5 shows the variation across regions in the share of the population living in each regime category. East Asia and the Pacific, South and Central Asia, the Middle East and North Africa (MENA), Sub-Saharan Africa, and Eastern Europe all have a majority of the population living in autocracies. South and Central Asia and MENA rank the lowest, with around nine out of 10 people residing in autocracies. Western Europe and North America remains the most democratic region of the world, and most of its inhabitants (96%) live in liberal democracies. The vast majority of Latin Americans (90%) also live in democracies. However, Latin America is also the region with the largest share of the population living in “grey zone” regimes: 24% of people reside in countries in the lower bound of electoral democracies that could be, or soon be, electoral autocracies.

Figure 5. Regional Shares of Population by Regime Type, 2023.

Note: Figure 5 plots the share of the population living in liberal democracies, electoral democracies, electoral autocracies, and closed autocracies, including “grey zone” subcategories, by regions of the world.

Elections are becoming less democratic

In this section, we drill down into the components of democracy, to dissect which aspects of democracy are affected the most by the current wave of autocratization. Figure 6 shows the number of countries declining (significantly and substantially meaningfully) in components of democracy from 2003 to 2013 (left panel) and 2013–2023 (right panel). If a component is above the diagonal line, it is improving in more countries than declining, and vice versa.

Figure 6. Democratic Aspects Improving and Declining, 2013 and 2023.

Note: For indices measuring components of democracy, Figure 6 shows the number of countries improving and declining significantly and substantially. An index is declining substantially and significantly if its 2023 value is at least 0.05 points lower than its 2013 value on a scale ranging from 0 to 1, and the confidence intervals do not overlap. The left panel shows changes in 2013 compared to 2003, and the right panel shows changes in 2023 compared to 2013.

Ten years ago, almost all component indices were above the diagonal line. In particular, the Clean Elections Index (CEI) was improving around the world overall: improving in 23 countries and worsening in 10, between 2003 and 2013. Over the past ten years, the situation has turned around. Counting only substantial and statistically significant changes, between 2013 and 2023 the CEI declined in 23 countries and improved in twelve. This is a new development: Democracy’s core institution, the election, is now increasingly undermined in more countries around the world.

For several years, we have documented that Freedom of Expression is the aspect of democracy most often under attack during autocratization, and often targeted first when countries decline democratically. It remains the worst affected aspect of democracy in 2023, and is deteriorating in 35 countries compared to ten years earlier, while improving in only eleven. Ten years ago, it was improving in twelve countries and declining in eleven. Freedom of Association is now deteriorating in 20 countries and improving in only three. Deliberation, Rule of Law, and Legislative Constraints are also declining in more countries than improving in 2023, again in contrast to the comparison between 2003 and 2013 (left panel), where almost all components were improving in more countries than they were declining.

Drilling down further into the indicators of these components, Figure 7 shows the top-20 indicators that have declined the most in the past decade. It sheds further light on the described patterns of change. The aspects of elections that deteriorate the most, are the extent to which they are free and fair, and the extent to which electoral management bodies (EMB) are autonomous. From media censorship and harassment of journalists to individual freedom of expression and cultural academic expression, a series of indicators of freedom of expression decline. In addition, civil society continues to be under pressure during the current wave of autocratization, with both CSO repression and CSO entry and exit among the top ten declining indicators.

Figure 7. Top-20 Declining Indicators, 2013–2023.

Note: Figure 7 plots the top 20 indicators that decline significantly and substantially in the largest number of countries between 2013 and 2023. The red line marks the top 10 indicators. An indicator is declining substantially and significantly if its 2023 value is at least 0.5 points lower than its 2013 value on a scale from 0 to 4 (for most indicators) or 0–5, and the confidence intervals do not overlap.

