Going technocratic? Diluting governing responsibility in electorally turbulent times

Abstract Technocracy has recently triggered growing scholarly interest, especially as an alternative form of ruling to both party government and populism. In the context of weakened parties-citizens links and increasing external constraints faced by Western European ruling parties, technocratic appointments might help deal with the responsibility-responsiveness dilemma highlighted by Peter Mair. However, research on the explanatory factors of technocratic appointments is still underdeveloped. This article argues that the recourse to technocracy is fuelled by electoral volatility. In contexts of high electoral turbulence – and even more when parties frequently enter or exit the party system – ruling parties turn to technocratic appointments to dilute responsibility. This expectation is tested through an original longitudinal multilevel dataset including 655 cabinets and 373 elections in 20 Western European countries from 1945 to 2021. The findings of this article contribute to the current debate on technocracy and shed new light on the general understanding behind political representation.

2018), and a voluntary reduction of the magnitude of the so-called partyness of government (Katz 1987) has been very unusual.
Things have begun to change in the past few decades, when the nature of party government has wavered (Andeweg 2000). The decline of partisanship of ministerial elites has been variously addressed, and technocrats have become more and more relevant both quantitatively within the ruling class and qualitatively in the sphere of government (Mair 2008).
Over the last decades, scholars have increasingly paid attention to the role that technocracy has played in the transformation of democracy (Caramani 2017;Mair 2013). This phenomenon has been studied by looking at several dimensions, including technocratic discourse and style of policy making (Bertsou and Caramani 2020a), social characteristics of the technocratic elite (Costa Pinto et al. 2018), and representation patterns vis-à-vis the party government model and implications for democracy (McDonnell and Valbruzzi 2014;Pastorella 2016).
However, relatively little investigation has been devoted to the determinants of technocratic appointments, with a privileged focus on economic and institutional determinants (Alexiadou and Gunaydin 2019;Amorim Neto and Strøm 2006;Schleiter and Morgan-Jones 2009;Wratil and Pastorella 2018). This article offers a novel perspective in explaining the recourse to technocracy in Western European governments. In the context of weakened parties-citizens links and increasing external constraints faced by Western European ruling parties, technocratic appointments might be a party strategy to cope with the dilemma highlighted by Peter Mair (2009Mair ( , 2013 between responsibility and responsiveness: the former means that ruling parties act prudently within the constraints provided by external agencies and institutions, while the latter means that parties act following voters' demands (see also Karremans and Lefkofridi 2020;Lefkofridi and Nezi 2020). Indeed, forced to be responsible towards supranational institutions and globalised markets, ruling parties are increasingly perceived by citizens as unresponsive to voters' demands. Thus, in a desperate attempt to bridge the gap between responsibility and responsiveness, ruling parties face a critical trade-off: holding all the governing positions and, therefore, assuming the risk of being electorally punished at the ballot box or 'diluting responsibility' by ceding government posts to technocrats which, unlike politicians, are not subject to democratic accountability (Cotta 2018).
Within this framework, we argue that the turbulence of the electoral environment defines the odds of parties in the responsibility-responsiveness bet. While in electorally stable contexts parties will feel relatively safe in handling responsibility on their own, as voters are relatively inelastic to governments' actions, as the context becomes electorally more unstable, the risk of being punished by voters for their responsibility-responsiveness gap increases. Therefore, following Friedrich's (1963) rule of anticipated reactions, ruling parties will increasingly recur to technocrats as a strategy to dilute responsibility and avoid paying a cost at the ballot box.
We test this argument on an original longitudinal multilevel dataset including 655 cabinets and 373 elections in 20 Western European countries from 1945 to 2021. The article finds that the recourse to technocracy in Western European governments is fuelled by electoral volatility. In contexts of higher electoral turbulence, ruling parties are more likely to turn to technocratic appointments. Moreover, the analysis also tests alternative hypotheses related to the specific source of the link between volatility and technocracy: the key driver of this mechanism is the unpredictability brought about by the disappearance of old parties and the emergence of new ones in the party system.
The article is structured as follows: the next section reviews the literature on the recourse to technocracy and presents our argument about the link between electoral volatility and the recourse to technocracy in Western European governments. The third section introduces the data and the method employed in the empirical analysis. The fourth section shows the results of the analysis and discusses the related findings. A concluding section follows.

Theoretical framework and hypotheses
Political science has recurrently paid attention to the relationship between technocracy and party government, with the former depicted as an alternative to both party government and populism: technocrats' democratic legitimation lies in the 'trustee' dimension of representation, while populism operates according to the 'delegation' model (Caramani 2017). Both governing configurations can be parsed as party government's correctives or even challenges (Caramani 2017;Mudde and Rovira Kaltwasser 2012). 1 Recently, scholars have also devoted attention to the aspects involving the electoral market's demand side, such as technocratic attitudes (Bertsou and Caramani 2020b;Bertsou and Pastorella 2017;Lavezzolo et al. 2021), the support for stealth democracy (Lavezzolo and Ramiro 2018), and public opinion favour for experts in government (Ganuza and Font 2020).
So far, however, the contextual drivers of technocratic appointments have been generally understudied. Within this framework, two strands of research have received attention. On the one hand, the impact of economic conditions on the recourse to technocrats in European democracies (Alexiadou and Gunaydin 2019;Brunclík and Parízek 2019;Semenova 2020;Wratil and Pastorella 2018). More specifically, it has been shown that economic recessions boost technocratic-led governments because parties can preserve their credibility without being affected by potentially treacherous crises (Wratil and Pastorella 2018). Moreover, some scholars have argued that political systems' poor economic performance and citizens' distrust exacerbate technocratic cabinets' occurrence (Brunclík and Parízek 2019), and economic crises drive technocrats' appointments in Western European cabinets' top economic portfolios (Alexiadou and Gunaydin 2019). Also, in Central and Eastern European countries, technocrats have high chances of being involved in cabinets during an economic downturn (Semenova 2020).
