A simple OLS regression between transformistic tendencies and lawmaking productivity suggests no correlation between MPs' political switching and their legislative activity (Table 2).
Table 2:
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
|
|
|
|
|
|
|
|
Party switch
|
-0.0286
|
-0.0969
|
-0.0854
|
-0.0583
|
-0.0886
|
-0.0401
|
-0.0572
|
|
(0.0776)
|
(0.0727)
|
(0.0697)
|
(0.0664)
|
(0.0614)
|
(0.0596)
|
(0.0554)
|
MP is female
|
0.232***
|
0.175***
|
0.147**
|
0.123**
|
0.141***
|
0.133**
|
0.126**
|
|
(0.0637)
|
(0.0593)
|
(0.0570)
|
(0.0539)
|
(0.0522)
|
(0.0514)
|
(0.0493)
|
MP’s age < 35 years old
|
0.131
|
0.130
|
0.111
|
0.155
|
0.176
|
0.126
|
0.0661
|
|
(0.123)
|
(0.134)
|
(0.131)
|
(0.125)
|
(0.122)
|
(0.119)
|
(0.114)
|
35 > MP’s Age < 45
|
0.0745
|
0.154
|
0.168*
|
0.183**
|
0.182**
|
0.130
|
0.0983
|
|
(0.0941)
|
(0.0962)
|
(0.0940)
|
(0.0887)
|
(0.0856)
|
(0.0814)
|
(0.0775)
|
45 > MP’s Age < 55
|
0.0677
|
0.0745
|
0.0492
|
0.0869
|
0.123
|
0.0871
|
0.0565
|
|
(0.0872)
|
(0.0919)
|
(0.0897)
|
(0.0835)
|
(0.0794)
|
(0.0762)
|
(0.0727)
|
55 > MP’s Age < 65
|
-0.0295
|
-0.0310
|
-0.0162
|
0.0302
|
0.0632
|
0.0550
|
0.0209
|
|
(0.0889)
|
(0.0944)
|
(0.0935)
|
(0.0863)
|
(0.0822)
|
(0.0800)
|
(0.0753)
|
MP’s age > 65 years old
|
-0.0859
|
-0.0393
|
-0.0843
|
-0.0634
|
-0.0482
|
-0.0983
|
-0.143
|
|
(0.142)
|
(0.124)
|
(0.121)
|
(0.116)
|
(0.112)
|
(0.108)
|
(0.0965)
|
Grade 12
|
-0.0794
|
-0.320
|
-0.234
|
0.00121
|
0.0936
|
0.0774
|
0.0790
|
|
(0.294)
|
(0.232)
|
(0.226)
|
(0.178)
|
(0.156)
|
(0.160)
|
(0.159)
|
University degree
|
0.130*
|
0.100*
|
0.0754
|
0.0793
|
0.0652
|
0.0779
|
0.0518
|
|
(0.0686)
|
(0.0578)
|
(0.0539)
|
(0.0523)
|
(0.0482)
|
(0.0475)
|
(0.0461)
|
Doctorate (Ph.D.)