Overall, the most notable change in Figures 6 and 7 compared to our analyses in the State of the World articles in previous years is that elections are now increasingly undermined by governments. This is also a notable change among the indicators that are worsening in the subsample of autocratizing countries (see Figure 8A, Online Appendix). As recently as the State of the World 2020 article,¹¹ we showed that electoral decline tends to come late in the autocratization process, following the derailing of a series of other aspects of democracy. Now the data paints a very different picture. It is also worth noting that we observe these declines in many relatively democratic, or until recently democratic, countries such as Ghana, Hungary, India, Mexico, and The Philippines. In a separate V-Dem thematic report, we study recent trends for electoral integrity across the world.¹²

Trends of regime transformations

Figure 8 shows the number of autocratizing and democratizing countries (left panel) and the respective share of the world population (right panel) over the past 50 years. We use the ERT methodology¹³ to identify episodes of autocratization and democratization.

Figure 8. Autocratizing vs. Democratizing Countries, 1973–2023.

Note: Figure 8 shows patterns of democratization and autocratization over the last 50 years. The left panel displays the number of countries in each category, and the right panel shows the share of the world’s population living in autocratizing or democratizing countries.

As of 2023, 42 countries are in ongoing episodes of autocratization while 18 countries are in ongoing episodes of democratization. The number of democratizing and autocratizing countries are practically showing an inverse trend since the mid-1990s. From a record 70, registered in 1992, the number of democratizers has been declining for more than 30 years. By contrast, the number of autocratizing countries was close to zero at the height of the third wave of democratization but has been inclining steadily for almost 30 years since then. Countries in episodes of ongoing autocratization now host 35% of the world’s population while democratizers are home to only 5%. A large respective share of autocratizers is India, which amounts to about half of the population living in autocratizing countries; but there are several other autocratizers with sizeable populations, like Indonesia, Mexico, and Pakistan. Brazil makes up more than half of the population among the democratizers.

The noticeable decline in the number of autocratizing countries and the slight increase in the number of democratizing countries in the last two years is potentially good news. It could indicate that the current wave of autocratization has crested. Yet, one should be careful not to immediately draw that conclusion. The ERT methodology is quite demanding for countries to qualify as manifest episodes of autocratization and it typically takes a few years of declines to accumulate to reach the thresholds. In short, the methodology strives to minimize the error of including false positives among the count. The data for 2023 shows that 25 countries registered declines in democracy levels over the past couple of years. This makes them “potential” episodes of autocratization, but the criteria for declaring them “manifest autocratizers” are not yet completely fulfilled. If deteriorations in just a handful of these countries continue in the coming years, the 42 recorded for 2023 would increase.

Figure 4A in the Online Appendix uses the population weight to show that many autocratizers, beyond just India, are populous, for example, Indonesia, Myanmar, Mexico, Pakistan, The Philippines, and South Korea. The democratizing countries are typically smaller in size, economy, and population, for example, The Gambia, The Maldives, and The Seychelles. As noted, the recent addition of Brazil is an exception to this, as is Thailand. Figure 5A in the Online Appendix sheds further light on the regional variation in autocratization and democratization.

Stand-alone autocratization and bell-turns

Quantitative studies of regime change typically analyse democratization and autocratization in time-series analyses with provincial one-year differentials, ignoring what precedes and comes after it.¹⁴ The new ERT methodology¹⁵ offers a considerable improvement on our analytical leverage by delineating full episodes of autocratization and democratization and making them the unit of analysis. At the same time, the ERT method for systematically identifying all episodes across time provides an unprecedented tool for case selection for comparative analyses.

Building on the work of Nord et al., we extend the ERT logic and coding rules, and distinguish here between two types of autocratization: “Stand-alone” autocratization, or episodes of deterioration in democracy levels that starts after a period of relative stability; and “Bell-turns,” or “periods of two-directional regime transformation along a democracy-autocracy continuum, in which democratization is closely followed by and inherently linked to subsequent autocratization.”¹⁷ Bell-turns should be especially interesting for our field of study as instances of “failed democratization.”

The complete list of the 42 countries that are at present in ongoing episodes of autocratization is presented in Figure 9. 42 autocratizing countries are grouped into the 23 stand-alone and 19 Bell-turn processes. Countries are ordered according to their levels on the LDI at the onset of autocratization. The LDI score for 2023 (marked in orange) reveals the total magnitude of deterioration by the end of 2023. The first important finding is that based on their classification by the RoW (not shown in the graph), 28 of the 42 were democracies when the episode started, and only about half of those (N = 15) remain democracies in 2023. This indicates that democratic breakdown is common.