On the other hand, some have stressed the influence of the institutional setting. Notably, presidential or semi-presidential regimes, where the directly elected President enjoys autonomous political legitimacy and holds substantive powers in government formation, enhancing non-partisan ministers' selection compared to parliamentary regimes (Amorim Neto and Strøm 2006). In such a system, the monocratic figure's dominance is accentuated, so the chief executive's delegation would prevail over parliamentary parties' control (Cotta 2018). Contrarily, government formation in parliamentary regimes is dominated by parties' influence over government posts (Saalfeld 2000). Recent evidence also suggests that presidential power, parliamentary fragmentation, and the cabinet's caretaker status significantly increase the share of non-partisan ministers in semi-presidential cabinets (Schleiter and Morgan-Jones 2009).
Despite their alleged importance, such factors cannot provide a systematic and compelling explanation: the recourse to technocrats is not limited to presidential or semi-presidential systems or periods of economic downturn. Additionally, although several European countries experienced stagnation or recession during the post-2008 period, technocrats have not replaced partisan cabinets (Cotta 2018). Thus, perhaps, technocratic appointments can be better understood from a different perspective.
In our view, returning to technocracy might be a party strategy to cope with the responsibility-responsiveness dilemma. According to Mair (2009: 12), responsibility means that ruling parties face relevant constraints due to external agencies and institutions, while responsiveness means that parties are more focussed 'on input-oriented legitimacy' . Such tension became apparent over time, putting parties under formidable pressure (Bardi et al. 2014). In the past, political parties governed by simultaneously meeting both responsibility and responsiveness. However, recently, due to party cartelisation (Katz andMair 1995, 2018), parties have reinforced their governmental role without preserving the representative one. Indeed, the responsibility-responsiveness dilemma becomes more severe when parties have to deal with constraints deriving from multilevel governance (Lefkofridi and Nezi 2020), as the Eurozone crisis has made clear (Karremans and Lefkofridi 2020). Furthermore, the increase in the number of veto and semi-veto players in the multilevel institutional setting progressively requires governments to act prudently and follow shared procedural norms and practices. In such circumstances, government responsibility implies that ruling parties' policy agenda will be constrained, and agents (i.e. parties) are forced into behaviour that neither they nor their domestic principal (i.e. voters) would have freely chosen (Mair 2009;Strøm 2003). In a nutshell, in contemporary Europe, 'the demands for responsiveness and the demands for responsibility [are] increasingly at odds with one another' (Mair 2009: 10).
The increasing dependence of Western European countries on external actors means ruling parties are less free to choose and implement public policies: they face a trade-off about when to pay the cost of being responsible vis-à-vis international and supranational institutions. Ruling parties can choose either to pay such a cost later -by holding all the governing positions and risking being electorally punished in electoral ballots -or immediately -by renouncing some power and ceding government posts to technocrats (Cotta 2018). 2 In other words, opting for the latter strategy can be seen as a compromise between their intrinsic office-seeking nature and the attempt to avoid the 'cost of ruling' (Narud and Valen 2008;Paldam 1986), namely, the electoral losses of government parties because of their activity in office.
This article makes a step forwards in clarifying the mechanism linking the responsibility-responsiveness gap to the increasing appointment of technocrats in Western Europe: we argue that it is the electoral context playing the lion's share in this story. 3 More specifically, the electoral context defines parties' odds in the abovementioned trade-off (paying a cost immediately by ceding power vs. paying it later for holding responsibility).
In electorally stable environments, voters' response is expected to be rather inelastic to governments' political outputs. In such contexts, parties will feel relatively safe in handling responsibility (i.e. keeping governing positions for themselves), as their alleged lack of responsiveness is less likely to get sanctioned at the ballot box. Conversely, in electorally 'turbulent times' (Chiaramonte and Emanuele 2019), namely electorally volatile contexts, the voters' response is expected to be elastic to a government's actions. 4 In line with Pedersen (1979) and Bartolini and Mair (1990), volatility is a sign of party system change and captures the degree to which the interactions between parties and voters become unstable and unpredictable. A volatile environment is, therefore, a context where previously established power positions are at risk, as voters tend to disconfirm their previous voting choice election after election. This is likely to occur because volatility can be seen as the 'behavioral translation' of a negative attitude of voters towards parties. Moreover, many studies underline that electoral volatility is a sign of voters' distrust of political parties (Dassonneville 2012;Söderlund 2008;Zelle 1995).
