Research

Polarization and Government Effectiveness: A look into European coalitions

Section 1. Introduction

Governance can be a tricky  thing to measure, not least because it houses an umbrella  of different ideas. Governance is sometimes thought of as the ability of a nation’s government to formulate and implement public policy, however the institutions in which authority is vested in matters, just as surely as how well a government takes care of its’ citizens sense of well-being and safety matters. As there are many parts to the whole that makes up governance, then there are many determinants of government effectiveness. This paper will focus on one potential influence on effectiveness, intra-coalition ideological polarization.

Coalition  formation   is  influenced  by  party  characteristics.  A  coalition  potentially  houses different ideologies under one roof. Jozwiak & Schneider (2006) find that cabinet duration, the length of the coalition’s stay in power,  is negatively  related to the degree of polarization. Greater polarization within a coalition leads to shorter cabinet duration.   It follows that the “degree of  polarization has implications for  the degree of  conflict and cooperation within society” (Inridason  2010).  Groups that are closer to each other in ideology  are less  likely to disagree with each other. When it comes to coalition governments that would mean that more closely-aligned  parties  are  willing  to  negotiate  and  push  through   policy.  In  a  society, individuals that are perceived to have similar values and beliefs are seen to be less of a threat to one another. The focus of this paper then, is intra-coalition polarization and how it affects the ability of a coalition to govern effectively.

Riker’s  (1962) seminal  work  in coalition  theory  determined  that  coalitions  are minimum- winning.  Parties that are not instrumental to coalition formation are left out. At the same time, Leisersen (1968) found  that large coalitions are less stable because of the potential for policy disagreements. We can apply this train  of  thought  to  polarization. More  highly-polarized coalitions  have the potential  for increased policy  disagreements.  This  would only inflate the time needed to come to policy  agreements and push through  policy.  I posit  that increased polarization within coalitions would negatively impact the ability of governments to effectively formulate and implement policy.

In  order  to  understand  the  important  of  polarization  within coalitions  and  its  ability  to influence  government  effectiveness, I  conducted  a  cross country  study  of  41 European countries. Data was gathered regarding the coalition composition of each one of the countries over a time period of 16 years, albeit with gaps.  The ideological positions of parties within coalitions were noted, so as to measure the degree of polarization, as well as the number of parties and the effective coalition size. The idea was to run a simple regression on a constructed measure of government effectiveness against the above mentioned variables as well as a few others mentioned in section 2, which details the data used in this paper.

The measure for government effectiveness is  constructed using  surveys, questionnaires, and various other assessment tools. Due to the nature of the data, it seemed prudent  to compare estimates yielded from regular panel data as well as dynamic panel data analysis. The finding of this paper is that there is nothing in the analysis to suggest that polarization has an impact on government effectiveness, that is, unless we are discussing developing countries. Even then, it is not clear whether polarization is at fault, or a political culture and institution that has been in place years and maybe even decades before a coalition has the chance to be born.

In  developed  countries, coalition size and government  effectiveness are positively linked. Coalition building in these nations is a careful process, where parties are likely to build medium sized,  ideologically  close,  coalitions  with an average of three parties. The process has been streamlined, perhaps the word  ‘perfected’ is strong, but it appears the relative stability of the institutions of developed countries are key in preserving future stability.

On the other hand, developing nations were more associated with lower effectiveness scores as well as lower rule  of law scores. Increases in coalition  size  and the degree of polarization yielded decreases  in effectiveness.  This suggests that already-present issues  associated with corruption  might hurt  political  stability  in these countries  and that in areas of less  political stability – there is more leeway for larger, more ideologically distant, coalitions to form.

The paper is organized as follows. Section 2 provides an in depth look at the different variables, dependent and independent, used in the paper and details the regression frame-work  utilized. Section 3 presents the results, Section 4 provides extensions, and Section 5 ties everything together.

