Erosion or decay? Conceptualizing causes and mechanisms of democratic regression

ABSTRACT Democratic regression has become a worrying phenomenon in the last years. Social science has provided a variety of explanations why democratic regimes have lost democratic regime quality. Against this backdrop, I take stock of the recent literature by putting forward two important analytical distinctions that we should make more explicit. First, I propose to classify our current explanations along the source where the cause for the malaise originated. By doing so, I introduce a distinction between erosion and decay type of arguments. While the former is a gradual process that is caused exogenously – like wind or water hitting a stone – the latter is caused endogenously – like the half-life in nuclear decay processes. Second, I draw a distinction between the endogenous or exogenous roots of the cause and the subsequent causal mechanism that connects the cause with the outcome. I outline the need for dissecting a causal mechanism into its constitutive components and highlight its underlying dimensions of temporality. Throughout the article, I use empirical case material as well as relevant secondary literature to illustrate these points.

to my readingdensely populated with in-depth case studies that seek to explain rather the particular, but that lack theoretical and conceptual coherence. This article is motivated by reviewing these insightful accounts and classifying them in a novel and hopefully helpful fashion that seeks to facilitate dialogue across individual cases. As such, the article can be understood as a taking stock endeavour with a constructive intent.
To be clear, there is an abundance of explanatory factors that are currently discussed. These factors are usually combined into bigger clusters, often ordered along the basic social science distinction between economic, societal, cultural, and political factors, 4 or, alternatively, between supply-side and demand-side explanations. 5 A more fine-grained distinction is made by Waldner and Lust. 6 They use the classic democratization theories that have been advanced in the last decades and turn them on their head for explaining the reverse trend of autocratization. In this light, they distinguish between agency-based theories, political culture approaches, institutionalist theories, theories of political economy, social structures and coalitional arguments, and international factors. Tying different explanatory factors to the wealth of democratization studies is intuitive, but represents, of course, only an approximation that, above all, needs to assume a symmetrical argumentation: When X has led to democratization, the absence of X must lead to autocratization. Remarkably, the editors of this special issue suggest to further distinguish between proximate and deep causes of democratic regressions, 7 harking back to Mayr's 8 influential distinction on the nature of causation in biology.
In this article, I take a different stance. I suggest that the basic distinction should be made between endogenous and exogenous factors, 9 proposing that the source of the cause could constitute a helpful axis along which we could classify arguments about democratic regression. This contribution explicitly adopts a historical institutionalist perspective and argues that the institutionalist repertoire of explaining change is a well-suited starting point for sorting our current explanations of democratic regression. More specifically, I use the approach by Thelen and colleagues 10 on gradual institutional change as a theoretical springboard to distinguish between two forms of gradual change. Borrowing metaphors from natural science, I draw a distinction between erosion and decay processes. While decay is endogenously caused, erosion is an exogenously driven process. I argue that it is of utmost importance to clarify in the first place where the causal driver for democratic regression is actually located. This is the major axis along which I classify the causes of democratic regression. This distinction has important implications not only for adequately explaining the empirical phenomenon of democratic regression, but also for sound policy advice.
The distinction between erosion or decay arguments has also important repercussions on the design of the causal mechanism that translates the causes to the outcome. As such, the causal mechanism is either being rooted in an endogenous or exogenous cause. In other words, the causes constitute the anchor points from which one needs to further develop the inner logic and unfolding of the concrete causal mechanisms. I outline the need for dissecting a causal mechanism into its components and highlight underlying dimensions of temporality. As such, I pay emphasis on marking a difference between the cause X and the causal mechanism M, arguing that they should be kept separate for analytical precision.
The same argument about analytical precision applies for an obvious concern. A complex phenomenon like democratic regression is not mono-causal. There is not only one cause that triggers one causal mechanism that leads to the outcome of democratic regression. Rather, it is a conjunction of several streams and mechanisms that might partially overlap temporally and substantially, that affect each other, be it in a mutually reinforcing or mutually undermining way. Yet, I think it is safe to argue that acknowledging the complex interplay of these mechanisms does not prevent us from disentangling them from each other. In contrast, analytical precision demands to dissect them form each other. As such, I see the proposed distinction between (a) endogenous and exogenous causes and (b) causes and causal mechanisms as important steps in taking apart complex explanations.
The article is structured as follows. In section II, I will go back to the historical neoinstitutionalist debate about explaining institutional change. I attempt to bring their toolbox back into the debate, enriching our understanding of democratic regression. Based on this discussion, I propose in section III a new axis along which we should classify our arguments. Depending on the source where the causes for the malaise originates, I distinguish between exogenous erosion and endogenous decay arguments. In section IV, I then zoom into the causal mechanisms M. I conclude in section V by first synthesizing major insights of this article and then highlighting future research avenues.