Our argument relies on the assumption that parties and voters act under a bounded rationality framework: similar to retrospective voting theory (Fiorina 1981), voters are expected to be able to assess parties' performance in office and attribute individual merits and blames. Their assessment is far from being perfectly rational and they make mistakes due to their political biases and lack of information (Healy and Malhotra 2013). However, what really matters for our argument is not that voters are able carefully to consider who is responsible for what, but that parties attribute credibility to this scenario. So, according to Friedrich's (1963) rule of anticipated reactions, parties will behave by trying to anticipate voters' future actions. Consequently, in a volatile environment, parties will likely prefer to 'dilute responsibility' by ceding governing positions to technocrats. Appointing technocrats is helpful because it shifts responsibility from distrusted parties to the hands of experts who are objectively qualified for the job and esteemed by international markets and supranational institutions (Alexiadou et al. 2022). This eventually allows parties to make technocrats appear as the scapegoat for the unresponsive policies of the government and therefore reduce the related electoral cost at the ballot box. Thus, our main hypothesis is the following: H1. The higher the electoral volatility, the higher the share of technocrats in governments Our argument is schematically summarised in Figure 1. Besides the general explanation outlined in H1, we also want to take a step forward in outlining the specific driver of technocratic involvement in governments. To do so, we disentangle the general concept of electoral volatility (Bartolini and Mair 1990;Pedersen 1979) into some internal components. The first one is the volatility accounted for by the electoral shifts between government and opposition parties. The second one captures the electoral shifts among established parties, while the third one focuses on the change produced by the birth of new parties and the death of old ones.
Therefore, we propose a corollary to the general hypothesis (H1). Three competing hypotheses originate from such a corollary, each emphasising the role of one of the components described above. The first and probably the most straightforward hypothesis is that the critical driver of parties' decision to hand over responsibility to technocrats is the extent to which voters move across the government-opposition dividing line. In those contexts where a large part of the electorate is available to head towards the punishment or the reward of governing parties (rather than choosing their preferred party only within the government or the opposition camp), such parties will be highly sensitive to pursue the abovementioned strategy of diluting responsibility. 5 An alternative hypothesis postulates that the crucial determinant behind ruling parties' choice to transfer responsibility to technocrats is the availability of voters to switch among all established competitors (rather than only between government and opposition parties), thus significantly altering the balance of power among established parties across elections.
Finally, a third driving mechanism can be imagined. Indeed, we may hypothesise that the driving force leading to technocrats' appointment is the availability of voters to move -rather than across the government-opposition line or among established competitors -away from old competitors and towards new ones. In other words, what matters is the unpredictability of the system brought about by the continuous disappearance of old parties and the emergence of new ones, an increasingly relevant phenomenon in Western European democracies (Chiaramonte and Emanuele 2017). This translates into a strong sense of vulnerability for ruling parties because actors and their interactions change, thus making the environment riskier and riskier. All in all, our three competing hypotheses H1a, H1b, and H1c are as follows: H1a. The higher the electoral volatility across the government-opposition dividing line, the higher the share of technocrats in governments H1b. The higher the electoral volatility among established parties, the higher the share of technocrats in governments H1c. The higher the electoral volatility due to the entry of new parties and the exit of old parties, the higher the share of technocrats in governments

Data and method
The empirical analysis starts from the operationalisation of the dependent variable. Unlike other studies focussing exclusively on fully technocratic governments (McDonnell and Valbruzzi 2014) or technocratic-led cabinets (Wratil and Pastorella 2018), we consider the overall weight of technocracy in Western European governments. The advantage of this operationalisation is that we can have a more comprehensive picture of the recourse to technocratic appointments. Indeed, a relevant phenomenon in contemporary democracies is not merely that certain governments are entirely made of technocrats, but that these latter hold governmental posts in political governments (trivially, governments led by politicians) (Amorim Neto and Strøm 2006). Moreover, our operationalisation allows having a more fine-grained assessment of the power of technocrats, compared to a dichotomous operationalisation dividing between governments with and without technocrats' presence (e.g. Semenova 2020).
Our dependent variable, Share of technocratic positions, is therefore measured as the share of ministries held by technocrats in a given cabinet at the time of government formation. Technocrats are defined as non-partisan ministers, namely ministers with no political affiliation. 6 Out of 698 governments formed between 1945 and 2021 in 20 Western European countries, the average of Share of technocratic positions is 7.8 percent. The Share of technocratic positions is on the rise in Western Europe: in the 2010s, the average share of technocrats is between two and three times higher than in the 1980s, 1990s, and 2000s (see Figures  A1 and A2 in the online appendix, supplementary materials). This finding is in line with our interpretation that follows Mair's responsibility-responsiveness dilemma (Mair 2009). Therefore, technocratic appointments can be seen as a way to dilute responsibility in times characterised by increasing international constraints and weaker parties (Mair 2008;Strøm 2003). In terms of cross-national variations, the appointment of technocrats is comparatively more frequent in Southern Europe but is also a relevant phenomenon in other national contexts (e.g. Austria, Finland, and France) (see Figure A3 in the online appendix).
Turning to the independent variables, to capture the extent to which Western European party systems are embedded in a turbulent electoral environment (H1), we use the Pedersen (1979) index of electoral volatility. Therefore, our focal predictor is Total Volatility (TV), measured as the net aggregate electoral shifts between two consecutive parliamentary elections. Moreover, we need to disentangle the index of TV into its internal components to test H1a, H1b, and H1c. The first one is the electoral volatility due to electoral shifts between government and opposition parties. For each legislature, we calculate the absolute difference in the aggregate vote share at time t and t-1 received by the parties forming the last government of the previous legislature. 7 It is a bloc volatility measure (Bartolini and Mair 1990) where the identity of parties composing the two blocs (government and opposition) changes across legislatures. This measure, by definition, can range between 0 (all electoral shifts take place within the government and opposition parties, respectively) and TV (all volatility is due to electoral shifts between government and opposition parties) but can never exceed it. We call this measure Government-Opposition Volatility (GopV).