Figure 9. All 42 Cases of Ongoing Autocratization, 2023.

Note: Figure 9 shows the LDI scores and confidence intervals at the start of autocratization. By the ERT methodology, these are technically “pre-episode” in the sense that the year after was the first recording a decline. The values in 2023 are highlighted in orange and reveal the total magnitude of change. The ERT is based on the Electoral Democracy Index (EDI) for identifying episodes, and we follow that but display countries’ values on the LDI. A country may change more on the EDI than on the LDI, which explains cases with overlapping confidence intervals (they do not overlap on the EDI).

Among the 10 stand-alone autocratizers to have declined the most (see Figure 6A, Online Appendix), eight were democracies before the start of autocratization. Democracy broke down in six of those eight cases – The Comoros, Hungary, India, Mauritius, Nicaragua, and Serbia. Only Greece and Poland remain democracies in 2023. This frequency of democratic breakdown coheres with recent research showing that 80% of democracies break down if they start autocratizing.¹⁸ It is moreover worth noting that autocratization in all these eight countries has taken place under democratically elected leaders who have gradually dismantled democratic institutions, although at different speeds and extent. This is different from democracies breaking down through military coups or invasion, which was more common in the past.¹⁹ Two countries among the top “stand-alone” autocratizers started out as autocracies. Afghanistan went from an electoral to a closed autocracy during the process, while the initially low level of democracy in Hong Kong’s closed autocracy continued to depreciate further in 2023.

We also assess the trajectories of the 10 Bell-turns that have autocratized the most (Figure 7A, Online Appendix). Eight of the top 10 Bell-turn autocratizers were democracies at some point during their episode: Armenia, Burkina Faso, El Salvador, Indonesia, Libya, Mali, Niger, and South Korea. At present only three – Armenia, Indonesia, and South Korea – remain democracies. Out of the five where democracy broke down, three – Burkina Faso, Libya, and Mali – are now closed autocracies. The survival rate among democracies undergoing the Bell-turn type of autocratization is thus about the same as for the stand-alone autocratization processes. It is worth noting that five of the Bell-turn countries – Armenia, Burkina Faso, Libya, Myanmar, and Niger – were among the top 10 democratizers of the last decade. South Korea – a country that was featured as a rare case of democracies bouncing back in last year’s State of the World article²⁰ – has now reversed back to its pre-democratization (deepening) episode levels but remains a liberal democracy. El Salvador and Indonesia went through very slow but steady episodes of democratization for more than a decade before the reversals. In other words, these were promising cases of democratization and their reversals into Bell-turns points to the uncertainty of holding on to democratic gains across very different contexts.

As instances of failed democratization, Bell-turns are of particular interest for understanding vulnerabilities of democratization processes and should be studied in-depth for that purpose. While beyond the scope of this article, we suggest that such analyses are a prime avenue for future research in the field.

Stand-alone democratization and U-turns

Building on Nord et al.²¹ and mirroring the analysis of autocratization, we divide the 18 democratizers into “stand-alone” democratization for episodes of improvement starting after a period of relative stability; and following Nord et al.,²² we use the term “U-turns” to define a “period of substantive two-directional regime transformation along a democracy-autocracy continuum, in which autocratization is closely followed by and inherently linked to subsequent democratization.”

The complete list of the 18 ongoing episodes of democratization grouped into the nine stand-alone and nine U-turn processes, is presented in Figure 10. Countries are ordered according to their levels on the LDI at the onset of democratization. The LDI score for 2023 (marked in blue) reveals the total magnitude of democratic improvement by the end of 2023.

Figure 10. All 18 Cases of Ongoing Democratization, 2023.

