Presentation on theme: "Fuzzy Sets Analysis: Applications for Comparative Corporate Governance Gregory Jackson Kings College London Presentation at Oxford University ESRC Methods."— Presentation transcript:
Fuzzy Sets Analysis: Applications for Comparative Corporate Governance Gregory Jackson Kings College London Presentation at Oxford University ESRC Methods Festival July 19, 2006
2 Corporate Governance Research Corporate governance –Corporate governance consists of the rules and beliefs that shape the relationships among stakeholders in the decision-making and control over firm resources. Comparative research –What differences in corporate governance exist across countries? –Why did diverse sets of institutions emerge historically, and change over time? –How does institutional diversity impact firm strategy and structure, performance and power of different stakeholders?
3 Methodological Issues Only two models across countries? –Conclusions drawn from strong correlations among elements of corporate governance in Anglo-American countries –But diverse configurations in other countries that are not well understood Methodological issues –Large number of variables, but small N –Much literature is based on variable based study of correlations, only a case-based attempts to explain why countries differ –conjunctions of different causal factors (e.g. Complementarities) –multiple paths to same outcome possible Most work in this area largely ignores these issues!
4 Application Explaining employee representation on corporate boards Jackson, Gregory, "Employee Representation in the Board Compared: A Fuzzy Sets Analysis of Corporate Governance, Unionism, and Political Institutions". Industrielle Beziehungen, Vol. 12, No. 3, pp. 1-28, 2005 Available at SSRN: http://ssrn.com/abstract=800525 http://ssrn.com/abstract=800525
5 Table 1 Codetermination Rights in 22 OECD Countries: Fuzzy-set Memberships Fuzzy ScoreCharacteristicsCountries 0No constitutional rights, and no statutory or tri-partite regulation Australia, Canada, Japan, New Zealand, South Korea, Switzerland, United Kingdom, United States 0.1Constitutional rights, but no statutory or tri-partite regulation Italy, Portugal 0.3Some statutory or tri-partite regulation of public sector firms Belgium, Greece, Ireland, Spain 0.7Legal right to attend board at private firms. France 0.9Legal rights to nominate some members to the board. Finland, Netherlands 1Legal rights to board-level representation in private firms. Austria, Denmark, Germany, Norway, Sweden
6 Possible Explanations Industrial relations factors –Union strength –Union centralization Corporate governance factors –Stock market activity –Investor protection –Ownership concentration, bank influence Political factors –Consensus oriented political systems –Left-wing government Legal origin –Civil law vs. common law
7 Steps in the Analysis Code conventional statistical indicators as 7-point fuzzy sets Analysis of necessary conditions Analysis of sufficient conditions using truth table approach Back to cases!
8 Table 4 Results of Fuzzy-set Tests: Necessary Conditions for Board-level Employee Representation ========================================== N Cause Observed Binomial Variable >= Outcome Proportion p ========================================== cg-market 7 0.50 CG-MARKET 3 0.21 cg-legal 6 0.43 CG-LEGAL 4 0.29 concentration 1 0.07 CONCENTRATION 8 0.57 bank 7 0.50 BANK 8 0.57 union 6 0.43 UNION 7 0.50 coordination 7 0.50 COORDINATION 10 0.71 0.584 centre-left 6 0.43 CENTRE-LEFT 5 0.36 electproportion 7 0.50 ELECTPROPORTION 9 0.64 parties 4 0.29 PARTIES 8 0.57 Consensus 5 0.36 CONSENSUS 8 0.57 commonlaw 13 0.93 0.047* COMMONLAW 1 0.07 =============================================== Number of Cases Tested (Outcome > 0): 14 ( 63.6% of Total) Test Proportion: 0.70 *p < 0.10
9 Both no Outcome but no cause (no necessity) Cause but no outcome (no sufficiency) Both yes
10 Analysis of Sufficient Conditions Test various combinations of the nine variables –But no more than 6 at a time (A.Marx 2004)! Using truth tables proposed in Ragin (2004) –Coverage: the sum of consistent Xi divided by the sum of Yi. "unique" coverage is the calculation of non-overlapping coverage for each solution term. –Consistency: the sum of consistent Xi minus the sum of inconsistent Xi divided by the Xi. This rewards near misses, but penalizes big misses
11 The only significant result was coordinated collective bargaining. Coverage is strong (7 out of 8 cases coverage score of 0.865) and high consistency of 79.4%. Only France is unexplained. Thus, coordinated collective bargaining passes the test as a sufficient condition.