The second and third internal components of volatility are instead complementary, as their sum equals TV. One measures the electoral shifts among established parties, namely, those parties that are part of the party system in the consecutive elections considered to calculate volatility. The other one captures the electoral volatility due to the entry of new parties into the party system or the exit of old parties from the party system. In the last years, a growing strand of literature has investigated such internal components of electoral volatility, offering slightly different names and operationalizations (e.g. Birch 2003;Mainwaring et al. 2017;Powell and Tucker 2014;Rose and Munro 2009). We focus on the measures by Emanuele (2017, 2022), who distinguish between Alteration Volatility (AltV) and Regeneration Volatility (RegV), respectively. Data on electoral volatility measures come from Emanuele (2015).
Our empirical analysis also controls for many potential alternative predictors. In particular, following the literature, we consider four sets of controls related to cabinet characteristics, institutional constraints, party system structure, and further socio-economic crisis-related indicators.
As for cabinet characteristics, we include the Type of cabinet, operationalised as a categorical variable, differentiating among multi-party minority, single-party minority, oversized coalitions, minimal-winning coalitions, and single-party majority governments (Müller and Strøm 2000). We included the type of cabinet as a control variable based on its established role when it comes to investigating government formation (see Bergman et al. 2021;Strøm et al. 2008) and because it has been shown that non-partisan appointments are more frequent under minority than under majority governments (Amorim Neto and Strøm 2006). Moreover, following the literature on technocratic appointments, we also control for the government's Left-right position. In this regard, research by Hallerberg and Wehner (2018) and Kaplan (2017) has pointed out that left prime ministers are associated with more appointments of experts during periods of economic distress (but see also Alexiadou and Gunaydin 2019). Measuring the cabinet's left-right position is not a straightforward task, especially when the cabinet is made by more than one party. Similarly to Woldendorp et al. (1998), we have created a categorical variable (distinguishing whether the government is left-leaning, centre-leaning, or right-leaning) that considers the weight of each party in each government. 8 Furthermore, following Amorim Neto and Strøm (2006), we account for the government's Parliamentary support by looking at the share of parliamentary seats held by governmental parties. Finally, we also add a dichotomous variable (First government) that takes value 1 for the first government of each legislature, in line with studies that find that the recourse to technocracy is less likely in governments formed immediately after the elections (e.g. Schleiter and Morgan-Jones 2010; Wratil and Pastorella 2018).
Concerning institutional constraints, we control for a distinct feature of the electoral system, namely its level of personalisation, a key predictor in Alexiadou and Gunaydin (2019). We have followed the well-known scheme by Carey and Shugart (1995) updated by Johnson and Wallack (2010). The resulting variable, Personal vote, has a theoretical range from 1 to 13, although it empirically ranges from 1 (closed-list systems) to 10 (single-member-district systems).
Moreover, the literature has found that the recourse to technocrats is more likely in semi-presidential or presidential systems than in parliamentary ones Morgan-Jones 2009, 2010;Wratil and Pastorella 2018). This is because the stronger the presidential powers, the higher the president's opportunity to appoint non-political experts (i.e. technocrats) who will therefore primarily have the president as their principal rather than partisan ministers, who instead would be more likely to respond to the parliament (Amorim Neto and Strøm 2006;Semenova 2020;Tavits 2009; but see Brunclík and Parízek 2019). Usually, the literature opts for a dichotomous distinction between parliamentary systems vs. presidential and semi-presidential ones (Alexiadou and Gunaydin 2019; Semenova 2020). Nonetheless, this operationalisation does not distinguish between parliamentary systems where, for instance, the monarch has a mere procedural role (e.g. Denmark, Sweden, or the United Kingdom) and other parliamentary systems where the president may have substantive powers in government formation (e.g. Italy). Therefore, consistent with Wratil and Pastorella (2018), we have used the measure of Presidential powers by Siaroff (2003), ranging from 0 to 9. 9 Furthermore, given the connection between the responsibility-responsiveness dilemma and external constraints (Mair 2009(Mair , 2013, we introduce two additional controls. The first is an ordinal variable for EU constraints (having value 0 if the country is not a member of the EU in a given year, value 1 if the country is an EU member, and value 2 if the country has also adopted the Euro). The second is a dichotomous variable (Crisis-related constraints) to tackle those cases that signed a memorandum of understanding with the so-called 'Troika' committee (the European Central Bank, the European Commission, and the International Monetary Fund) and were subjected to severe austerity programs (Lefkofridi and Nezi 2020).
The third set of controls concerns the structure of party competition, which can be captured by two elements: the number of competitors within a party system and their level of ideological polarisation. Both factors can complicate the bargaining process during government formation (Golder 2010;Martin and Vanberg 2003;Schleiter and Morgan-Jones 2010;Wratil and Pastorella 2018). In contexts with many competitors or ideologically distant parties, the recourse to technocrats can be a way to get out of the impasse in the bargaining (but see Amorim Neto and Strøm 2006). We measure party system fragmentation via the Effective Number of Parliamentary Parties (ENPP) by Laakso and Taagepera (1979) and party system polarisation through Dalton's (2008) Polarisation index. 10 Another important set of controls is related to the socio-economic context. First, previous literature has emphasised the role of economic conditions, particularly economic crises, for technocrats' appointments, with economic crises positively associated with the share of technocrats (Brunclík and Parízek 2019; Schleiter and Morgan-Jones 2010; Semenova 2020). 11 In particular, the literature focussing on the 2008 Great Recession underlines that the appointment of technocrats (or of technocratic-led governments) is more likely in contexts of harsh economic conditions because, in this situation, the need to implement unpopular reforms is more likely and parties prefer not to handle the responsibility on such inflammatory issues (Alexiadou and Gunaydin 2019;Wratil and Pastorella 2018). We use the GDP growth rate measured one year before the formation of each government. 12 Second, the level of corruption has been associated with an increase in votes for new parties (Engler 2016), and we might imagine a high level of corruption can delegitimate political parties and push them to cede responsibility to technocrats to reduce the cost of the voters' sanction in the following elections. We have used the Political Corruption Index from the Varieties of Democracy (V-Dem) dataset (Coppedge et al. 2021), ranging from 0 to 1.