Section 2

2.1. Regression frame-work

We will use the insights gleaned from  the discussion above to create a simple   theoretical framework.  The idea is that if government effectiveness is merely  a product of policies chosen and the subsequent implementation of these policies (Ahlerup, Hansson 2010), then it stands to reason that any potential for gridlock in coalition bargaining due to increased polarization or coalition size could influence government effectiveness negatively . The aim of the paper is to understand how coalition composition affects effectiveness.

The empirical analysis can be separated into two parts. The initial analysis makes use of regular OLS methods and then because of the nature of the data I chose to make use of a dynamic model.  The data collected on government  effectiveness is culled from  yearly surveys; the perceptions ,on quality of government policy implementation and formulation, of which are likely to be adjusted based on previous years’ calibrations.

The regression takes the form of

research0

 

when modeling dynamic panel data.

2.2. A cross-country study

In western Europe, coalition cabinets are the most common form  of government.   I test the hypothesis  using  data  on  the  composition  of  coalition  governments  across 41 European countries between the years 1996 to 2011. The resulting time series has a total of 533 (41×13) observations.

Table 1 provides a glimpse into the incidence and type of coalitions within European parliaments. The breakdown details whether, across the time period in question, a country has single party or coalition rule and whether or not the government formed by that coalition/party ruled with majority/minority or partisan leadership.

As we can see, in the examples above, the most common form of government takes the form of a coalition, majority or otherwise. The Czech Republic has the only incidence of non-partisan coalition, and Malta  stands out as the only country  to have had only majority single party leadership throughout.

The data on coalition composition comes from  Elections in Europe: A  Data Handbook,  a comparative text on elections and electoral systems.

research1

2.3. Government effectiveness

I use a composite variable as the dependent variable in this study.  Worldwide Governance Indicators is an index built by the world bank to evaluate country governance. Based on long- standing  research, WGI  identifies  six  dimensions  of governance. Voice  and Accountability, Political Stability and Lack of Violence, Government effectiveness, Regulatory Quality, Rule of Law and Control of corruption.

The index provides an empirical measure of governance in over 200 countries and territories. The indicators  were  updated  every  two  years between 1996-2002  and  on  a yearly  basis thereafter. The WGI data is based on perceptions based governance data sources; using surveys and other subjective  assessments from  various  sources – ranging  from  NGOs to multilateral organizations.  The many different measures are combined into a composite indicator that is reflected in the yearly estimate for each of the six identified dimensions of governance.

The composite dependent variable in this paper, Government Effectiveness, is constructed simply by using the average of the yearly estimates from  the WGI  data set for  both Government Effectiveness and Regulatory Quality; as both capture perceptions of the ability of government to formulate policy and the quality of said policy.

2.4. Polarization

All the values for  government  polarization are calculated using data from  Mapping Policy Preferences II, which provides estimates on the ideological placement of political parties on the left-right continuum.  The book only provides data from 1990-2003 and so additional data was culled from the related online database created by Manifesto Research Group.

Table 2 lists 8 countries with the highest and lowest values of coalition polarization. Only countries with a polarization score greater than zero were included, excluding countries with exclusively single party rule throughout the time period 1996-2011. Unsurprisingly many of the countries, including the United Kingdom, Moldova, and Portugal, with the lowest scores are ones that have had a mixture of single party and coalition leadership. Ireland, and Italy are the standouts;  having  had  multi-party  coalition  leadership throughout.   Italy   is  a  surprising inclusion, having had a relatively high average of 4.69 parties.

research2

On the high list, Slovakia and Bulgaria are the run-away winners while the rest of the pack are closely bunched together. Switzerland’s ‘magic formula’ policy of dividing their executive seats among four ruling parties give them an inherent and automatically high degree of polarization.