|
0.258*
|
0.195
|
0.182
|
0.0203
|
0.0219
|
0.0375
|
0.0295
|
|
(0.139)
|
(0.143)
|
(0.144)
|
(0.0902)
|
(0.0886)
|
(0.0887)
|
(0.0853)
|
Center-right affiliation
|
0.0230
|
0.0111
|
0.0718
|
-0.00267
|
-0.0387
|
-0.0752
|
-0.0595
|
|
(0.126)
|
(0.123)
|
(0.120)
|
(0.105)
|
(0.0989)
|
(0.0936)
|
(0.0849)
|
Center-left affiliation
|
0.120
|
0.169
|
0.198
|
0.110
|
0.0659
|
0.0106
|
0.0188
|
|
(0.132)
|
(0.125)
|
(0.123)
|
(0.106)
|
(0.101)
|
(0.0947)
|
(0.0858)
|
MP > 4 mandates
|
0.705***
|
0.353*
|
0.252
|
0.127
|
0.122
|
0.0873
|
0.0166
|
|
(0.148)
|
(0.182)
|
(0.164)
|
(0.140)
|
(0.137)
|
(0.135)
|
(0.121)
|
1 ≥ MP mandates ≤ 4
|
0.229***
|
0.0896
|
0.0852
|
0.0403
|
0.0250
|
0.00623
|
-0.0176
|
|
(0.0706)
|
(0.0613)
|
(0.0595)
|
(0.0576)
|
(0.0549)
|
(0.0540)
|
(0.0525)
|
Gov. undersecretary
|
0.179
|
-0.191**
|
-0.206**
|
-0.187**
|
-0.174**
|
-0.191**
|
-0.129
|
|
(0.109)
|
(0.0904)
|
(0.0899)
|
(0.0879)
|
(0.0882)
|
(0.0863)
|
(0.0798)
|
Presidential secretary
|
-0.0423
|
-0.110
|
-0.118
|
-0.115
|
-0.116
|
-0.130
|
-0.108
|
|
(0.127)
|
(0.126)
|
(0.122)
|
(0.126)
|
(0.127)
|
(0.128)
|
(0.126)
|
Alpha trim level
|
None
|
2.5%
|
5%
|
7.5%
|
10%
|
12.5%
|
15%
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
Adjusted R-squared
|
0.116
|
0.051
|
0.046
|
0.029
|
0.038
|
0.028
|
0.030
|
Centered R-squared
|
0.140
|
0.0759
|
0.0719
|
0.0559
|
0.0650
|
0.0562
|
0.0586
|
Clustered standard errors - at MP level - in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table presents the empirical coefficients of the OLS regressions conducted between party hopping (in Italian context described as “parliamentary turncoat”) and parliamentarian productivity (OpenPolis lawmaking index), along with a number of variables included in a control vector. The analysis is based on a sample of 629 MPs from the XVI legislature (2008-2013) of the Italian Chamber of Deputies.
However, it is known that attitudes toward party hopping can differ significantly, depending on a set of institutional background conditions (Kreuzer & Pettai, 2009) that make this variable highly endogenous (Ferland & Dassonneville, 2021).
From an historical point of view, historiographers have concentrated significant attention to trace the root of this behavior in the Italian unification as the end of intertwined political, personal, family, factional, territorial, and social divisions between the center-north and center-south, where pluralism was not tolerated (Fruci, 2019).
One of the most remarkable aspects following unification was the fluid political positions (Rogari, 1998) of MPs hailing from the former Kingdom of the Two Sicilies (KTS). Several studies have concluded that such fluidity was a result of an "ideology based on small and big private interests," with parliamentary turncoat being the typical strategy of the rearmost bourgeois class to gain power (Carocci, 1995). The primary reason behind this fluidity was the absence of a consolidated active bourgeoisie, replaced by a persistent feudal nobility that readily adapted to secure its hegemonic position during the transition from the Bourbons to the Savoy (Lupo, 2011).
Therefore, in the pre-unitary Bourbon south, weak bourgeois classes played a marginal role and lacked political representation (Felice, 2019). The monarchy's inability to modernize the state, coupled with its closed and aloof nature, further compounded the situation (Caglioti, 2002). Hence, it's not surprising that historians highlight a tendency for parliamentary turncoat, which was "pathological in the North and the Center but physiological in the South," where "political strife concealed limited and undisputed personal interests" (Carocci, 1956), and the prevalent "distorted political mentality" revolved around "wheeling and dealing" (Colapietra, 1956).
Parliamentary turncoat faced greater resistance in the North due to the presence of a more cohesive and homogeneous middle class (Fontana, 1984). In contrast, the South viewed unification as an unwelcome intrusion from the North (Benigno, 2019), leading to the establishment of a dominant local clientelism system to bypass central authority and the rule of law (Petraccone, 1994). This resulted in a stark divide between the North and South, particularly evident in the realm of politics (De Francesco, 2020).
Contemporary accounts of the time indicated that the KTS was marked by exclusive privileges in public positions (Rocco, 1837) and a constant fear of "shameless betrayal" by authorities (De' Sivo, 1868). Eventually, a tendency for parliamentary turncoat prevailed because the Kingdom was plagued by traitors and corrupt officials who were both corrupted and corrupting (Bruno Guerri, 2020).