Finally, as the literature has shown that technocracy has recently become more relevant in government formation (Pastorella 2016), we also add a control for time (Year). 13 Table A1 in the online appendix reports the descriptive statistics of our variables. Therefore, our dataset is based on cabinet-level, legislature-level, and country-level variables. We also have a hierarchical data matrix (655 cabinets nested within 373 legislative terms, in turn, nested within 20 countries 14 ). With such a structure, a simple linear regression method would return biased results (Rabe-Hesketh and Skrondal 2008) because errors among units belonging to the same cluster will likely covary (e.g. cabinets within the same legislature or legislatures within the same country). Therefore, it is important to specify a model that considers the hierarchical nature of the data structure. Following Schmidt-Catran and Fairbrother (2016: 25), our data structure fits their Model D reported in Figure 2.
The figure considers a three-level structure, where governments (G in the figure) are nested within country*legislative term dyads (CL in the figure), which, in turn, are nested within countries (C in the figure). We have performed an ICC (intra-class correlation) test on the null model that has suggested us opt for this specification against alternative ones. 15 So, in the regression analyses of the next section, we perform multilevel mixed regression models, with country and country*legislative terms set as random intercepts. 16 Table 1 reports the results of the regression analysis. Model 1 is the null model, with legislature and country*legislature random intercepts, Model 2 adds all the control variables, while Model 3 also includes our focal predictor, TV. Before discussing the empirical test of our main hypothesis, let us first look at the overall picture provided by the effect of our control variables in Models 2 and 3. Consistently with most literature, Presidential powers emerge as a substantive predictor of technocratic presence in governments: thanks to our variable operationalisation as an ordinal measure, we can precisely distinguish between, on the one hand, the effect on cases with low presidential powers -substantively indifferent compared to the reference category of no presidential powers at all (assigned to monarchies) -and, on the other hand, the strong effect on cases where presidents have a more substantial influence on the country's politics. 17 Surprisingly enough, the economic conditions of a country have no effect, although a large literature has underlined a negative effect of economic growth on the recourse to technocrats, with specific emphasis on the years following the onset of the Great Recession (Alexiadou and Gunaydin 2019; Semenova 2020; Wratil and Pastorella 2018). Although we do find the expected negative effect, the latter is never significant, 18 thus showing that, as we broaden the temporal perspective beyond the specific Great Recession scenario, the role of the economy fades away.

Analysis and results
Conversely, the two controls related to the presence of external constraints (EU constraints and Crisis-related constraints) show a positive and significant effect, thus confirming the issue of responsibility imposed by supranational institutions (e.g. Cotta 2018; Mair 2009Mair , 2013. Similarly, also the Political Corruption Index exerts a positive and significant impact on the Share of technocratic positions, witnessing that the appointment of technocrats is more likely in more corrupt contexts. 19 Finally, consistently with Amorim Neto and Strøm (2006: 640), we find that technocratic appointments are more frequent under minority cabinets compared to our reference category of minimal winning coalition cabinets. Moreover, the analysis also shows that the share of technocrats is comparatively higher in single-party majority cabinets. Intuitively, in a single-party majority cabinet, handing over some ministerial positions to technocrats helps the ruling party dilute responsibility, but is less costly than a coalition government, where the number of ministerial positions potentially available to each party is lower. In other words, while, in single-party majority cabinets, the ruling party might have more room for manoeuvre in ministerial appointments, in coalition governments, where the governmental prize shall be divided among a high number of players, the presence of technocrats is likely to result in a violation of the proportionality norm in portfolio allocation, the so-called Gamson law (Gamson 1961).
Our analysis makes a further step forward in outlining the specific mechanism linking the responsibility-responsiveness gap to the increasing appointment of technocrats. Indeed, we find that electoral volatility plays the lion's share in this story. Model 3 introduces our focal predictor, TV.
This leads to a remarkable improvement in the model's goodness of fit, as witnessed by the increase in Model 3's Wald Chi-square statistics compared to Model 2's. This is because, as we hypothesised, TV plays a substantive role in explaining the recourse to technocrats in Western Europe. Its coefficient, significant at the highest level of confidence (p < 0.001), shows that as TV increases by one point, the Share of note: Multilevel mixed regressions with country and country*legislative term as random intercept; standard errors are reported. + p < 0.10; * p < 0.05, ** p < 0.01, *** p < 0.001. technocratic positions increases by 0.41 points. Substantively, by moving from a situation where there is no volatility to a situation that meets Mair's threshold of high volatility (Mair 2011), namely, a TV of 20, the percentage of technocrats grows by more than 8 points. This is particularly impressive, as this effect is obtained after controlling for all the previously discussed factors, most of which were at the centre of previous studies on the topic. 20 The inclusion of TV clarifies the mechanism leading from the responsibility-responsiveness dilemma to technocrats' appointment. The abovementioned process, leading ruling parties to hand over governing positions to technocrats to dilute responsibility, occurs especially in turbulent times.