The historical institutionalist toolkit: forms of gradual change
In this article, I draw on the rich toolkit of historical neo-institutionalism. 11 Historical institutionalism provides not only an ample, but also a well-sorted arsenal in explaining change. Against this backdrop, I argue that it can be made fruitful for systematically thinking about processes of democratic regressions.
Yet, it should be noted that historical neo-institutionalism has long been plagued by a stability bias. 12 Using its typical instrumentarium, it is easier for historical institutionalists to explain the long phases of institutional reproduction than to actually account for changes. The "punctuated equilibrium", the dualism of path dependency and critical juncture, is almost essential knowledge for this theoretical approach. 13 In most accounts, long phases of institutional stability that are often portrayed as being based on increasing returns to scale 14 are interrupted by brief moments of change in which decisive action is taken in either direction. 15 The Achilles heel of this explanatory dualism, however, has always been how the brief interruptions that enable change come about in the first place. This task was often relegated to the mere occurrence of critical junctures.
Against this backdrop, it is particularly the research agenda of Kathleen Thelen and colleagues who challenged this approach. 16 She deviates from the "common wisdom" of historical institutionalism in two important regards. First, she is more interested in explaining how change comes aboutrather than assuming that there are windows of opportunities in which "permissive" and "productive conditions" make change likely. 17 Second, she is more interested in gradual forms of change. 18 While the "punctuated equilibrium" approach focuses on a sudden interruption, the Thelen agenda adopts a longer time horizon. The duration as the temporal length of a process in which change takes place is extended. Framing it differently: while the longue durée that is characteristic for many works in the historical institutionalism tradition refers often to the development of the outcome after the critical juncture is taken, this approach uses extended and stretched periods of time in order to explain how the outcome came actually about over time.
I think it is fair to argue that the Thelen research agenda could be divided into two broad parts. 19 Particularly in her work with Streeck, Thelen accounts for different forms of change. 20 In her work with Mahoney, she further develops the idea by adding types of "change agents" that stand behind these forms of change. 21 Five forms of gradual change have been identified by Thelen and her co-authors: displacement, layering, drift, conversion, and exhaustion. Displacement sees the gradual emergence of new models that call into question old ones; layering is the "active sponsorship of amendments, additions, or revisions to an existing set of institutions"; drift is the slippage of an institution; conversion is the redeployment of an old institution to a new purpose; and exhaustion refers to the overextension and gradual breakdown of an institution. 22 These five types aptly capture the broad variety of gradual change processes. Yet, these five processes are rather descriptive and procedural in nature and answer the question of how change unfolds, but not why these processes take place.
The respective actors behind (four of) these five processes are specified in a later work. 23 Layering processes are driven by subversive actors. The most conducive environment for the emergence of these actors is a weak institution with low discretionary leeway for the actors, but one in which the political context has strong veto powers. Displacement sees insurrectionaries as major actors that act also in weak institutions with a lot of discretionary leeway, but with rather weak outside veto points. Drift is driven by parasitic symbionts that act under a tight institution with low leeway and in a strong context with veto power. Lastly, Thelen and Mahoney assume opportunistic actors in a tight institution, but with a weak veto environment.

Endogenous or exogenous causes for democratic regression?
I take these insightful considerations of historical neo-institutionalism as a theoretical starting point. The different forms and actors of gradual processes that recent research has addressed give answers to the questions of how change takes place, by whom change is performed, and under what circumstances these developments are to be expected. I add to this insightful discussion the source of the cause, distinguishing between exogenous and endogenous causes.

Erosion and decay arguments
While the above discussed literature is usually interpreted as forms of endogenous gradual change, the distinction between endogenous and exogenous causes can already be read into it. 24 Layering, for example, is better to be understood if the change comes from without and is added onto an old institution. Similarly, exhaustion invokes an understanding of endogenous decay. When it comes to displacement, drift, and conversion, the picture becomes more complicated. Expressis verbis, Streeck and Thelen remain ambivalent about exogenous and endogenous drivers for these types of change. Displacement, for example, can happen either due to a "rediscovery or activation of previously suppressed or suspended possibilities" or due to an "invasion" that is characterized by a "supplanting of indigenous institutions and practices with foreign ones". 25 The former would be an example for endogenous gradual change, the latter for an exogenous one. In my language, the former is an endogenous decay that needs to be analytically separated from exogenous erosion.