In this context, top-down patronage structures and clashes between closed client networks (Blando, 2019) may still have lingering effects today, as self-reinforcing feedback loops and long-term processes follow the principles of increasing returns (Pierson, 2000). It has already been demonstrated, indeed, that past arrangements constrain present economic opportunities (Libecap, 2011) through behavioral spillovers (Bednar & Page, 2018).
To test this hypothesis, we employ a two-stage least squares (2SLS) estimating method, instrumenting politicians' parliamentary turncoat with a binary variable that takes a value of 1 if the MP was elected in an electoral mandate that overlaps with the KTS' territories (see Figure 2) before the unification of Italy in 1860, and 0 otherwise.
If there is a form of path dependency in these territories that shapes current civic cultures (Sabetti, 1996), we should observe a positive and statistically significant relationship between our instrumental variable and parliamentary turncoat. Simultaneously, if OLS results are influenced by endogeneity, we should obtain different results in the second stage of the regressions. Considering that parliamentary turncoat has widely been regarded as a despicable betrayal (Galli della Loggia, 2016), our expectation is that defecting MPs are less likely to introduce new bills, garner support from their peers, and witness a positive final vote regarding their legislative inputs.
A reduced form of the empirical evidence of the first and second stage of the 2SLS regressions is reported in Table 3. In contrast, extended results are reported and commented on in the Appendix (Tables 5-6).
Table 3:
1st stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Kingdom of two Sicilies
|
0.128***
|
0.135***
|
0.143***
|
0.141***
|
0.149***
|
0.147***
|
0.160***
|
|
(0.0360)
|
(0.0365)
|
(0.0370)
|
(0.0375)
|
(0.0376)
|
(0.0377)
|
(0.0377)
|
Cragg-Donald Wald first stage F statistic
|
14.34
|
15.29
|
16.97
|
16.06
|
18.05
|
17.52
|
21.21
|
Controls
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
Adjusted R-squared
|
0.065
|
0.067
|
0.070
|
0.068
|
0.073
|
0.070
|
0.085
|
Centered R-squared
|
0.0888
|
0.0910
|
0.0950
|
0.0938
|
0.0990
|
0.0969
|
0.112
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
|
|
|
|
|
|
|
|
2nd stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Party switch
|
-1.381**
|
-1.269**
|
-1.113**
|
-1.108**
|
-0.928**
|
-0.981**
|
-0.773**
|
|
(0.644)
|
(0.516)
|
(0.438)
|
(0.435)
|
(0.376)
|
(0.389)
|
(0.317)
|
|
|
|
|
|
|
|
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
Adjusted R-squared
|
-0.364
|
-0.436
|
-0.393
|
-0.497
|
-0.336
|
-0.465
|
-0.282
|
Centered R-squared
|
-0.329
|
-0.399
|
-0.355
|
-0.456
|
-0.298
|
-0.423
|
-0.244
|
First satge Cragg-Donald Wald F statistics
|
14.34
|
15.29
|
16.97
|
16.06
|
18.05
|
17.52
|
21.21
|
Alpha trim level
|
None
|
2.5%
|
5%
|
7.5%
|
10%
|
12.5%
|
15%
|
Clustered standard errors - at MP level - in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table presents the empirical coefficients of the 2SLS regressions conducted between party hopping (in Italian context described as “parliamentary turncoat”), which is instrumented with a dummy variable that equals one if the parliamentarian was elected in the territories of the former Kingdom of the Two Sicilies, and parliamentarian productivity (OpenPolis lawmaking index). The analysis also includes a number of control variables in a vector. The sample consists of 629 MPs from the XVI legislature (2008-2013) of the Italian Chamber of Deputies.
Before discussing the evidence reported in Table 3, we need to assess whether the instrumental variable (IV) is valid and sufficiently robust. A typical rule of thumb for validity is represented by a first-stage F-statistic exceeding 10 (Staiger & Stock, 1997). Stock & Yogo (2005) critical values for first-stage F-statistics are set at a 1% threshold and a 15% maximum IV size, with the lower threshold of 8.96 for both chi-squared and F-statistic. Importantly, all first-stage F-statistics in our analysis exceed both of these thresholds. Furthermore, the Cragg-Donald Wald F-statistics, a key measure of instrument strength (Sanderson & Windmeijer, 2016), exhibits a maximum IV size increase to 10% once the dataset is trimmed, surpassing the 5% alpha-trimmed level threshold (in which case, the Stock-Yogo weak ID test critical value is set at 16.38).