Indeed, as already discussed above, in contexts of 'inelastic' voters' response (i.e. with low volatility), parties will feel relatively safe from handling the responsibility-responsiveness dilemma, as their need to be responsible at the expense of voters' demands is less likely to get sanctioned in the electoral ballot. Conversely, in electorally turbulent times, ruling parties are conscious that voters are more likely to punish them for being unresponsive and, consequently, will be more prone to 'pass the hot potato' of responsibility to technocrats.
Moreover, the fact that TV is significant despite the presence of different institutional and non-institutional constraints and crisis-related indicators in the model means that it is not simply the fact that the need for responsibility has become more urgent in recent times, but it is the presence of a turbulent electoral environment that has a more critical say in this matter.
However, the relationship between TV and Share of technocratic positions is not just the product of recent years, characterised by increasing international constraints, especially after the 2008 Great Recession and the subsequent more active role of supranational institutions in countries' public policy. Conversely, such a relationship holds (at the same level of confidence) also when we exclude from the analysis the post-2008 years (see Table A4 in the online appendix), thus showing that the electoral context has always played an important role in our story. Thus, our result is consistent with Amorim Neto and Strøm (2006), who found a positive effect of TV on governments' technocratic appointments in their analysis of European republics in the 1990s. However, their explanation was based on the specific relationship between the prime minister and the president in government formation and was tailored to cases where the president had a substantive role in government formation. Instead, our analysis shows that volatility has a widespread effect across time and space, as we consider the whole post-1945 period and also include Western European monarchies. Also, our analysis presents a more comprehensive explanation of the role of electoral volatility. The latter is not limited to the specific moment of the bargaining between the prime minister and the president in government formation. Indeed, by running the analysis exclusively on parliamentary regimes or excluding those cases where the president has a substantive role in government formation (Siaroff 2003), the effect of TV remains as strong and significant as in our main model (see Table A4 in the online appendix). Finally, the effect of TV is robust to country fixed-effects (Table A5 in the online appendix) and other model specifications and alternative controls. 21 The role of turbulent times in fostering the recourse to technocrats is not limited to the electoral context but also concerns the level of instability in the parliamentary arena. Indeed, if we replace TV with parliamentary volatility (TPV) ( Table A8 in the online appendix), we still find the same effect on the dependent variable. From another viewpoint, this means that the effect of turbulent times is robust to different measurements.
All in all, our H1 finds empirical confirmation. Nonetheless, we are also interested in providing a more fine-grained explanation of the specific driver linking turbulent times to the appointment of technocrats. We have conjectured the presence of three alternative mechanisms, with H1a predicting the driving impact of the volatility across the government-opposition dividing line (GopV), H1b hypothesising the leading role of electoral instability in the balance of power among established parties (AltV), and H1c putting forward the pre-eminence of the electoral unpredictability due to the entry of new parties or the exit of old parties from the system (RegV). To test which of the three drivers has the most important role in our story, we have replicated Model 3 in Table 1 with, respectively, GopV, AltV, and RegV (Model 4, 5, and 6 in Table 2).
The analysis clearly shows the prominence of Regeneration vis-à-vis Government-Opposition and Alteration. While all are statistically significant, the effect of a 1-point-increase in RegV is almost 2.5 times bigger than that in GopV and four times bigger than that in AltV. Figure 3 reports the coefficient plot that compares the effect of GopV, AltV, and RegV. Regeneration volatility has an impressive effect on technocrats in Western European governments: the Share of technocratic positions increases by 1.18 points for a unitary increase in RegV. So, by moving from a situation with no Regeneration at all to one where Regeneration is extremely high (say, RegV equal to 15), the expected share of technocrats increases by more than 17 points. 22 In other words, the electoral change due to voting shifts between government and opposition parties or among all established parties is much less important than the electoral change brought about by the emergence of new parties and the disappearance of old ones. Therefore, the critical driver leading ruling parties to devolve governing responsibility to non-partisan experts in electorally turbulent times is what Emanuele (2017, 2022) call 'party system unpredictability' . 23 Such unpredictability means the ruling parties are more and note: Multilevel mixed regressions with country and country*legislative term as random intercept; standard errors are reported. + p < 0.10; * p < 0.05, ** p < 0.01, *** p < 0.001. more vulnerable in the increasingly riskier electoral competition: every time a new and unknown party appears or a known one dies, interactions among actors are likely to change, and ruling parties feel they are stuck in a 'no country for old men' . Therefore, it is even more urgent for ruling parties to dilute responsibility by giving technocrats more ministries to play on the safe side of responsiveness.

Conclusion
This article has investigated the determinants of technocratic appointments in Western European governments. The recourse to technocrats has been on the rise in recent years, and a growing body of literature has been investigating a few connected aspects such as the patterns of representation vis-à-vis the traditional model of party government, the social characteristics and policy-making style of technocrats, and voters' attitudes towards them. However, little has been said about the explanatory factors of this manifest phenomenon.
The main attempts to explain the recourse to technocrats have focussed either on the President's powers and autonomy in government formation or on economic downturns, but they offer limited empirical evidence from a comparative longitudinal perspective. As the recourse to technocrats is not limited to presidential and semi-presidential regimes or periods of economic downturns, we have provided an original framework to reach a more encompassing explanation.
We have focussed on the turbulence of the electoral environment as a contextual factor that plays a crucial role in the trade-off that political parties face to cope with the responsibility-responsiveness gap illustrated by Peter Mair (2009). As ruling parties are forced to deliver policy outputs that comply with supranational institutions and globalised markets even at the expense of the satisfaction of voters' demands, responsibility and responsiveness are increasingly perceived to be at odds. Therefore, ruling parties have to choose between two different strategies: holding responsibility to maximise their office-seeking goals and accepting the risk of being punished by voters at the ballot box or, instead, renouncing some power by ceding government posts to technocrats to dilute responsibility.