Literally, endogenous and exogenous to an institution means that the cause for a process is not only located within or outside the institution under study, but that the cause is also generated within or outside an institution. This driver for change works so from the outside inor vice versa. I suggest distinguishing between internally generated causes and externally generated causes, the former associated with forms of decay, the latter associated with forms of erosion. Borrowing from natural sciences, I argue that erosion-type arguments use a different explanatory architecture than decaytype arguments. While erosion is a phenomenon in which wind or water hits a stone, making it porous over time and carrying gradually away its substance, decay is a phenomenon in which the seeds of its own destruction are already implanted in the radioactive chemical element itself. Compared to stable chemical elements, radioactive elements are inherently instable and the varying number of neutrons per nucleus determine the length of the respective half-lives. What is important to note here is that radioactive decay is an internally driven disintegration process, while erosion is set in motion by outside forces. 26 Needless to say, in the institutionalist literature, there is an abundance of definitions of what an institution actually is. These definitions range from very narrow and often formalized understandings to broader conceptions that include informal arrangements as well. 27 Douglass North has proposed one of the most widespread and comprehensive definitions when he famously argued that institutions are "the rules of the game" and the "humanly devised constraints that shape human behavior". 28 This definition is elastic enough to subsume various strands of institutionalist thinking. 29 For the study of democratic regressions, democratic regimes are usually perceived as a set of intertwined sub-institutions that range from formal institutions like party, electoral or representation systems, to informal ones that are usually unwritten and include diverse institutions like societal norms of for example fairness, forbearance, how to treat the political opposition, to shared behavioural equilibria, and business-government regulations. It is therefore the first and foremost analytical task of scholars to clearly define the borders of the institution or institutional ensemble that they actually study. Based on this decision, the core question in identifying endogenous decay processes is then to what extent the (interplay of sub-) institutions directly generate (unintended and often perverse) incentives, produce inner tensions, or only indirectly enable actions by leaving open regulatory loopholes and not closing gaps that actors can then exploit for subverting this very institution. 30 In contrast, democratic institutions can be washed out by outside forces, marking the opposite pole of erosion processes.

Erosion and decay arguments for democratic regression
This distinction between erosion and decay becomes important when ordering current accounts of democratic regression. I set forth that this major dividing line helps structuring our explanations for the empirical phenomenon of democratic regression.
Unfortunately, social science has already adopted a metaphorical language that does not pay sufficient attention to the analytical difference. 31 For example, the excellent conceptual discussion by Tom Gerald Daly insightfully condenses the current state of the art, but defines decaylike others did for erosionas the "incremental degradation of the structures and substance of liberal constitutional democracy", paying less attention to the source of the malaise. 32 At the end of the article, Daly even highlights two prominent research projects, a multi-university consortium "Democratic Erosion" and the "Democratic Decay (Dem-Dec)" project, 33 that are both equally committed to studying democratic regression processes. While these projects share the gradual unfolding of these processes, they do not focus on the question of endogenous or exogenous causation.
Yet, it seems that the erosion metaphor is currently prevailing. In her insightful article, Bermeo 34 states for example that "troubled democracies today are more likely to erode rather than to shatter". Ginsburg 35 discusses what needs to be done for an "anti-erosion jurisprudence", while Lieberman and colleagues 36 see an "erosion of democratic norms" in the United States. Diamond 37 describes the "significant erosion in electoral fairness, political pluralism, and civic space for opposition and dissent" in Turkey, and the V-Dem scholars 38 describe recent "democratic erosion processes" in, among others, the Czech Republic, Hungary, Poland, the United States, India, Bulgaria, and Brazil.
Despite the frequent usage of the word, real erosion arguments are rather rare in explaining democratic regression. 39 It might not be too far-fetched to say that its terminological usage is inversely related to actual erosion arguments. One example for exogenous causation is the modification of the idea of "linkage" and "leverage", originally put forward by Steven Levitsky and Lucan Way. 40 In an attempt to "turn the tables", the argument runs as follows: the more democratic regimes are linked to autocratic ones, via economic trade, but also communication channels, visits, and cultural affinities of any sort, the stronger the leverage of the autocratic regime to influence and autocratize the democratic regime. While Levitsky and Way have provided ample empirical evidence for the democratization perspective, the autocratization perspective still needs to be tested. As of now, there is no systematic empirical evidence that linkage and leverage actually work also in the other direction. But, beyond doubt, the linkageleverage-thesis is one that represents the most clear-cut erosion-type of explanation. The autocratic regime is like wind or water the outside force attacking the democratic substance. In Thelen and Streeck's descriptive terms, democratic institutions can be so displaced, converted, or be layered by active foreign sponsorship. Of course, Russia and China are discussed in this context as the most prominent of these foreign sponsors 41 or "gravity centers". 42 Yet, what is important to note here is that in all these processes, the cause is exogenous to the democratic institution that is targeted, marking so a clear-cut erosion process.