In light of these results, the robustness of our instrumental variable (IV) is implied. These collective findings provide strong evidence supporting the robustness and reliability of our instrumental variable, reinforcing its suitability for the two-stage least squares (2SLS) estimation method employed in our analysis.
In the first stage, the coefficient for the instrumental variable consistently appears positive and statistically significant across all columns, indicating that MPs elected in territories pertaining to the former Kingdom of the Two Sicilies (KTS) is linked to a higher likelihood of engaging in party switching. The estimates reveal that MPs elected in the KTS territories exhibit a greater propensity to switch parliamentary groups and cross the floor. Specifically, on average, the probability of a parliamentarian elected in the KTS territories switching parties is 14 percentage points higher compared to their colleagues from other electoral districts.
In the second stage of the analysis, the coefficients for party switching are consistently negative and statistically significant, signifying that parliamentarians who participate in parliamentary turncoat tend to exhibit lower parliamentary productivity, as measured by the OpenPolis lawmaking index.
The statistical significance of the coefficients of interest remains robust to various model specifications (refer to Tables 7-8 in the Appendix) and when limiting the data to the borders of the KTS territories (refer to Table 9 in the Appendix).
Overall, the results point to a negative correlation between party switching and parliamentary productivity. Parliamentarians elected in the KTS territories are more likely to change political affiliation during their term, and this behavior is associated with reduced legislative productivity. While the negative impact of party switching on productivity holds across different model specifications, there may be slight variations in the magnitude of the effect. These results can be elucidated from a historical standpoint. To begin with, the KTS territories were characterized by weak institutions (Brosio, 2018) and a "vertical patron-client relations of dependence" on the feudal monarchy (Putnam, 1993).
The persistence of specific clientelistic cultural values in these regions is a result of their historically ingrained hierarchical structure, in which individuals – and not the authority – typically oversee other morally unscrupulous officials (Banfield, 1958). In these areas, it is unusual to witness penalized politicians pursuing vested interests (Alesina & Giuliano, 2015).
The reasons for this – only apparently – surprising result can be found in the literature suggesting familism and clientelism in political participation in these areas (Foschi & Lauriola, 2016). There is extensive evidence that KTS developed a distinct and localized cultural response, a factor that is known to influence shifts in political power (Acemoglu & Robinson, 2021). Nevertheless, given the overlap of the former KTS territories with economically disadvantaged regions (Graziani, 1978), it could be argued that socioeconomic factors, beyond clientelistic networks, might impact the results.
Therefore, to address such concerns, we use an alternative instrument, namely, the results of the Italian 1946 Institutional Referendum (IR), which was conducted to determine whether Italy would adopt a Republican or Monarchic form after World War II. The referendum results is a robust measure of social capital (Crescenzi et al., 2013) representing long-term institutional characteristics of the past (Uleri, 2012). Historians suggest that the monarchic votes in the 1946 referendum were rooted in a "purely clientelistic" political tradition (Fonzo, 2019), symptomatic of the division between the Center-North, aligned with republican and secular traditions of the "Risorgimento" (Aimo, 2021), and the Center-South, which aimed to continue with the former KTS (Mastropaolo, 2016). In this sense, the votes for the monarchy at the referendum were a signal to continue previous institutional and cultural distinctions, characterized by aristocratic privileges, extractive rents, and political clientelism (Di Martino et al., 2020).
For a visual representation of the geographical distribution of monarchy preferences in the 1946 IR, please refer to Figure 3, which is based on the data available from the historical archive of the Ministry of the Interior ("Archivio Storico delle Elezioni").
The image illustrates the distribution of monarchy votes in the 1946 IR, revealing a distinct division between the North and South of the peninsula (Corbetta & Piretti, 2009). Instead of the previous instrumental variable, which denoted the simple overlap between electoral districts and the former territories of the Kingdom of the Two Sicilies (KTS), we now employ the rate of monarchic votes in the IR as a proxy for the inclination to perpetuate the clientelistic tendencies that prevailed in the KTS, as suggested in the literature mentioned above. This modification yields the primary empirical findings outlined below (refer to Table 4 for detailed estimates). The comprehensive set of estimations can be found in the Appendix (refer to Tables 11-14).