Note: Figure 10 shows the LDI scores and confidence intervals at the start of democratization. By the ERT methodology, these are technically “pre-episode” in the sense that the year after was the first recording an increase. The values in 2023 are highlighted in blue and reveal the total magnitude of change. The ERT is based on the Electoral Democracy Index (EDI) for identifying episodes, and we follow that but display countries’ values on the LDI. A country may change more on the EDI than on the LDI, which explains cases with overlapping confidence intervals (they do not overlap on the EDI).

Of the 18 democratizers, 14 were autocracies by their RoW classification (not shown in graph) at the start of their episode and four were already democracies, and thus experienced further democratic deepening. By 2023, 15 of the 18 are democracies, an increase by 11. This overview shows that democratization can take off in different circumstances, which is reflected by the distribution of starting levels on the LDI in Figure 10.

The nine stand-alone democratizers (see Figure 9A, Online Appendix, for more detail) are all relatively small countries with a combined population of about 30 million people, or 0.38% of the world’s population. Dominican Republic (11 million) and Honduras (10 million) account for 70% of those 30 million, speaking to the dominance of autocratization as a more influential global trend. Seven out of nine “stand-alone” democratizers were autocracies before the start of democratization – only Dominican Republic and Kosovo began their episodes as democracies to improve further. All nine countries are democracies in 2023, but only The Seychelles has developed into a liberal democracy so far.

The nine U-turns show a somewhat different pattern (see Figure 10A, Online Appendix). Three of the nine U-turns have by 2023 restored pre-autocratization levels of democracy – Lesotho, The Maldives, and Zambia. The other six – Benin, Bolivia, Brazil, North Macedonia, Thailand, and Tunisia – remain at lower levels in 2023, and some substantially so. Since we focus on ongoing U-turn democratization episodes, there is still a possibility that their trajectories continue upwards.

Two countries – Brazil and Lesotho – managed to halt autocratization before a democratic breakdown and are now “bouncing back.” Four countries – Bolivia, The Maldives, North Macedonia, and Zambia – went through democratic breakdowns but U-turns are turning them back to democracies. Two countries – Benin and Tunisia – suffered democratic breakdowns and while their LDI levels are improving, they remain quite distant from regaining democracy. Finally, Thailand is recovering from an authoritarian regression.

U-turns are also of particular interest for understanding why and how some countries successfully halt and revert autocratization. Although we do not provide answers to such questions in this article, we believe these are topical questions for the current scholarly development on democratic resilience.²³ Nord et al.²⁴ identify and provide descriptive analyses of all 98 cases of U-turns since 1900, which could be the basis for further research development to probe and rejoin with in-depth, comparative, and statistical analyses of what helps to turn autocratization into processes of democratization.

New findings on waves of regime transformation

Going beyond the Democracy Report 2024, here we provide the first ever re-analysis of the well-known three waves of democratization and autocratization by unpacking them in terms of stand-alone regime transformations vs Bell-turns and U-turns. These analyses show that the current development of regime changes is growingly contested and highly volatile.

Starting by re-approaching the three waves of autocratization,²⁵ Figure 11 shows the frequency of stand-alone and Bell-turn autocratization episodes from 1900 to 2023. The first wave of autocratization (circa 1922–1942) consisted mostly of episodes of stand-alone autocratization, as visualized by the light-red predominance in Figure 11. Prominent cases of stand-alone episodes are the electoral victories of Hitler and Mussolini leading to the democratic breakdowns in Germany and Italy.²⁶

Figure 11. Three Waves of Autocratization, by Stand-Alone and Bell-Turn Episode Types, 1900–2023.

Note: This figure shows the historical development of autocratization between 1900 and 2023. Autocratization episodes are divided into stand-alone and Bell-turn episodes. Note that only the autocratization part of the Bell-turn episodes is included in the count.

During the second wave of autocratization (circa 1960–1975), Bell-turn episodes became common making up from 50% (1961) to 72% (1966) of all autocratization episodes (58% on average across the wave). This is a new finding regarding the second wave of autocratization. For instance, Lührmann and Lindberg’s²⁷ analysis suggested that autocratization during the second wave predominantly consisted of military coups and took place in already authoritarian contexts. Although that is true, it is only one part of the story. Introducing the analytic lens of Bell-turn episodes, we can now show that the largest share of autocratization episodes during the second wave were in fact the result of failed democratization attempts that were foiled by coups.