While the linkage-leverage hypothesis demands strategic action, this can be relaxed for a related argument. Value-based diffusion can be unintentional. 43 Yet, in parallel to the above, the diffusion of democratic norms is much more researched than the other way around. 44 The current state of the art is still rather cautious about the effect of autocratic diffusion. 45 Future research needs to fill this research lacuna and clarify to what extent an inherent attractiveness of authoritarian values, often catchily coined as the Chinese or Singaporean model, 46 an unintentional diffusion of authoritarian practices, or even the explicit learning from other's authoritarian example 47 explains a deterioration of democratic quality.
Please note that erosion arguments are exogenous arguments, which should not be equated with international arguments. While the linkage-leverage and the diffusion literature focus in their work on international factors, the exogenous force for erosion does not necessarily need to come from the international arena. It can also be bred domestically. The major dividing line is as to whether the cause is generated within or outside the democratic institutions that one studies. To name just a few instances of domestic erosion, an anti-democratic influence of the national military forces, an overly conservative Catholic Church, strong national Islamist traditions, or deep ethno-political cleavages can also constitute domestic exogenous causes.
If taking a historical perspective, exogeneity can also stem from a different angle. If for example the timing, a certain "populist Zeitgeist", 48 is taken for explanatory purposes and not only as a descriptor of an era in which we live in, it constitutes also rather an exogenous than an endogenous factor for explaining democratic regression. Yet, more empirical research is needed here as well that demonstrates how timing actually develops causal power within domestic democratic regression processes, let alone how contagious it is across countries. 49 Moreover, if we adopt an explanatory perspective in which we treat the past as exogenous to our institutions under study, then today's democratic regressions can be explained by previous low-quality democratizations. These democratizations suffered particularly from economic and social inequality and weak political structures, providing already a fertile ground for democratic regressions. Adopting such a perspective means to treat this history as an exogenous cause for today's situation. 50 The causes for today's deterioration of democratic quality are here generated outside (i.e. temporally prior to) the (current) institutions under study. 51 Take for example the case of the Philippines. It is argued here that the former military dictator Ferdinand Marcos established a system of patronage, corruption, cronyism and wide-spread pork-barrel politics in which political institutions were destroyed or severely weakened. 52 This authoritarian shadow that is still tangible until today continues to erode current democratic substance, both in terms of political culture and government institutions. Hutchcroft and Rocamora 53 aptly summarize that in the post-Marcos era "the logic of patronage remains central to understanding Philippine politics, and political parties remain weak, ill-defined, and poorly institutionalized".
In turn, rivalling arguments are of endogenous nature. Democratic regression happens often within democratic core institutions that have been initially built to guarantee horizontal and vertical accountability. The arguably most influential work today on this topic, the book on "How Democracies Die" by Steven Levitsky and Daniel Ziblatt,54 uses the classic and seminal work by Juan Linz 55 which can also be read more in terms of endogenous than exogenous causation. Linz suggested a list of four key indicators. First, Linz spoke about a weak commitment to the democratic "rules of the game", as evidenced in undermining the legitimacy of elections, taking extraconstitutional means like military involvement in politics or violent insurrections. A second warning signal refers to the denial of the legitimacy of the political opponent and to what extent they are defamed as criminal, subversive, foreign agents or constituting a threat to national security. A third aspect is toleration or even an encouragement of violence by actors close to the incumbent. A final dimension of authoritarian behaviour in democratic politics is the willingness and readiness to restrict civil liberties of political opponents, including press freedom, freedom of movement, and freedom of association. These indicators by Linz can be used to identify countries that are at high risk or that already entered democratic regression. What is important to note here is that these causes are rather generated within than outside the democratic institutions.