Table 4:
1st stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Monarchy votes IR 1946
|
0.00362***
|
0.00350***
|
0.00356***
|
0.00339***
|
0.00320***
|
0.00320***
|
0.00318***
|
|
(0.000955)
|
(0.000946)
|
(0.000942)
|
(0.000952)
|
(0.000937)
|
(0.000938)
|
(0.000933)
|
Cragg-Donald Wald first stage F statistic
|
14.34
|
15.29
|
16.97
|
16.06
|
18.05
|
17.52
|
21.21
|
Controls
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
Adjusted R-squared
|
0.065
|
0.067
|
0.070
|
0.068
|
0.073
|
0.070
|
0.085
|
Centered R-squared
|
0.0888
|
0.0910
|
0.0950
|
0.0938
|
0.0990
|
0.0969
|
0.112
|
|
|
|
|
|
|
|
|
2nd stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Party switch
|
-1.816**
|
-1.857***
|
-1.788***
|
-1.675***
|
-1.350***
|
-1.487***
|
-1.109**
|
|
(0.764)
|
(0.657)
|
(0.652)
|
(0.596)
|
(0.478)
|
(0.529)
|
(0.438)
|
|
|
|
|
|
|
|
|
Observations
|
629
|
612
|
599
|
584
|
569
|
557
|
536
|
Adjusted R-squared
|
-0.664
|
-1.363
|
-1.335
|
-1.266
|
-0.903
|
-1.287
|
-0.769
|
Centered R-squared
|
-0.621
|
-1.298
|
-1.270
|
-1.202
|
-0.847
|
-1.217
|
-0.713
|
First stage Cragg-Donald Wald F statistics
|
11.58
|
12.49
|
12.11
|
12.78
|
13.96
|
12.82
|
13.28
|
Alpha trim level
|
None
|
2.5%
|
5%
|
7.5%
|
10%
|
12.5%
|
15%
|
Clustered standard errors - at MP level - in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table reports the reduced form of the empirical coefficients for the first and second stages of the 2SLS regressions between party hopping (referred to as "parliamentary turncoat" in the Italian context) – instrumented with the proportion of votes for the monarchy in the 1946 institutional referendum – and parliamentary productivity (measured using the OpenPolis lawmaking index) – along with a number of controls included in a vector – based on data from 629 MPs in the XVI legislature (2008-2013) of the Italian Chamber of Deputies.
Table 4 presents a summarized view of the outcomes from both the first and second stages of our instrumental variable analysis, investigating the association between parliamentary turncoat and MPs’ productivity. The instrumental variable used in this analysis is the proportion of votes in favor of the monarchy in the 1946 IR. Across all seven columns, the instrumental variable – the share of monarchy votes in the IR – is consistently statistically significant. The F-statistics exhibit robust distribution, surpassing the 1% threshold and 15% maximal instrumental variable size, with p-values consistently below 0.01. The coefficients related to this instrumental variable are uniformly positive, signifying that a one-unit alteration in the share of preferences favoring the monarchy leads to an average 0.3 percentage point change in the likelihood of party switching.
Thus, in all model specifications, the empirical coefficients affirm a positive correlation between an increased proportion of pro-monarchy votes and the propensity for MPs to engage in parliamentary turncoat. The rationale for this initial-stage correlation can be traced back to the political subcultures of the former Kingdom of the Two Sicilies (KTS). The Kingdom's legacy gave rise to "dignitary networks capable of transcending the most profound political divisions in national history" (Adorno, 2020). The referendum itself stemmed from a longstanding ideological, social, and material schism between the North and South, wherein a deep-seated connection between the monarchy and a lengthy tradition of clientelistic rules persisted (Aterrano, 2020).