Bell-turns constituted an even greater share during the first years of the third wave of autocratization (from circa 1994 and ongoing) with a peak at 88% in 1996. On average, Bell-turn episodes make up almost half (49%) of all autocratization episodes during the third wave. Episodes of failed democratization turning into autocratization thus continue to be frequent. This was overlooked in the literature on autocratization, which has paid more attention to the mechanisms through which autocratization takes place and expands in more recent times (e.g. “executive aggrandizement”²⁸ and attacks on accountability measures²⁹), rather than to the roots of autocratization. Importantly, Bell-turns of the third wave of autocratization take place not only via coups in electoral autocracies, but also as reversals of democratization that happen in already established and seemingly consolidated democracies as in recent cases of Türkiye and South Korea. Future in-depth analyses of the causes and consequences of “third-wave” Bell-turns happening in democratic settings could shed light on whether these are outcomes of “ceiling effects” or being driven by different origins.

The fact that almost half of all autocratization episodes during the third wave are, in fact, episodes of reversed democratization (Bell-turns) is one of the most important underexplored patterns of regime transformation in recent academic and practitioner’s debates.

U-Turns on the rise

In our re-analysis of the three democratization waves,³⁰ Figure 12 shows the frequency of stand-alone and U-turn democratization episodes from 1900 to 2023 and presents new insights.

Figure 12. Three Waves of Democratization, by Stand-Alone and U-Turn Episode Types, 1900–2023.

Note: This figure shows the historical development of democratization episodes between 1900 and 2023. Democratization episodes are divided into stand-alone and U-turn episodes. Note that only the democratization part of the U-turn episodes is included in the count.

Overall, episodes of U-turn democratization are much less frequent among instances of democratization than Bell-turns are among episodes of autocratization. The first (circa 1900–1926) and the second (circa 1945–1965) waves of democratization were characterized mostly by authoritarian countries improving on democratic levels.³¹ There was a short period in the 1930s when the share of U-turns was 61% on average but in absolute numbers there were very few. U-turn episodes were a marginal proportion of all episodes of democratization after that and touched the lowest point in 1960 with only 3% of all democratization episodes.

In 1974, when the Portuguese Carnation Revolution initiated the third wave of democratization,³² 89% of all episodes of democratization were stand-alone democratizations. After the fall of the Berlin Wall, the number of democratization episodes peaked with 70 countries at one point and over 90% of these were stand-alone. Similar to processes in previous waves, they were the result of the fall of previously relatively stable dictatorships. This finding demonstrates that democratization is usually an independent regime transformation (i.e. moving from more autocratic to more democratic political systems) rather than a reaction to autocratization.

Yet, the U-turn mode of democratization started to gain traction in relative terms by the beginning of the twenty-first century rising from 13% in 2001 to 50% in 2023. Figure 12 shows that the proportions of stand-alone democratizations and U-turns have become progressively similar over the last two decades. On average, 30% of all episodes of democratization of the twenty-first century are U-turns, with the average share rising from 18% in the 2000s to 46% during the last several years. U-turns today are, on average, as common as a “classic” stand-alone type of democratization despite the overall fall in the number of countries in democratization episodes at all.

We interpret this progressive increase in relative numbers of U-turns over stand-alone democratizations in two ways. First, from a historical perspective, it means that democratic resilience is unprecedented and halting and reversing autocratization becomes ever more common. It also signals that “stage two” resilience³³ against anti-democratic incumbents is now much more frequent than during the twentieth century. This finding strengthens previous literature focusing on recent case studies on how autocratization can be stopped and, more importantly, reversed.³⁴

Second, there is a growing effort from opposition actors to maintain democratic institutions in place, implying that opposition to anti-democratic incumbents might be increasing in the future. This could take several forms, including voting anti-democratic incumbents out of office, as the recent case of Brazil shows,³⁵ growing opposition from social movements, as in South Korea and North Macedonia,³⁶ or the rise of elite oppositions, as in Benin.³⁷

Conclusions and outlook

This article adds to the on-going scholarly debate about the state of democracy in the world. We analyse the global averages over time, changes in regime types, developments in sub-indices and indicators, and ongoing regime transformations in terms of number of countries undergoing autocratization and democratization. Regardless of how the data is sliced, the general pattern is that democracy is globally in decline. As of 2023, 42 countries with 35% of the world population are autocratizing while only 18 countries with 5% of the world population are democratizing. We have also shown that this decline was driven in the past decade by, in part, significant declines in freedom of expression and reduced numbers of clean elections.