To get at the key question of whether polarization affects government effectiveness, there needs to be a way to measure the level of polarization within government coalitions. If we think about the political spectrum, we intuitively understand that polarization is at its highest when two parties exist on either ends of the continuum. A strict ideological division measure, which relies on calculating the distance between the parties in question can yield misleading measures, especially when small extremist parties are present (Indridason, 2010).

Naturally, a weighted  measure of polarization is required. In this paper I use the family of measures developed by Esteban & Ray (1994). The measure is based on 4 underlying axioms that attempt to capture an intuitive understanding of the term ‘polarization’.  The authors show that the only measure that satisfies their axioms (for a detailed statement of the axioms please refer to Esteban & Ray (1994))  takes the form of :

research3

where research18ߨ௜   is the size of, and yi is the ideological   position of party i. research17   is the ideological distance between party i and j. K is a scalar, which  is set to 1 in this paper and  α is a measure of the  sensitivity  to  polarization.  α  is  fixed  for  the  purposes of  this  analysis  and  set at 1.3 (Indridason, 2010).

I  included dummy  variables for  low, medium and high measures of polarization. Table 3 provides descriptive statistics on polarization. After the computation for polarization for every country coalition in the data set, the values were ranked in ascending order and I picked the 50th percentile value , which turned out to be 1.05. All values from 0-1 were then classified to be a low measure of polarization. A coalition with a polarization value, 1<P<2, was considered to display medium levels of polarization, and P>2 meant the coalition fell into the high bracket. Note that 75% of the observations on degree of polarization fall between 0 and 2.4. The majority of countries in the sample have had coalitions with low and medium levels of polarization.

reserach4

2.5. Coalition size & Effective size

To understand whether or not the size of a coalition has the potential to destabilize a political system, we need to define coalition size. In addition to including a variable for coalition size, the number of parties within a ruling bloc, we must include a dependent variable that takes into account their relative size.  Thus, effective size reflects the total number of seats the coalition holds in comparison to the entire parliament. The larger the coalition is, the tighter their hold over the issues of the day.

I also include various other independent variables in the paper, such as the natural log of GDP, the Worldwide Governance indices for Corruption,  Rule of law, as well as a lagged dependent variable in the dynamic modeling part of the analysis. In addition there is a variable called GDP/Can11; it represents GDP numbers for different countries divided by the GDP of Canada for 2011 as a way of setting a standard for comparison.

Section 3. Results

3.1. Summary Statistics

Table 4 on page 12 presents the descriptive statistics for  a sample of the countries in the analysis. The mean effective size of a coalition government stands at 0.56, indicating the most commonly  occurring  form  of government  is  majority rule.  The lowest  scores are found  in Russia, which holds the minimum effective size in the sample (0.12). This, while strange, can be understood to be a product of the country’s semi-presidential system where the second most powerful official  in government is  appointed  by the president.  Low  scores are found  in the Scandinavian countries, where minority governments are common. On the other end of the spectrum  is  tiny   Liechtenstein  which  owns  the  largest  score  for   effective  size  (0.96). Liechtenstein has only three parties in current existence, two of which dominate the political landscape, which helps explain the near total control of parliament.

The mean coalition  size is  2.8 and the median is  3. Again there is plenty of variation in the sample. Countries such as Italy, Latvia, Belgium, and Albania to name a few as identified as having a history of large coalitions incorporating a number  of parties   (often 5 or more). Meanwhile countries such as Spain, the United Kingdom and Greece have had mostly single party  governments  between  1996 and  2011. Looking  at  the  statistics  for  the  degree of government  polarization,   we can see that the sample  is  right skewed, and noting  that the median degree of polarization is 1.05, we can now paint a picture. The most common form of European government  is a majority cabinet coalition government  with a medium level of polarization.

research5

We would expect to  see the perception  of  government  effectiveness increasing as general political and social conditions, as measured the indices for rule of law and corruption, improve.