Based on this premise, we have tested the hypothesis that ruling parties opt for a strategy of diluting responsibility the more the electoral context becomes turbulent, as this is an indicator that voters are available to change their previous vote choice and prone to sanction the government's actions. In other words, electorally volatile contexts should be positively associated with the appointment of technocrats in governments. We have tested this expectation on a longitudinal multilevel dataset including 655 cabinets and 373 elections in 20 Western European countries from 1945 to 2021.
The results of our empirical analysis strongly support our expectation: electoral volatility profoundly impacts the recourse to technocratic appointments in governments. This result is robust to many different controls and model specifications. Interestingly, as further evidence that economic downturns and presidential powers tell only a part of the story, the effect of electoral volatility on technocratic appointments is also found in the pre-Great Recession period and in parliamentary democracies.
Moreover, the article has also investigated the specific driver of technocratic appointments by testing three alternative hypotheses related to three different sources of volatility: the electoral shifts between government and opposition parties, those shifts that occur more generally among established parties, and those where instead voters move away from old competitors and towards new ones.
The analysis shows that the electoral volatility brought about by the emergence of new parties and the disappearance of old ones is far more important than the electoral volatility due to voting shifts across the government-opposition dividing line or among established parties. In other words, party system unpredictability is the key driver for technocratic appointments, thus confirming that such a solution should be conceived as a strategic choice to dilute responsibility to anticipate voters' future reactions.
These findings may open new research paths on the role of the electoral context on government formation. Specifically, three main implications from this study deserve some attention. First, the need for a deeper analysis of the link between party system change and technocratic drift. The increasing recourse to technocratic appointments may be a stable scenario for the years to come, as Western Europe has entered an era of electorally turbulent times. Under this point of view, further expansion of the comparative research in the explanation of technocratic appointments -widening the analysis' geographical scope and even including governments in the post-pandemic scenario -will likely become critical. Second, as our article also focuses on the impact of the electoral volatility due to the emergence of new parties, an in-depth analysis of the specific effect provided by the entrance into the sphere of executive power of new populist parties will be needed. Third, and most importantly, the article's quantitative analysis may be fruitfully corroborated by qualitative analyses of the roles played by the different actors involved in government formation (party leaders, parliamentary parties, supranational actors and public opinion). Such a renovation of the research agenda on parties, governments and technocracy should be an urgent concern for all those dealing with one of the most critical questions of modern comparative politics: who governs? Notes 1. According to Caramani (2017), technocracy and populism criticize party democracy, as both share the same non-pluralistic view of society and the un-mediated relationship between elites and citizens, and, therefore, technocrats and populists in government constitute a challenge for party government. 2. While party politicians are heavily dependent on democratic accountability, technocrats' behaviour in government is essentially subject to their peers' judgments concerning professional reputation (Cotta 2018). Ceding posts to technocrats is, of course, costly under the assumption that political parties are office-seekers that aim at maximizing their political power (Downs 1957). 3. The connection between electoral environment and technocracy was already made by Amorim Neto and Strøm (2006), albeit tailored to the specific context of the bargaining between the prime minister and the president in government formation and tested only on European republics in the 1990s. 4. Previous research has found that electoral change has primarily damaged incumbent parties (Hernández and Kriesi 2016;Müller and Strøm 2000). 5. This argument is in line with Bartolini's conceptualization of vulnerability as a dimension of competition (Bartolini 2000: 54). Thus, a high electoral availability along the incumbent-opposition line is a key condition to ensure that incumbent parties feel vulnerable to the electoral outcomes, a crucial prerequisite for responsiveness.
6. Our focus is, therefore, on political independence rather than on knowledge or expertise (see Bertsou and Caramani 2020a;McDonnell and Valbruzzi 2014). At any rate, our hypotheses have also been tested on slightly different operationalisations of the dependent variable (see next section and the online appendix). The calculation of technocrats' share is based on the total number of ministries (including the prime minister position) and not on the total number of ministers as the denominator. Although it has seldom occurred, the rationale behind our calculation is weighting posts, not people, because if the same person holds two different ministries, there is not one but two positions to be considered, held by the same minister. Data on cabinet composition come from Casal Bértoa (2021) and, in case of missing data, from Sonntag (2015). 7. For instance, for the 1997-2001 UK legislature, we calculate the absolute difference in the vote share of the Conservative Party (the only party supporting the last government of the previous legislature, the Major II government) in the 1997 and 1992 general elections. 8. See the online appendix for more details on control variables' operationalization. 9. Siaroff (2003) points out the president's specific powers, from being or not popularly elected to her role in government formation, up to her ability to dissolve the legislature. On this last point, see also Schleiter and Morgan-Jones (2010) and Semenova (2020). Notice that we have given the value 0 to monarchies. 10. Data on ENPP come from Gallagher (2020), while data on Polarization, related to seat distribution in parliament, come from Parlgov (Döring and Manow 2020). 11. But see Amorim Neto and Strøm (2006). 12. Data come from the Total Economy Database (Conference Board 2020). 13. Year is measured at the time of each parliamentary election. In cases of multiple governments within a legislative term, Year is the same across these governments. A time variable is also helpful in time-series data to avoid spurious correlations among variables that vary in a consistent direction over time (Tavits 2005). Further controls and specifications are present in the online appendix, see the next section. 