Going beyond that observation, this article provides an innovative re-evaluation of the three waves of autocratization and democratization by dividing them into stand-alone autocratization and Bell-turns, and stand-alone democratization and U-turns, respectively. This re-appraisal of the waves of regime changes shows one central aspect: countries’ political regimes in recent years are increasingly volatile. In historical perspective since 1900, no other period has had such a high frequency of Bell-turns and U-turns than the last decade. As recent literature shows,³⁸ there is a growing spread of disinformation feeding societal and political and polarization putting the question of democracy progressively at the center of those divisions.³⁹ The analysis here suggests, but does not conclusively show, that this results in an uptake of two-directional episodes of regime transformation. It is worth keeping in mind that this increased volatility takes place within a context where autocratization is the dominant trend.

Looking forward, we especially encourage scholars to build on the descriptive findings of Bell- and U-turns to unpack their origins and mechanisms. Our first descriptive analysis suggests that in-depth, comparative case studies of these cases of reversals and their counterfactuals would be an important area of future research.

Looking at the present year, twenty twenty-four is “the biggest election-year” in over a century, with citizens in 60 countries⁴⁰ called to cast their votes worldwide. Countries holding national elections make up nearly half of the world’s population (45%), and the list includes seven of the world's ten most populous countries: Bangladesh, India, Indonesia, Mexico, Pakistan, Russia, and the United States. This might translate into a make-or-break year for democracy in the world.

Figure 13 groups the 60 countries that hold elections in 2024 into declining, improving, and a category of “others.”⁴¹ More than half – 31 countries – are in periods of decline on democracy. Eleven of them are “stand-alone” episodes of autocratization while nine are “Bell-turn” autocratizers. The ERT thresholds deliberately exclude smaller changes because they could stem from measurement uncertainty in the underlying indicators. However, by engaging in an exercise of lowering the ERT threshold from a hard 0.1 to a slightly more permissive range from 0.05 to 0.1, we show the existence of eleven “near-misses” – countries that are on the verge of autocratization. This points at the potential of high relevance of elections in these countries to determine whether an autocratization episode takes hold, or not.

Figure 13. Elections in 2024.

Note: This figure shows the number of countries holding elections in 2024 for which V-Dem data is available. Countries are grouped into declining, improving, and other regimes. Note that in this figure we include “near misses” in both categories of changing regimes.

Only 5%, or three out of the 60 countries holding elections in 2024, are improving on democracy, and all three are processes of U-turn democratization, meaning that they are recovering from previous democratic decline. Finally, 26 countries fall into the “other” category, meaning that they do not register substantial and ongoing improvements or declines in democracy levels.

Among the eleven “stand-alone” autocratizers, we find India, Mexico, and Senegal. The nine “Bell-turn” autocratizers include El Salvador, Indonesia, and Pakistan who held elections earlier in February this year. Elections in the eleven countries who are “near misses” of autocratization, for example South Africa, are expected to be consequential. If the incumbents win the elections and stay in power, it seems reasonable to expect that further deteriorations would follow. The three democratizers holding elections in 2024 are all “U-turn” processes: North Macedonia, The Maldives, and Tunisia.

Among the group of “others,” we find the United States with the likely run-off between President Biden and former President Trump looking to be highly consequential. The expected Republican nominee and former President Donald Trump undermined American democracy substantially during his previous tenure, not least regarding media freedom, judicial independence, and executive oversight. A second presidential term for Trump could have significant ramifications also for democracy internationally given the United States’ key role in world affairs and the affinity for autocrats Trump sported during his first term.