The referendum results were the signal of a localized hierarchical structure reminiscent of a feudal system with concentrated power, steeped in social practices that evolved over centuries (Morese, 2020a) that made the South, according to some historians, inherently "monarchic by instinct" (Cristina, 2020). This set of cultural and political values unmistakably favored the monarchy as the preferred institutional form for postwar Italy (Schininà, 2020). The prevalence of monarchic tendencies stemmed from the historical links with the Bourbon monarchy and the enduring system of clientelistic powers held by dignitaries, willing to adopt the most convenient approach to rebuild the KTS based on ambitions, personal interests, opportunism, and contingencies (Morese, 2020b). Therefore, it should come as no surprise that an increase in pro-monarchy votes in the 1946 referendum correlates with transformistic tendencies among MPs.
Eventually, the votes in favor of the monarchy in the 1946 Institutional Referendum aimed to reinstate the ancient aristocratic privileges of the KTS and its clientelistic system, regardless of the reigning House (Ridolfi, op. cit.). It is unsurprising that historians characterize this all-out push for monarchy as "new institutional transformism" (Gheda, 2020). The empirical evidence aligns with the literature on the distinctive Italian cultural values that underlie institutional change and the rules of the political game (Persson & Tabellini, 2021). The results reinforce the notion that culture evolves gradually (Gorodnichenko & Roland, 2021) and in the long term (Besley, 2020).
In the second stage, where we assess the impact of party switching on parliamentary productivity (lawmaking), all coefficients related to party switching are, again, negative. These coefficients hold statistical significance, implying that party switchers tend to exert a detrimental influence on parliamentary productivity. This indicates that the transformation of the party-switching dummy variable from 0 to 1 results in a change in the lawmaking productivity index, ranging between 110% and 185%, contingent on the specific level of dataset outlier trimming à la Koenker & Bassett (1978). Ultimately, these results confirms that institutional structures mold the organizational framework of politics (Besley & Persson, 2019).
Up to this point, our analysis has primarily focused on historical factors such as the boundaries of the KTS and preferences in the 1946 IR as potential indicators of the legacy of clientelism in specific regions. However, there might be a reasonable concern that these historical factors alone might not be sufficient to fully explain current political behaviors, given the considerable passage of time. To gain deeper insights into this issue and determine whether the influence of clientelism persists in these regions, we enhance our instrumental strategy by introducing an interactive term: the mean number of convictions for the crime of "concussione" in each electoral district during the years 2000-2013.
"Concussione" is an Italian legal term that roughly equates to embezzlement (Giommoni, 2021), referring to a form of extortion in which a civil servant blackmails someone else to obtain advantages in exchange for favors (Ferrari, 2003). Public corruption, including extortion, embezzlement, and misappropriation, has been widely acknowledged as a common occurrence (Bobbio, 1980) in a country where these practices have informally institutionalized over time, severely undermining the democratic stability of the system (Sberna & Vannucci, 2013). Concussione has become a symbol of politicians exceeding their political roles (Lorch, 1994) and serves as a sign of a pervasive landscape of political illegality involving all political groups (Vannucci, 2021).
There is literature suggesting that personal relationships between voters and politicians, which can deteriorate into clientelism (Asquer, 2015), have historically been more prominent in Southern Italy (Asquer et al., 2020). We aim to investigate whether this alleged legacy of clientelism, influencing party switching, is robust to ongoing dynamics of corruption and exploitation of public office for private interests.
To do so, we interact our previous instrumental variables with a more recent and concurrent measure of political corruption. This additional robustness check is intended to assess whether clientelistic trends, registered in the same time frame, substantially alter the statistical significance or the direction of our results.
Eventually, we still observe a significant relationship between historical clientelism and contemporary party switching (see a reduced form of the results in Table 5 and Table 6, and the complete results in the Appendix, Table 17 and Table 18).