And  as expected, Table 5 shows us that Rule of Law and Corruption  are highly correlated (0.8672, 0.9224) with Government effectiveness. The initial findings are encouraging, coalition size and polarization have negative relationships with government effectiveness while effective size  has a positive  link to effectiveness. A  possible  explanation  for  the negative correlation between coalition size and effectiveness is that larger coalitions are commonly more polarized than coalitions made up of fewer parties.  The negative link of polarization with effectiveness supports this.

research6

 

research7

Column 1 of table 6 tells us that the more parties in a coalition, the more polarized that coalition is likely to be. The same is true of effective size. We could posit that to gain substantial control over parliament, powerful parties will align with other parties, but not with ones that are too ideologically distant . Keep in mind, the average coalition in our sample is characterized by a medium degree of polarization and is typically made up of 2 to 3 parties.  Straying from this configuration appears to be a bad idea, but of course there are already exceptions that can be identified. Italy  stands out  as one such example; a country  that  has historically had large coalitions made up of ideologically similar parties.

3.2. OLS, Random Effects and Fixed Effects

Let us start our analysis  with regular OLS regression  using government effectiveness as the dependent variable and including coalition size, effective size, ln GDP, Corruption, Rule of Law (RoL), and the square of Rule of Law as independent variables. I included the square of Rule of Law to try and get at the true nature of the relationship between Rule of Law and Government Effectiveness. As noted earlier, it is rare for coalitions to be made of homogenous blocks; if one imagines a coalition consisting of just one party,  then the inclusion of an extra party  in a coalition naturally increases the polarization of your faction. This is obvious of course, but is a good way to think about why polarization is increasing with the increasing coalition size. And although increased coalition size is associated with less government effectiveness, as shown in the second column of table 7, the effect is not significant at the standard levels of significance (p- value 0.095). It should also be noted that the effect associated with an increase in effective size was higher than initially expected (0.4762).

The result is statistically significant (p -value 0.009) and the initial regression analysis tells us that gaining substantial control over parliamentary proceedings through a larger effective size has a large impact on government effectiveness. It makes sense that having a near-monopoly over government would allow coalitions to push through legislation with little opposition.

To sum up, while the effect of coalition size is not statistically significant it still gives us an important result.  To increase  the effectiveness of government, the governing  coalition  must limit the number of parties while simultaneously maintaining majority control over parliament.

research8

It was expected that Rule of Law and Corruption  would be highly correlated with Government effectiveness. The results  are somewhat  surprising.  While  Corruption   is  highly positively correlated with effectiveness (1.4233), the same does not hold true for Rule of Law (-0.3636). The data on both Corruption  and Rule of Law come from  surveys in which respondents answer questions through  a ranking procedure. Higher ranks always indicate  a better standard of living, whether it is measured through Corruption,  Government Effectiveness, Rule of Law or Regulatory Quality. Thus, we would expect improved confidence in how the rules of society are uphold to be associated with improved government effectiveness. Instead, the results presented in Table 2 show a clear nonlinear relationship between Rule of Law and Government Effectiveness. This is the reason why the square of ROL was included in the analysis, to get at the true relationship between ROL and GE. The positive link (0.521) shown in table 7 of the square of Rule  of Law  and effectiveness show that there is  a linear  relationship  at higher, positive values of Rule of Law.

Fig. 1 illustrates the parabola shaped relationship, For countries in the lower end of the Rule of Law measure, corresponding to the lowest quartile of the sample, a perceived increase in how well the rules of law are being upheld is not necessarily associated with increased government effectiveness. Depending on where a country starts out, effectiveness could either increase or decline.  Nations  with a rule  of  law index  very  close  to  0 would expect an  increase  in effectiveness with perceived increases in the quality and integrity of law enforcement.