Elections in autocratizing countries are critical events that can shape the future of the country’s political regime. Brazil is a recent example of the key role of elections in upending autocratization. Last year’s State of the World article shows that this is a common pattern in recent cases of U-turn democratization.⁴² Surprise victories for a democratic opposition in critical elections can lead to the ousting of an incumbent, even in autocratic settings. The Maldives and Zambia are two recent examples of this. Contrastingly, elections can also serve as powerful instruments of legitimation and spur further autocratization when challengers fail,⁴³ such as in Hungary and Türkiye in recent times.

The fact that a majority of elections during the “record election year” 2024 take place in such contested spaces makes this year likely to be critical for the future of democracy in the world. How many of the 2024 national elections end up legitimizing further autocratization, and how many lead to a halting – or even reversing – of that process remains to be seen.

Acknowledgements

This article is based on the V-Dem Institute’s Democracy Report 2024: Democracy Winning and Losing at the Ballot (Nord et al., 2024). We thank Cecilia Borella, Lisa Gastaldi, Ana Good God, and Natalia Natsika for their contributions to the Democracy Report 2024.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.

Additional information

Funding

This research was made possible through support by Swedish Research Council, Grant [2014-00038]; Knut and Alice Wallenberg Foundation to Wallenberg Academy Fellow Staffan I. Lindberg, Grant [2018.0144]; and European Union, represented by the European Commission, Grant [2023/442-497]; as well as by internal grants from the Vice-Chancellor’s office, the Dean of the College of Social Sciences and the Department of Political Science at University of Gothenburg.

Notes on contributors

Fabio Angiolillo

Fabio Angiolillo is a postdoctoral research fellow at the V-Dem Institute, University of Gothenburg. His research focuses on comparative politics, political institutions, authoritarian regimes, political parties, and political behaviour.

Martin Lundstedt

Martin Lundstedt is an assistant researcher at the V-Dem Institute, Department of Political Science, University of Gothenburg.

Marina Nord

Marina Nord is a postdoctoral research fellow at the V-Dem Institute, University of Gothenburg. Her research focuses on comparative politics, democratization, autocratization, and the interaction of political and economic systems around the world.

Staffan I. Lindberg

Staffan I. Lindberg is Professor of Political Science and Director of the university-wide research infrastructure at the V-Dem Institute, University of Gothenburg, founding Principal Investigator of Varieties of Democracy (V-Dem), founding Director of the national research infrastructure DEMSCORE, Wallenberg Academy Fellow, author/editor of several books, and 60 + articles on democracy, elections, democratization, autocratization, accountability, clientelism, sequence analysis methods, women's representation, and voting behavior in journals such as APSR, AJPS, Political Analysis, Democratization, World Development, World Politics, etc.

Notes

* The article draws on the V-Dem Institute’s Democracy Report 2024: Democracy Winning and Losing at the Ballot (Nord et al., 2024) and has been reviewed by the co-editors-in-chief of Democratization before publication. Authors are listed in alphabetical order (except for Lindberg as the research team leader) to reflect the collective effort. Equal authorship applies.

1 Coppedge et al., “V-Dem Country-Year Dataset v14”.

2 Maerz et al., “Episodes of Regime Transformation.”

3 The confidence intervals, or strictly speaking highest posterior density regions, are estimates of upper and lower bounds of range within which the ‘true’ value may be located taking uncertainty and measurement error into account.

4 For example, Bermeo, “On Democratic Backsliding;” Lührmann and Lindberg, “A Third Wave of Autocratization;” Mounk, “The People vs. Democracy;” Runciman, “How Democracies End.”

5 For example, Brownlee and Miao, “Why Democracies Survive;” Levitsky and Way, “Democracy’s Surprising Resilience;” Weyland, “Democracy’s Resilience to Populism’s Threat.”

6 Knutsen et al., “Conceptual and Measurement Issues in Assessing Democratic Backsliding.”

7 Lührmann, et al., “Regimes of the World (RoW).”

8 Ibid.

9 There is a variant of the RoW-measure that models uncertainty and identifies countries close to thresholds, see the variable v2x_regime_amb in the V-Dem Codebook, v14.