Another prominent argument of inner tensions within the institutional design of democracies has lately been put forward by Dan Slater and Aries Arugay. 56 They argueusing Thailand, the Philippines, Indonesia, and Taiwan as empirical examples that democratic institutions demand vertical accountability and horizontal accountability at the same time, thereby following simultaneously the conflictive aim of inclusion and constraint, causing internal institutional conflict that sparks institutional polarization. In a similar vein, Milan Svolik 57 compellingly demonstrates that the causes for democratic regression "lies in a vulnerability that is inherent to democratic politics." He develops an argument in which authoritarian populists exploit a dilemma that is ingrained into the democratic institution of elections. Elections provide voters with two choices that can be conflictive, particularly in countries with deep political cleavages and acute tensions: democratic principles on the one hand and partisan interests, embodied by populist leaders, on the other hand. Using experimental methods, Svolik 58 shows for Turkey, Venezuela, and the United States that a significant portion of the population "act as partisans first and democrats only second", ranking partisan interests higher than democratic principles. Like for the Slater and Arugay study, the cause of democratic regression is seen by Svolik in endogenous tensions rather than in exogenously driven processes.
The relevance of endogenous explanations of democratic regression are bolstered by empirical evidence. Svolik 59 argues that democratic subversions are either caused by military coups or executive turnovers. In his account, between 1973 and 2018 88 of 197 cases of democratic declines have been caused by executive takeovers, while only 46 cases can be attributed to military involvement into civilian politics. The time trend is clear: Since 2005 four out of five democratic regressions are caused by democratically elected incumbents that abuse their power. And one of their most important tools for undermining democracy isparadoxicallylaw and the liberal democratic constitution. 60 Kim Lane Scheppele 61 captivatingly demonstrates in this context how "legalistic autocrats" exploited the inherent tension between a (vulgarized understanding of) democracy and constitutionalism to the detriment of liberal values.
To sum up, erosion arguments, if taken seriously, refer to a cause that is generated outside the institution under study. Despite the high usage of the term, this type of argument seems to be rather rare. Linkage-leverage-arguments arguments are prototypical for such an endeavour. In contrast, decay arguments about democratic regression that focus on endogenous tensions within democratic designs seem to be overall more widespread. This might be open to debate, but the distinction between the locus of the cause should be made. The source of the malaise needs to be located in the first place. This is not only for the sake of analytical clarity but might have also important repercussions on advising political praxis.

Zooming into the causal mechanism of democratic regression
The locating of a cause has direct implications for detecting causal mechanisms. In an admittedly very simplified version, an explanation of a phenomenon includes causes X, a causal mechanism M, and an outcome Y (X → M → Y). In the previous section, I have introduced the major distinction between endogenous and exogenous causes. This section is dedicated to detailing the causal mechanism M. The set of causes X and their inherent properties impact the working of the causal mechanism. Yet, while properties of the cause influence the mechanism, both should be kept separate from each other for analytical purposes.
Let me first clarify my understanding of a causal mechanism. It goes almost without saying that its meaning is hotly contested. In an insightful review article, John Gerring 62 lists at least nine different meanings of what a causal mechanism can be. Similarly, James Mahoney 63 lists 24 slightly varying definitions by 21 authors. If there is any common ground between these differing meanings, then I would follow Gerring's suggestion and see it in a "pathway or process by which an effect is produced". 64 It seems to be intuitive that a mechanistic explanation focuses on what happens between X and Y. In close analogy to physical mechanics, Hedström and Ylikoski 65 suggest therefore to think of the "cogs and wheels (…) through which the outcome to be explained was brought about". Taking this analogy further, the idea of disentangling the complex processes of democratic regression is to dissect the bits and pieces in order to see how they interrelate and work together. In analytical terms, this is the task of a clockmaker who opens the clock case in order to see how the cogs and wheels interact.
As a further general caveat in zooming into the study of causal mechanisms, their generative and repeated nature need to be emphasized. Both attributes are highlighted by Renate Mayntz. 66 A causal mechanism should generate the outcome and not remain as a mere concurrent epiphenomenon. While this seems to be less controversial for democratic regression mechanisms, she pays emphasis that a mechanism invokes notions of regularity and should not be limited to idiographic case-specific explanations. For her, a causal mechanism should be, at least in principle, transportable from one specific context to another. A causal mechanism should produce the same outcome when specified initial circumstances are met. The recurrent nature of regression mechanism needs to be empirically demonstrated. This condition might meet more scepticism in the case-oriented literature, but nevertheless remains a useful thought exercise for abstracting from the specific for the sake of the general. 67 Components of a causal mechanism I follow a substantive understanding of causal mechanisms as complex systems of interlocking and interacting parts. In philosophy of science, a widely used definition of causal mechanism proves to be helpful here. A causal mechanism consists of "entities and activities organized such that they are productive of regular change from startup to finish or termination of condition". 68 Activities are understood as the "producer of change" while entities are "the things that engage in activities". 69 This dualistic understanding and the explicit distinction between activities and entities is crucial here. 70 Moreover, the analytical sociological tradition would add that we also need to look for actor-centered micro-foundations of the macro-relationship. 71 When theorizing processes of democratic regression, explicitly clarifying between (micro-)entities and their related activities of a causal mechanism constitutes an important empirical step.