Table 5:
1st stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Interaction Concussione & Kingdom Two Sicilies
|
0.0535***
|
0.0582***
|
0.0589***
|
0.0604***
|
0.0646***
|
0.0620***
|
0.0610***
|
|
(0.0149)
|
(0.0154)
|
(0.0153)
|
(0.0155)
|
(0.0156)
|
(0.0158)
|
(0.0160)
|
Cragg-Donald Wald first stage F statistic
|
15.47
|
17.25
|
17.70
|
18.21
|
20.70
|
15.47
|
17.25
|
Controls
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Observations
|
617
|
583
|
577
|
569
|
544
|
523
|
502
|
Adjusted R-squared
|
0.067
|
0.070
|
0.073
|
0.072
|
0.077
|
0.072
|
0.080
|
Centered R-squared
|
0.0910
|
0.0956
|
0.0986
|
0.0983
|
0.104
|
0.101
|
0.109
|
|
|
|
|
|
|
|
|
2nd stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Party switch
|
-1.372**
|
-1.192***
|
-1.070***
|
-1.070***
|
-0.871***
|
-0.815**
|
-0.527*
|
|
(0.626)
|
(0.440)
|
(0.411)
|
(0.396)
|
(0.337)
|
(0.341)
|
(0.303)
|
|
|
|
|
|
|
|
|
Observations
|
617
|
583
|
577
|
569
|
544
|
617
|
583
|
Adjusted R-squared
|
-0.351
|
-0.518
|
-0.434
|
-0.480
|
-0.344
|
-0.351
|
-0.518
|
Centered R-squared
|
-0.316
|
-0.476
|
-0.394
|
-0.439
|
-0.305
|
-0.316
|
-0.476
|
First stage Cragg-Donald Wald F statistics
|
15.47
|
17.25
|
17.70
|
18.21
|
20.70
|
15.47
|
17.25
|
Alpha trim level
|
None
|
2.5%
|
5%
|
7.5%
|
10%
|
12.5%
|
15%
|
Clustered standard errors - at MP level - in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table reports the reduced form of the empirical coefficients for the first and second stages of the 2SLS regressions between party hopping (referred to as "parliamentary turncoat" in the Italian context) – instrumented with the interaction between the mean of concussione convictions (between 2000-2013) and a dummy equal to one if the parliamentarian was elected in the territories of the former Kingdom of the Two Sicilies – and parliamentary productivity (measured using the OpenPolis lawmaking index) – along with a number of controls included in a vector – based on data from 629 MPs in the XVI legislature (2008-2013) of the Italian Chamber of Deputies.
Table 5 provides a detailed presentation of the results from the initial and subsequent stages of the 2SLS regressions that use the interaction between "concussione" convictions and a dummy variable that equals one if the parliamentarian was elected in the territories of the former KTS.
In the first stage, across all seven columns, the interaction variable involving "concussione" convictions and the KTS is consistently statistically significant, with positive coefficients. These results indicate a positive connection between historical and concurrent clientelism and party switching. The first stage F statistics is again higher that the Stock-Yogo weak IV critical values for a 15% maximal IV size, across all specifications.
In the second stage, which investigates the impact on parliamentary productivity, the coefficients for party switching are uniformly negative and statistically significant. These findings confirm that party switchers tend to be less productive in the legislative process. The magnitudes of these coefficients vary depending on different alpha-trimmed levels, emphasizing, nonetheless, the negative impact party switchers have on parliamentary productivity.
Table 6:
1st stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Party switch
|
Interaction Concussione & Monarchy votes IR 1946
|
0.000970***
|
0.000988***
|
0.000976***
|
0.00100***
|
0.00103***
|
0.000951***
|
0.000927***
|
|
(0.000264)
|
(0.000271)
|
(0.000271)
|
(0.000275)
|
(0.000279)
|
(0.000284)
|
(0.000286)
|
Cragg-Donald Wald first stage F statistic
|
14.60
|
15.08
|
14.65
|
15.18
|
16.52
|
14.28
|
14.06
|
Controls
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Yes
|
Observations
|
617
|
583
|
577
|
569
|
544
|
523
|
502
|
Adjusted R-squared
|
0.065
|
0.065
|
0.067
|
0.066
|
0.067
|
0.061
|
0.069
|
Centered R-squared
|
0.0892
|
0.0908
|
0.0926
|
0.0921
|
0.0946
|
0.0899
|
0.0983
|
|
|
|
|
|
|
|
|
2nd stage
|
|
|
|
|
|
|
|
|
(1)
|
(2)
|
(3)
|
(4)
|
(5)
|
(6)
|
(7)
|
VARIABLES
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Lawmaking
|
Party switch
|
-1.781***
|
-1.615***
|
-1.596***
|
-1.448***
|
-1.157***
|
-1.305***
|
-0.889**
|
|
(0.680)
|
(0.553)
|
(0.556)
|
(0.507)
|
(0.410)
|
(0.469)
|
(0.389)
|
|
|
|
|
|
|
|
|
Observations
|
617
|
583
|
577
|
569
|
544
|
523
|
502
|
Adjusted R-squared