research9

Fig.2  demonstrates the link between the ln(Rule of law) and effectiveness. The rate of increase in effectiveness associated with the natural log of rule of law takes the form of an exponential curve. The 40th percentile corresponds roughly with ln(Rule of law) = 0.03. If rule of law were the main driving force behind government effectiveness, it would appear that the majority of the countries in the sample can increase government effectiveness at a near exponential rate by focusing on improving the standard of the rule of law. This is a striking proposition and I expand later on regarding the idea of improving existing institutions within a country in a bid to increase overall effectiveness.

research10

The effects of  inclusion of  the measure of  coalition polarization are presented in table 8. Polarization  has a slightly  positive  link with effectiveness, although  the effect is  statistically insignificant. The signs of the effects associated with coalition size and effective size have not changed, and thus are in accordance with the results found in the earlier part of this section. The test predicts that the fixed effects model provides the best fit when (Rule of law)2  is omitted while random  effects is safer when  it is included. The fixed effect model  has non-random parameters, and we see that coalition size is negatively correlated at significant levels (-0.033, 0.024). At the same time, the effect of effective size is not as pronounced (0.2073) as in previous specifications of the model.

The alternating signs of the different coefficients for coalition size and the fluctuations between statistically significant and insignificant outcomes are likely to indicate a non linear relationship between coalition size and government effectiveness. Figure 3, which plots the values of Government Effectiveness against that of ideological polarization (natural log values), shows us that nearly every level of government effectiveness can be reached at any coalition size, suggesting independence. The negative relationship that we have found in the analysis appears to be the result of a lack of observations at the higher levels of coalition size.  However, the fixed effects model would suggest that the negative correlation would remain with more observations for coalitions made up of more than six parties.

research11

research12

The values for the degree of polarization are not as stratified as that of coalition size, but we can see again that a lack of observations at the higher end of the spectrum is again impeding a complete analysis. If anything, excluding the tail observations, the best fit line would have a positive slope. As we saw earlier, polarization does appear to have a positive correlation with effectiveness – if at statistically insignificant levels.

3.3 Dynamic panel data estimation

Dynamic panel data models are useful when the dependent variable depends on its own past realizations (Branas-Garza, Bucheli, Garcia-Munoz 2011). In this case, GLS and OLS estimates are biased  and  inconsistent.  The Generalized  Method  of  Moments  (GMM)  however  will produce  unbiased  estimates in  this  case. Let  us  examine  the  results  when  we  regress Government  Effectiveness while  including  a  temporally  lagged  dependent  variable.  The variable, lag-government effectiveness, is lagged by one year.

The results, displayed in table 9, do not shed much more light than the initial static panel data analysis.  Rather, it only serves to strengthen the conclusions  that we arrived at during the previous investigations. It seems hard to draw any true conclusions about the impact coalition size, effective size, and the degree of polarization have on government effectiveness. The signs on coalition and effective size are changing, and none of the results are valid at the usual significance level.   The introduction of the lagged variable dramatically reduces the effect of effective size on government effectiveness, keeping it more in line with the results presented in bold for the random effects model in table 8.

If anything, looking at column 2 in table 9, we can say that the degree of polarization has a slight, if imperceptible, positive effect on government effectiveness (0.008). This trend shows up across the board, in both the results of the dynamic model estimation and earlier results in static panel model analysis using regular OLS, fixed effects and random effects.

research13

Section 4. Extensions

4.1. Developing and Developed countries

The mean government effectiveness score for developed countries is 2.11 and for developing countries  that drops by about 1.87. The mean effectiveness score for  countries  with a high degree of polarization  was 1.88, decreasing  by 0.67 and 0.71 for medium and low levels  of within-coalition polarization.

Gov Effectiveness = 1.88 -0.67(MediumPolar) – 0.71(LowPolar) Gov Effectiveness = 2.11 -1.87 (Developing)

The final element of  the analysis involves examining and determining whether  there is a difference  between developed  and  developing  nations  in the  sample.  After  noticing  that removing (Rule of law)2, and thus all large and positive values, from the regressions yielded results suggesting a negative link between coalition size and  government  effectiveness; it seemed likely that there might be a divide in the sample between developing and developed nations. A developing nation is more likely to have lower Rule of law scores and thus focusing on developing and developed nations separately might provide insight on the matter.