14. Austria, Belgium, Cyprus, Denmark, Finland, France, Germany, Greece, Iceland, Ireland, Italy, Luxembourg, Malta, Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, and United Kingdom. 15. For further robustness checks on different model specifications, see the online appendix. 16. We have performed a Wooldridge test of serial correlation (Drukker 2003), which excludes issues of autocorrelation in the data. We have also performed a VIF test that has reported no relevant multicollinearity issues (mean VIF equal to 2.08). 17. The effects of our focal predictor and Presidential powers do not change if we treat the latter as a continuous variable. 18. If we test alternative measures of the state of the economy, such as the unemployment rate or public-debt-to-GDP ratio ( Table A2 in the online appendix), calculated one year before the formation of each government (data come from the International Monetary Fund), the models' substantive results do not change. However, the number of observations drops significantly to 514 and 492 respectively, as there are no available data for earlier cases. At any rate, the effect of unemployment is not significant, while only that of public-debt-to-GDP ratio is significant at p < 0.05. A final alternative for the state of the economy is, following Wratil and Pastorella (2018), the 'misery index' . We have built two alternative versions of the index: the first is obtained by summing the standardized scores of (inverted) GDP growth rate and unemployment. The index shows high values when economic conditions are negative, and vice versa. The total number of observations is 514 (all observations reporting a value of GDP growth rate and unemployment). The second measure of the misery index also adds a standardized variable of public debt. This time, the number of observations drops to 469 (all observations reporting a value of GDP growth rate, unemployment, and public debt). In both cases, while our substantive results do not change, the index shows a positive but not significant effect on our dependent variable, showing once again the overall weak effect of the state of the economy on technocrats (see Table A2 in the online appendix). 19. Another indicator of political crisis that could have been connected to the delegitimization of political parties and their consequent urgency to cede responsibility to technocrats in governments may be the presence of political scandals. Indeed, Wratil and Pastorella (2018) find that political scandals trigger the formation of governments led by technocrats. By including their dichotomous variable of political scandals, which we have updated for the post-2013 years, the number of observations in our models falls to 371 cases (there are no available data about political scandals before 1977). In any case, such a phenomenon does not significantly influence our dependent variable, while our substantive results are confirmed ( Table A3 in the online appendix). 20. The mere descriptive statistics return an even more remarkable difference: in the 599 governments formed after a parliamentary election with a TV lower than 20, the average Share of technocratic positions is less than a third compared to that in the 99 governments formed after an election having a TV higher than 20 (6.1 vs. 18.3). 21. In Table A5 in the online appendix, we have re-run Model 3 in Table 1 with country fixed-effects to control whether our results are robust to potential omitted variable bias at the country level; moreover, given that our dependent variable is a proportion, with a lower and an upper bound, we have run a tobit regression model and a generalized linear model with a logit link function. Second, we have also replicated Table 1's analyses through Model A specification by Schmidt-Catran and Fairbrother (2016: 25), namely, with only country*legislative term dyads set as random intercepts (Table A6 in the online appendix). Furthermore, we have also tested the robustness of TV to alternative controls (Table A3 in the online appendix): replacing the continuous Time variable with a categorical variable for each decade; using fragmentation and polarization data calculated on votes instead of seats; replacing our measure of Left-right position with an alternative measure calculated starting from the rile variable in the Comparative Manifesto Project (Volkens et al. 2021) weighted by the cabinet seats held by each party in the government; using a dichotomous variable distinguishing whether the prime minister is left-leaning or not (see Alexiadou and Gunaydin 2019) instead of Left-right position; replacing Presidential powers with a dichotomous variable distinguishing whether each country has a semi-presidential system or not; adding a control for the age of democracy, usually included by those scholars working on young democracies (Amorim Neto and Strøm 2006;Brunclík and Parízek 2019;Semenova 2020;). We have also added two other controls to account for the organizational crisis of political parties and the alleged parallel rise of the personalization of politics: we have tested two additional variables. The first one captures the overall membership/electorate ratio: membership data come from the MAPP dataset (Van Haute and Paulis 2016), Marino (2021) for Greece, and other party-related sources; electorate data come from Nohlen and Stöver (2010) and, for recent years, official electoral authorities. The second variable (ranging from 0 to 10) captures the degree to which the politics of a given country is personalized, with higher values indicating a stronger impact of personalization. Data come from Marino et al. (2022) (see Table A7). The effect of our focal predictor is robust across all models. 22. The results of the analysis in Tables 1 and 2 are substantively confirmed through slightly different operationalisations of our dependent variable that emphasize the expertise of non-partisan ministers. Indeed, we have excluded from the pool of non-partisan ministers those politically independent ministers without any expertise or even those who, despite having specific policy expertise, have been appointed in a ministry which is not related to their expertise. The replication of the analysis with such different operationalisations confirms the impact of our independent variables in terms of coefficient sign and significance, with the only partial exception of AltV, which turns out to be significant only at p < 0.10 in the model with the more restricted definition of expertise. This finding further reinforces the effect of TV and, in particular, RegV on the appointment of technocrats (not only independents but also experts) in Western Europe (see Tables A9 and A10 in the online appendix). 23. As a further confirmation of the importance of the emergence of new parties of the recurrence to technocrats, our result is robust also by replacing RegV with a simple share of votes obtained by new parties in parliamentary elections or with the Parliamentary Volatility by Regeneration (RegPV) ( Table A8 in the online appendix). Further notice that all the previously mentioned robustness checks that confirmed the effect of TV also confirm the effect of RegV.