16 Robustness Do the thresholds for being in a category influence the results? –A lower threshold for ownership concentration or coordinated bargaining might make these factors necessary –But both factors are common to the two configurations meeting condition of sufficiency Missing cases –eight variables yield 2*8 = 256 possible configurations, of which only 20 configurations have empirical cases. –results become less robust because cases are not sufficiently similar to falsify a particular configuration –Experiments with random data suggest robustness as long as a certain ratio of variables to cases (5-6 variables are ok for 22 cases) (Marx 2005) Parsimony vs. Complexity? –Best solution can cover the most cases consistently – 7 out of 8 cases explained, and the unexplained French case is much weaker
17 Conclusions Back to the cases! The results confirmed importance of numerous factors from variable-based studies, but rejects their sweeping conclusions Underlines strong potential of fuzzy sets
18 Application 2: Determinants of Ownership Structure Corporate Ownership Blockholders (families, state, inter-corporate, banks) vs. dispersed investors (individuals, institutional investors, speculators) Comparative Explanations Legal origins (LLSV) Political systems (Roe) Employee power (Roe) Role of banks vs. markets as alternative source of finance Salience of M&A (Mayer, Franks and Rossi)
19 Model: DISPERSEDGJ = EMPBOARD + CG-MARKET + CG-LEGAL + UNION + COORDINATION + CENTRE-LEFT + CONSENSUS + COMMONLAW + PRIVPENSION + GINI Cases Read: 22 *** NECESSARY CAUSE ANALYSIS *** Number of Cases Tested (Outcome > 0): 17 ( 77.3% of Total) Method: Probabilistic Test Proportion: 0.70 *p < 0.05 N Cause Observed Binomial Variable >= Outcome Proportion p ========================================== empboard 9 0.53 EMPBOARD 8 0.47 cg-market 9 0.53 CG-MARKET 10 0.59 cg-legal 9 0.53 CG-LEGAL 11 0.65 union 9 0.53 UNION 8 0.47 coordination 11 0.65 COORDINATION 9 0.53 centre-left 6 0.35 CENTRE-LEFT 10 0.59 consensus 11 0.65 CONSENSUS 10 0.59 commonlaw 12 0.71 0.597 COMMONLAW 5 0.29 privpension 12 0.71 0.597 PRIVPENSION 5 0.29 gini 12 0.71 0.597 GINI 6 0.35 =============================================== No conditions are necessary for dispersion. In particular, neither common law nor investor protection are necessary.
20 Pathways to dispersion? ********************** *TRUTH TABLE ANALYSIS* ********************** Model: DISPERSEDGJ = EMPBOARD + CG-LEGAL + COORDINATION + CONSENSUS + GINI ** TRUTH TABLE SOLUTION ** raw unique coverage coverage consistency ---------- ---------- ----------- CG-LEGAL*coordination*consensus*gini+ 0.329787 0.085106 0.815790 empboard*CG-LEGAL*coordination*GINI+ 0.457447 0.212766 0.826923 EMPBOARD*CG-LEGAL*COORDINATION*CONSENSUS*gini 0.276596 0.191489 0.722222 solution coverage: 0.744681 solution consistency: 0.777778 Legal protection plus weak labour and high inequality (Anglo-Saxon case) Legal protection plus strong labour, low inequality, and consensus politics (Nordic case).
21 Table 3 Fuzzy Membership Scores: 22 OECD Countries country empboar d cg- marketcg-legalconcentrationbankunioncoordination centre- left consensu s commonla w Australia00.9 0.3 0.70.30.570.321 Austria10.30.1100.910.410.550 Belgium0.300110.90.300.840 Canada00.9 0.30 00.910.321 Denmark10.70.30.90110.910.950 Finland0.90.30.7 10.9110.950 France0.70.30.7 10.10.30.5700 Germany10.30.10.910.30.70.570.840 Greece0.30.1 10.30.100.410.320 Ireland0.3 0.70.300.70.300.841 Italy0.10 100.70.30.570.840 Japan000.3 0.70.3 00.840 Netherlands0.90.3 0.700.30.70.410.950 New Zealand00.30.7 00.3 0.8201 Norway10.9 0.70.910.910.840 Portugal0.1 010.70.3 00.840 South Korea000.3 00 0.700 Spain0.3 110 0.570.550 Sweden10.7 11110.840 Switzerland00.70.3 0.910.950 United Kingdom00.910.10.30.700.410.321 United States00.90.7000.100.910.321