Figure 5 shows us that developed nations have exclusively positive rule of law scores while developing nations have both positive and negative scores. Clearly a rule of law score below 1 is more likely to belong to a developing nation. Regressing on Rule of Law, the results in table 10 also show that while the mean Rule of law score is calculated to be 0.5654 and 0.0752 in OLS and GMM  estimations respectively, there is a stronger negative link attributed to Developing nations versus Developed ones ( rows 4 and 6 in table 10).

research14

4.2. The divide

After sub-setting the data, tailoring it to include only observations on developing nations – the results show us the predicted negative link between coalition size and government effectiveness. The OLS estimate in column three in table 11 predicts a correlation of -0.02, and the corresponding GMM estimate stands at -0.005, albeit at non-significant levels. This remains an important outcome however; in developing  countries,  increased coalition  size is  likely to lead  to  decreased government  effectiveness. The  corresponding  analysis  for  developed countries show positive links with coalition size and effectiveness, lending some credence to the hypothesis of a divide in the sample.

research15

 

research16

Section 5. Concluding Remarks

The theoretical  literature  establishes  that  polarization  has the  potential  to  affect  political decision  making  (Lindqvist,  Ostling  2010). Tsebelis  (2002) shows that  increased  ideological distance among a coalition of several parties can lead to increased gridlock, as realized by a fewer number of changes in the national budget. However, it is also the case that, depending on the  assumptions  made,  both  positive  (Azzimonti  2011) and  negative  (Svennson  1998) relationships between polarization and size of government can be rationalized.

The empirical  evidence  suggests a causal, if tentative,  link between government  size,  as measured by level of public spending and polarization. In similar spirit, this paper also suggests a positive correlation  between polarization and government  size, whether measured by the number of parties in a coalition or the coalition’s effective size.  Unfortunately  the results in this study  were  not  able  to  find any  conclusive  evidence  of  any  causality  with government effectiveness.

What we were able to arrive at is a clearer picture of how European coalitions have operated over the last sixteen years. The most common form of European coalition takes the form of a majority coalition; usually comprised of 3 parties with a medium level of polarization. At most we can infer from the findings presented in the results section that an increase in coalition size and polarization will likely lead to slightly increased government effectiveness.

It  appears that the independent variables chosen are not ideal for  explaining trends in the change of government effectiveness over time. It must be noted that the sample   has a short time  period  (T=13) relative  to  population  (N=41),  which  makes identifying  trends  more difficult. The high correlation between government effectiveness and the other included indices from  the Worldwide Governance Indicators project suggests that  increases in government effectiveness may be due to institutions that were put in place years before the increase is ever recorded.   The U shaped relationship between Rule of Law  and Government Effectiveness scores indicated the presence of a divide between the developing and developed countries in the sample. Developing countries are more likely to have lower effectiveness and Rule of Law scores. Furthermore, coalition  size is  negatively correlated with Government Effectiveness in developing countries. We could infer that cabinet formation in developing countries should aim to minimize coalition size and the degree of polarization that goes hand in hand with increasing size of government. The decreased opposition would serve to push through policies that aim at increasing effectiveness.

Developed countries, contrastingly, appear to increase government  effectiveness when polarization and coalition size increase. However, the result is misleading. Developed countries are more likely to form ideologically close, medium sized coalitions. The stability provided by already existing institutions enable these nations to get on with the task of governing. In developing countries it may be that the incidence of more polarized coalitions are causing lower effectiveness scores but the lack  of causality  leads  this  researcher to conclude  that it is  the presence of corruption that is interfering with governance.