As argued above, the properties of the cause X do impact the causal mechanism M and the way it is studied. It is straightforward to argue that endogenous or exogenous causation makes us look at different locales. It is like marking a cross on a map that tells us where to dig deeper. When digging deeper into a decay process, the focus of our search for entities and activities lies automatically within the institution under study. Take the example of the autocratic legalism argument that I mentioned above. 72 It is endogenously anchored. When detailing the causal mechanism behind this argument, the place to look for lies in inherent institutional weaknesses and loopholes that make it possible, sometimes even incentivize, formally democratically elected leaders to non-use, ab-use, and/or mis-use existing, but dormant laws to the detriment of liberal democracy. While in the legalism mechanism the "non-, ab-and mis-use" of laws refers so to the concrete activities, the involved entities are the governments that instrumentalize lower-level courts as well as supreme and constitutional courts and reform them in such ways that like-minded judges are appointed, and that critical voices are suppressed, seizing incrementally power over courts, undermining judicial independence in the alleged name of the people, curtailing then the rights of the opposition, and violating subsequently civil liberties and political rights.
Consider the empirical example of South Korea under Lee Myung-bak and Park Geun-hye. 73 It illustrates the inner working of the "autocratic legalism" mechanism that Scheppele 74 and Corrales 75 showed for Hungary and Venezuela, i.e. the gradual conversion of existing laws for illiberal purposes. Stephen Haggard and Jong-Sung You 76 argue for example that the South Korean democratic regression lies in the conjunction of several legal tactics by the government. In this context, they refer to the abuse of criminal defamation laws, to abuse of election laws, and the abuse of the National Security Law. Using the language of historical institutionalism, this line of argument can be subsumed under endogenous displacement in which an insurrectionary type of change agents activates "previously suppressed or suspended possibilities", 77 being enabled by a weak institutional context. 78 Under the presidency of Lee, it is shown that the number of indictments for defamation doubled, often involving massive attacks on the freedom of expression. 79 Also, electoral laws were bended, ranging from illegal use of money and gifts to illegal propaganda, and violations of neutrality orders. 80 While the National Security Law that entails the prohibition of so-called "anti-government organizations" has been in great decline during the liberal governments between 1998 and 2007, it saw a sharp increase when Lee entered office. 81 The South Korean case shares many similarities with the Indonesian democratic regression episode between 2007 and 2014, exhibiting the recurrent nature of the autocratic legalism mechanism. Warburton and Aspinall 82 speak aptly of an "illiberal drift". They argue that not only "draconian laws on defamation and blasphemy" 83 have been used in more instances, but that the Indonesian government also relied on the 2013 Law on Societal Organizations ("NGO Law") that prohibit, sanction, and censor non-governmental organizations critical of the state doctrine Pancasila. Also like the South Korean example, the Indonesian democratic regression is explained by "electoral narrowing" 84 in which electoral laws have been used by the incumbents to raise the hurdles for competition thresholds, nomination procedures, and financing and campaigning, transforming Indonesian democracy "into a system of intra-incumbency contestation that excluded other forces trying to penetrate it". 85 These policies can be adequately subsumed under an endogenous causal mechanism of autocratic legalism.
In contrast, when exogenous erosion processes are traced, scholars need to concentrate on external actors outside the institution under study. The research focus should lie then on the strength of the external forces that work from the outside in. As such, erosion arguments seem to be generally more uni-directional, depending largely on the magnitude and intensity of the outside force that gradually weakens democratic substance. In turn, decay arguments are characterized by internal dynamics that often involve self-reinforcing feedback loops. As such, the source of the cause does not only tell us where to search for entities and activities for spelling out the causal mechanism, but also give us hints to probable inner workings, highlighting already implicit temporalities that should be made explicit.

Accounting for temporality in causal mechanisms
Beyond dissecting a causal mechanism into its components and demonstrating its recurrent nature, temporality is a crucial, but often underrated feature of a causal mechanism in generaland of democratic regression mechanisms in particular. 86 While the dissecting into components of a causal mechanism is often at least implicitly done, the temporality question remains underspecified in the current literature. Temporality covers the tempo of a process, as well as its duration, acceleration, sequencing, and timing. In her admirable article, Anna Grzymala-Busse 87 not only clarified the differences between these temporal dimensions, but also gave concrete advice on how to operationalize them and how to methodically approach them.