10 Percentages are rounded throughout this article. The 71% mentioned here builds aggregating rounded figures for liberal and electoral autocracies in Figure 4. Population figures come from the World Bank.

11 Hellmeier et al., “State of the World 2020.”

12 See Nord et al., “Democracy’s Core Institution” for further discussion about which aspects of electoral integrity deteriorate the most during process of autocratization.

13 Maerz et al., “Episodes of Regime Transformation.”

14 Nord et al., “When Autocratization is Reversed.”

15 Maerz et al., “Episodes of Regime Transformation.”

16 Nord et al., “When Autocratization is Reversed.”

17 This wording is borrowed from Nord et al., “When Autocratization is Reversed.” U-Turn is defined as a “period of substantive two-directional regime transformation along a democracy-autocracy continuum, in which autocratization is closely followed by and inherently linked to subsequent democratization” (p.9). Reduced to the bare bones, one can say that the ERT method (see Maerz et al., “Episodes of Regime Transformation”) works by first identifying small annual changes (>0.01 on the 0–1 scale) in a country’s level of democracy measured by V-Dem’s Electoral Democracy Index (EDI). If they accumulate within several years to substantial (>0.1), it is an episode. If changes are smaller, it is deemed mere fluctuations without substantial change that could result from uncertainty in the underlying measures.

18 Boese et al., “How Democracies Prevail.”

19 Lührmann and Lindberg, “A Third Wave of Autocratization.”

20 Wiebrecht et al., “State of the World 2022.”

21 Nord et al., “When Autocratization is Reversed.”

22 Ibid, p. 9.

23 Laebens and Lührmann, “What Halts Democratic Erosion;” Gamboa, “Resisting Backsliding;” Tomini et al., “Standing up Against Autocratization;”. Wiebrecht et al., “State of the World 2022.”

24 Nord et al., “When Autocratization is Reversed.”

25 Lührmann and Lindberg, “A Third Wave of Autocratization.”

26 Juan et al., “War and Nationalism;” Acemoglu et al., “War, Socialism, and the Rise of Fascism.”

27 Lührmann and Lindberg, “A Third Wave of Autocratization,” p. 1103.

28 Bermeo, “On Democratic Backsliding.”

29 Laebens and Lührmann, “What Halts Democratic Erosion;” Lührmann and Lindberg, “A Third Wave of Autocratization.”

30 Huntington, “The Third Wave.”

31 Ziblatt, “Conservative Political Parties;” Huntington, “The Third Wave.”

32 Huntington, “The Third Wave.”

33 Boese et al., “How Democracies Prevail.”

34 Gamboa, “Resisting Backsliding;” Laebens and Lührmann, “What Halts Democratic Erosion;” Wiebrecht et al., “State of the World 2022.”

35 Nord et al., “Democracy Report 2024.”

36 Laebens and Lührmann, “What Halts Democratic Erosion;” Wiebrecht et al., “State of the World 2022.”

37 Laebens and Lührmann, “What Halts Democratic Erosion.”

38 Angiolillo et al., “Democratic-Autocratic Party Systems;” McCoy et al., “Polarization and the Global Crisis of Democracy;” Somer et al., “Pernicious Polarization;” Svolik et al., “In Europe, Democracy Erodes from the Right.”

39 In light of the challenges to democracies worldwide, the V-Dem Institute summarized the wealth of research pointing to the benefits of democracy. See Lundstedt et al., “Case for Democracy Report.”

40 We discuss the 60 countries for which data is available in the V-Dem dataset v14. Four additional countries also hold national elections in 2024: Kiribati, Palau, Tuvalu, and San Marino. Additionally, V-Dem does not code elections to the European Union Parliament, so countries holding only those elections during 2024 are also excluded.

41 “Other” refers to countries that do not reveal any ongoing substantial changes in democracy levels according to the ERT rules. The most prominent country in this category is the United States of America.

42 Wiebrecht et al., “State of the World 2022.”

43 Knutsen, et al., “Autocratic Elections.”