|
-0.636
|
-1.011
|
-1.047
|
-0.928
|
-0.647
|
-0.986
|
-0.485
|
Centered R-squared
|
-0.594
|
-0.956
|
-0.990
|
-0.874
|
-0.599
|
-0.925
|
-0.438
|
First stage Cragg-Donald Wald F statistics
|
14.60
|
15.08
|
14.65
|
15.18
|
16.52
|
14.28
|
14.06
|
Alpha trim level
|
None
|
2.5%
|
5%
|
7.5%
|
10%
|
12.5%
|
15%
|
Clustered standard errors - at MP level - in parentheses
*** p<0.01, ** p<0.05, * p<0.1
Note: The table reports the reduced form of the empirical coefficients for the first and second stages of the 2SLS regressions between party hopping (referred to as "parliamentary turncoat" in the Italian context) – instrumented with the interaction between the mean of concussione convictions (between 2000-2013) and the proportion of votes for the monarchy in the 1946 institutional referendum – and parliamentary productivity (measured using the OpenPolis lawmaking index) – along with a number of controls included in a vector – based on data from 629 MPs in the XVI legislature (2008-2013) of the Italian Chamber of Deputies.
Table 6 reinforces the conclusions drawn from Table 5. The first stage coefficients illustrate a positive relationship between the interaction of historical clientelism, represented by monarchy votes in the 1946 referendum, and current clientelistic trends, related to concussion. Once again, the coefficients are statistically significant across all seven columns, underscoring the positive relationship between forms of historical and recent clientelism and party switching. The interaction variable again exhibits robustness, as confirmed by the first stage F-statistic.
In the second stage, where we examine the impact on parliamentary productivity, the coefficients for parliamentary turncoat are again negative and statistically significant. The results reaffirm the notion that parliamentary turncoat has an adverse effect on lawmaking productivity. The results are consistent even if we combine other forms of immoralities related to corrusptive crimes (see Table 19-20, in the Appendix).
To summarize, regardless of the specific empirical specifications used, all empirical findings consistently underscore a significant and negative impact of parliamentary turncoat on the productivity of lawmakers. From an economic perspective, party-switching within the legislature can be considered a rational and strategic behavior (Snagovsky & Kerby, 2018), enabling individuals to secure greater personal benefits or attain positions of higher prestige. Legally, parliamentary turncoat is well within the bounds of the constitution.
Nevertheless, this freedom negatively affects the primary role of MPs to foster the legislative power and represent their electorate mandate within a specific and designed coalition. Therefore, the problem is not only that MPs have no qualm in switching from the right to the left without hesitation, and vice versa (Montanelli, 2006), but deviates from MPs’ primary societal obligation (Siddali, 2015) and core responsibility (Reid, 1980) of representing the legislative power. The erosion of representatives has contributed to the diminishing the image of the parliamentary institution (Morelli, 2018), undermining the expected programmatic coherence for constituents (Massari, 2009) and proving detrimental to the legislative function that citizens anticipate their representatives should fulfill as parliamentarians (Pansardi, 2015).
While parliamentary turncoat may be viewed as a response to evolving aspects of civil society (Principato, 2012) and may contribute to changes in parliamentary composition aligned with constituents' preferences, it is crucial to recognize that MPs should not be obligated to heed all constituents' demands (Pacini et al., 2021). Instead, the legislative power is expected to function as a tool for government oversight and direction (Franchino, 2006), serving to constrain executive contingent rule-making (Rose-Ackerman et al., 2018).
From a policy perspective, the constitutional legality of parliamentary mobility – originally set right after the fascist discipline of economic and political elites (Stah & Popp-Madsen, 2022) – has faced widespread criticism (Pinelli, 2022, 2018; Sulpizi, 2022; Belletti, 2020; Mannino, 2002). Recently, the parliament has taken initial measures, implemented only in the Senate, to minimize parliamentary turncoats (Pinto, 2023).