In conclusion, it does not appear that polarization matters in the big picture, when it comes to government effectiveness. Large coalitions are rare, and most parties aim to create factions of low to medium polarization. The increases in perceptions of effectiveness are most likely due to improvements  in the  quality  of  life  that  were  put  in place  years before the  increase  in effectiveness is measured.

 

 

References

Ahlerup, Pelle,  and Gustav Hansson. “Nationalism  and government effectiveness.” Journal  of

Comparative Economics 39.3 (2011): 431-451.

Arellano,  M.  and S.  Bond (1991). Some  tests of  specification  for  panel data: Monte  Carlo evidence and an application to employment equations, Review of Economic Studies 58:277-297.

Azzimonti,   Marina.  “Barriers  to  investment  in polarized  societies.” The  American  Economic

Review 101.5 (2011): 2182-2204.

Brañas-Garza, Pablo,  Marisa  Bucheli,  and Teresa Garcia-Muñoz. Dynamic  panel  data:  A useful technique in experiments. No. 10/22. Department of Economic Theory and Economic History of the University of Granada., 2011.

Dahl, Robert A. On democracy. Yale University Press, 2000.

Esteban, Joan-Maria & Debraj Ray. 1994. “On the Measurement of Polarization.” Econometrica

62(4):819{851.

Holtz-Eakin,  D., W. Newey  and H. S.  Rosen (1988). Estimating  vector autoregressions  with panel data. Econometrica 56. pp. 1371 – 1395.

Indridason,  Indridi H.  “Coalition  formation   and  polarisation.” European  Journal  of  Political

Research 50.5 (2011): 689-718.

Jozwiak, Eva, and Gerald Schneider. “Polarization and Government Survival: A Comparative

Analysis.” manuscript, University of Konstanz (2006).

Kaufmann,   Daniel,  Aart   Kraay,  and  Massimo  Mastruzzi.   “The  worldwide  governance indicators: methodology  and analytical issues.” World Bank Policy Research Working Paper 5430 (2010).

Laakso, Markku, and Rein Taagepera. “Effective number of parties: A measure with application to West Europe.” Comparative political studies 12.1 (1979): 3-27.

Leiserson, Michael. 1968. “Factions and Coalitions in One-Party Japan: An Interpretation  Based on the Theory of Games.” American Political Science Review 62(3):70{87.

Lindqvist,     Erik,     and    Robert    Ostling.     “Political     polarization     and    the    size     of government.” American Political Science Review 104.3 (2010): 543-565.

Mapping Policy  Preferences  II.  Estimates for  Parties, Electors,  and Governments in Eastern

Europe, the European Union and the OECD, 1990-2003 Oxford: Oxford University Press

Mershon. 1996. “The Costs of  Coalition:  Coalition  Theories  and Italian Governments.” The

American Political Science Review , Vol. 90, No. 3 (Sep., 1996), pp. 534-554

Riker,  William H. 1962. The Theory of Political  Coalitions.  New Haven, NJ:  Yale University

Press.

Roodman, D. (December 2006). How  to do xtabond2: an introduction  to “Difference”  and

“System” GMM in Stata. Center for Global Development Working Paper Number 103.

Strøm,   Kaare,  and  Wolfgang   C.  Müller.  Torbjörn   Bergman,  eds.  2003.”Delegation and accountability in parliamentary democracies”.

Strøm, Kaare, Wolfgang C. Müller, and Torbjörn Bergman. “Cabinets and coalition bargaining:

the democractic life cycle in Western Europe.” (2008).

Svensson, J. 1998.  “Investment, Property Rights and Political Instability: Theory and Evidence.”

European Economic Review 42 (7): 1317–41.

The   Manifesto    Data   Collection.    Manifesto    Project   (MRG/CMP/MARPOR)       Berlin: Wissenschaftszentrum Berlin für Sozialforschung (WZB)

Tsebelis,   G.  2002. Veto  Players:  How   Political  Institutions   Work.  Princeton,   NJ:   Princeton

University Press.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s