In most empirical instances, democratic regression is a slow process (tempo) that takes a long time to develop (duration). Usually, democratic regression isalready in its namea process in which democratic substance is gradually weakened or hollowed out over a longer period of time. While we might agree that these parameters are often shared, they come with considerable empirical complications. Duration covers the temporal length of how long a process operates, necessitating so not only the identification of a clear starting and end point, but also establishing what Grzymala-Busse 88 calls a "temporal baseline". We do not only need to justify the boundaries of an event or a process, but also how we temporally relate this process to other processes. Do we argue in terms of objective time, measured for example in months or years, or do we relate it to the length of previous episodes? Moreover, what counts as democratic regression, given that this is a complex phenomenon consisting of several temporally overlapping sub-phenomena? Also, how do we deal with phenomena that slowly accumulate over time, remaining under the radar of both academia and praxis, before they finally enter a threshold and become visible? How should we classify these invisible processes? 89 In the introduction to this special issue on democratic regression in Asia-Pacific, Croissant and Haynes 90 propose for example a 10% decline in V-Dem's Electoral Democracy Index and V-Dem's Liberal Democracy Index to define a democratic regression. The starting point of the episode is marked by such a drop and is interrupted when either a stagnation is observed over four years or when a substantial increase is observed. This means that Croissant and Haynes use an objective time (years) and do not relate it to the length of previous episodes. Also, it should be noted that the indices that Croissant and Haynes use are situated on a rather high aggregation level in which profound changes have already manifested themselves in the respective countries. This should not be misunderstood as a critique of their approach, but rather as an attempt to make the implicit explicit.
Justifying decisions for delineating the respective regression episodes is not a trivial task that gets even more complicated when we consider tempo. Tempo of democratic regression can be defined as the amount of decline in democratic quality per time period (duration). While tempo (fast and slow) should not be conflated with duration (short and long), it should also be kept separate from arguments about acceleration. Acceleration is change of tempo per time unit, taking into account the nonlinear nature of democratic regression processes. The (unexpected) increase in tempo might not only bring about turbulence, but also fosters situational logics in which the emergence of new actors or the sudden re-shuffling of power asymmetries in actor constellations is catalysed. 91 While it still needs to be systematically tested, a hypothesis could be formulated: Democratic regressions that can be traced back to exogenous erosion remain rather on a constant tempo as the exogenous effect often remains invariant over time. In turn, decay processes are more inclined towards internal dynamics and self-reinforcing feedback loops, suggesting acceleration processes.
Finally, sequencing and timing are two further crucial aspects of causal mechanisms. 92 To my reading, they remain underexplored, but should be on the future research agenda on democratic regression as well. A sequence is nothing more than an "ordered list of elements". 93 Yet, for crafting convincing causal sequencing arguments about democratic regressions, it needs to be shown that it is the order of events that actually matters. In other words, the burden of proof lies in empirically demonstrating that a sequence of X 1 → X 2 → X 3 leads to the outcome Y, but a sequence of say X 1 → X 3 → X 2 would not. It is the order of events that develops causal power. I am not aware of empirical case studies that focus on this type of explanation. Yet, it seems to be a promising route, particularly for an endogenous type of explanation.
In contrast to sequence arguments, timing does not refer to the relative ordering of events within a sequence, but it is anchored in an exogenous context. For timing arguments, it is the change of the context that matters. It is the temporal circumstances that cause a process. Applied to democratic regressions, the Zeitgeist argument that I mentioned above illustrates this thinking. 94 Yet, again, to my reading of the state of the art, this is an underexplored topic that should be addressed more systematically. While we might all agree that we recently observe a clustering of democratic regressions, it still needs to be shown that the timing does have an independent causal impact on (or at least serve as a catalyst or facilitator for) these phenomena, both being effective within a country and contagious across countries.
To sum up, when zooming into M, the burden of empirical proof lies (a) in demonstrating the generative and (in principle) recurrent nature of the causal mechanisms; (b) in decomposing the causal mechanism into its entities and activities and potentially accounting for micro-level foundations; and (c) in integrating previously underexplored aspects of temporality that range from duration, tempo, and acceleration to issues of timing and sequencing.

Conclusion
This article is written in light of the current rise of democratic regression phenomena. It seeks to classify existing approaches and explanations in a novel way, pledging on the one hand for analytical clarity and synthesizing on the other hand their insights into a "checklist" how we could methodologically approach democratic regressions. While I maintain that democratic regression processes are not only multi-faceted, but also multi-causal, I argue that we should even be more analytically precise and dissect the complex phenomenon into its constitutive parts. Three points are central.
First, I seek to establish a new axis along which we can classify existing accounts. Using historical neo-institutionalism as a theoretical starting point, the crucial dimension is the source where the cause for democratic regression originates. I broadly distinguish between endogenous and exogenous explanations to organize and categorize existing explanatory factors. To illustrate the different logics, I use the natural science metaphors of endogenous decay and exogenous erosion.
Second, I illustrate these considerations with selected secondary literature and empirical cases. As such, I argue that despite the frequent usage of the term "erosion", decay arguments are not only more widespread, but also more convincing. Exogenous causes that are found in the literature employ often a reversed linkageleverage argument. In contrast, endogenous causes argue more with inherent institutional tensions, be it between conflictive goals of inclusion and constraint or between democratic and partisan principles.
Third, I argue that it is key for analytical precision to not only distinguish the sources of the cause X, but separate the cause X from the causal mechanism M. I do not question that X and M are interconnected and that properties of X influence the inner working of M. To the contrary, I argue that we should take advantage of this interconnectedness as the source of the cause gives us an orientation where to dig deeper. It is the marking on a map that shows us where to dig deeper. The study of causal mechanism details the concrete pathway by which the cause leads to the outcome, with its preeminent task lying in decomposing its constitutive components, the involved activities and entities, and in accounting for varying temporalities.
While the current state of the art provides ample empirical material for dissecting the mechanisms into its components, the temporal dimension seems to be still underexplored. When systematically detailing the inner workings of a causal mechanism, basal temporal dimensions of duration, tempo and acceleration, but also of timing and sequencing should be made more explicit. They constitute a useful toolbox for refining our causal mechanisms with which we seek to explain democratic regression. Yet, as much as future research could pick up these currently neglected dimensions, it needs to be seen to what extent the classifications that I propose here proves to be helpful guidance in ordering a fuzzy field. Notes 9. In his work on democratic breakdowns, Maeda also distinguishes between endogenous and exogenous termination of democratic rule, the latter being caused by democratically elected incumbents, the latter being caused by forces outside the democratic government. I share with him the emphasis on the source of the cause, but rather look at gradual forms of change instead of breakdowns, and expand the distinction between endogenous and exogenous causes to institutions more generally. See Maeda, "Two Modes of Democratic Breakdown". 10. Streeck and Thelen, "Introduction"; Mahoney and Thelen, "A Theory of Gradual Institutional Change". 11. Fioretos, Falleti and Sheingate, The Oxford Handbook of Historical Institutionalism. 12. Immergut, "The Theoretical Core of the New Institutionalism". 13. Collier and Collier, Shaping the Political Arena; Mahoney, The Legacies of Liberalism. 14. Pierson, "Increasing Returns, Path Dependence, and the Study of Politics"; Arthur, Increasing Returns and Path Dependence in the Economy; Arrow, "Increasing Returns". Theoretical Core of the New Institutionalism"; Hall, "Politics as a Process Structured in Space and Time". 28. North, Institutions, Institutional Change and Economic Performance, 3. 29. For some scholars, the wide usage of "institutions" might seem inadequate. Some might prefer concepts like system, structure, field, or regime as a bounded whole. Yet, to make the argument about endogenous and exogenous causation, it suffices to establish a border of what counts as within and what counts as outside of the entity. 30. For example, Maeda skillfully analyses the impact of presidential vs parliamentary systems (the institution) on democratic breakdown. Drawing on the classic debate initiated by Linz in the 1990s, he empirically demonstrates that there must be something within the institution of the presidential system (polarization, winner-takes-all-attitude, deadlock, rigidity, dual legitimacy etc.) that generates perverse incentives so that "presidents in presidential systems are more likely to become authoritarian than prime ministers in parliamentary systems", Maeda, "Two Modes of Democratic Breakdown", 1141. 31. A remarkable conceptual discussion can also be found in Sadurski's recent book on "anti-constitutional populist backsliding" in Poland. Here, Sadurski criticizes the term "decay" for being too passive and actorless, neglecting the enthusiasm and energy that key actors have invested to dismantle the constitutional state. I agree with Sadurski here that decayas well as erosioncould be understood in this way. Yet, I would like to alleviate these concerns. Moreover, it might be particularly the study of causal mechanisms that re-focuses again on the centrality of actors. See Sadurski, Poland's Constitutional Breakdown, 12-4. 32. Daly, "Democratic Decay", 17.