Impact Assessments in Europe Preliminary report Ioannis Lianos & Mihály Fazekas & The Gutenberg project team http://www.ucl.ac.uk/cles/research_initiatives/impact_assessments_in_eur ope Structure of the presentation • Motivations for the research – broader research programme • Objectives of the research project and organization of the team • Theoretical framework and principal hypothesis • Results on the typology of IAs • Data • Regressions • Clustering • Conclusions and further work 2 Motivations for the research Broad research landscape • George (1972) multiple advocacy framework • Impact assessment exercises the function of the “honest broker”…? • The need for dialectical exploration of Facts and Values • Elaboration of rules and procedures governing the exchange of normative arguments • Towards uniform rules of evidence? 3 Objectives of the project and organization of the team Intermediary objectives • develop and explain a robust typology of IAs – Identify characteristic types of IA across 19 jurisdictions – Explore the determinants of IA quality • examine the impact of IAs in the decision-making process and the relation between various actors in the IA process • the impact of IAs on the quality of normative production • explore the emergence of policy areas rationalities in IA and how these could fit into the general framework of RIA Ultimate objectives • understand how the wider environment of (scientific) evidence appraisal in policy or judicial decision making (the “knowledge ecosystem”) affects the development of the tool of IA. • understand how the introduction of the tool of IA transforms the role set of bureaucrats and more generally “governmentality” 4 The Gutenberg project team 5 Theoretical framework The evidence/science politics nexus Four models of interaction between science/expertise and politics/law (Stehr, 2012) •Expertise drives politics •Politics drive expertise •Symbiosis of politics and expertise •Politics and expertise as separate domains (indifference) • Challenging the distinction between policy-makers and knowledge producers • “boundary work” (Halffman, 2003; Hoppe, 2009) • The “argumentative policy analysis” or postempiricist alternative (Fischer & Forester, 1993; Hoppe, 1999) • “participatory policy analysis” – measurement becomes a “social process” (Fischer, 2003) • Different models of boundary arrangements between science and politics 6 The politics – science/expertise nexus Hoppe, 2005 + the importance of “participatory policy analysis” in Europe 7 The six hypothetical models of IAs • The “ticking box” or rudimentary model of IA • The sophisticated rational-instrumental model of IA – CBA style • The shallow rational-instrumental model of IA – CBA style • The rational-instrumental model of IA –cost effectiveness style • The participatory model of IA • The symbiotic model of IA 8 Hypothesized IA clusters and their key characteristics 9 Schematic logic model of impact assessment quality and its effects on the quality of regulatory environment 10 Distribution of IA types per country, 2005-2012 (weighted, up-scaled) 11 Descriptive statistics of country IA production, 2005-2012 (non-weighted) 12 The data • Time period: 2005Q1 – 2012Q2 • 18 countries + EU • Data collection about IAs as well as countries’ institutional frameworks • Data collection templates can be found in the report’s appendices 13 The sample and weighting • Stratified random sampling jurisdiction/year Bulgaria Czech Republic Denmark EU France Germany Greece Hungary Ireland Italy Netherlands Norway Poland Romania Serbia Slovakia Spain Sweden UK Total 2005 0 0 0 0 0 15 0 0 0 0 1 1 2 0 0 12 0 0 0 31 2006 0 0 10 76 0 16 0 0 5 13 1 11 18 0 0 8 0 10 17 185 2007 0 3 12 88 0 15 0 0 4 44 2 9 20 22 6 13 0 10 41 289 2008 0 22 24 114 0 18 0 0 7 16 5 12 20 0 12 11 14 11 17 303 2009 14 18 15 77 3 15 0 0 5 23 1 7 20 0 11 13 12 10 20 264 2010 0 23 22 51 14 16 10 2 9 20 2 10 24 0 9 11 10 9 20 262 2011 0 17 20 119 9 15 12 32 7 9 5 10 21 0 10 13 10 10 11 330 2012* 1 20 21 32 2 18 12 68 12 1 3 10 26 0 11 2 10 10 27 286 Total 15 103 124 557 28 128 34 102 49 126 20 70 151 22 59 83 56 70 153 1950 14 • Weighting to reflect production per year, country Lead indicators: overview • High degree of complexity: 125 elementary indicators of IAs • Need for parsimony • Development of 5 lead indicators – – – – – Scope Sophistication Pseudo-quality Consultation accountability 15 Lead indicators: scope • Scope score = ( budget + business + SMEs + consumers + innovation + employment + health + competition + environment + society + geographic equity + EU institutions + single market + acquis communitaire + corruption + administrative burdens + subsidiarity + proportionality + implementation + transition) / 20. 16 Lead indicators: scope bulgaria • Marked cross-country differences 0.48 eu 0.44 germany 0.37 uk 0.36 denmark 0.36 greece 0.29 netherlands 0.24 spain 0.23 0.22 sweden france 0.20 Mean scope poland 0.18 serbia 0.18 ireland 0.17 slovakia 0.17 czechr 0.16 0.13 hungary norway 0.11 italy 0.10 romania 0.06 0.00 0.10 0.20 0.30 0.40 0.50 0.60 0.70 17 Lead indicators: sophistication • Sophistication score = ( cost/benefit categories explicitly stated + total costs/benefits calculated + scoring of options + use of net benefit calculations + use of cost effectiveness calculations + at least one alternative policy option + do nothing option included + comparison of alternatives + taking account of uncertainty + doing robustness checks + use of simple/complex calculations + comparison with other countries + explicit reference to non-monetisable factors/methodological limitations + explicitly stating a discount rate + using consistent figures)/18 18 Lead indicators: sophistication • Large cross-sectoral differences 0.4 0.34 0.4 0.3 0.31 0.27 0.26 0.25 0.3 0.2 0.2 0.23 0.21 0.21 0.21 0.19 0.18 0.18 Mean sophistication 0.1 0.1 0.0 19 Lead indicators: pseudo-quality • Pseudo quality = scope * sophistication * log length 20 Lead indicators: consultation • Extensiveness of consultation = ( any mention of consultation + evidence submitted through consultation + opinions submitted through consultation + whether consultation took place after the IA completion or not + comprehensiveness of consultation + openness of consultation + consultation shorter than 60 days + number of public parties consulted + number of societal parties consulted + number of external experts consulted)/13 2. czechr 3. denmark 4. eu 5. france 6. germany 7. greece 8. hungary 9. ireland 10. italy 11. netherlands 12. norway 13. poland 14. romania 15. serbia 16. slovakia 17. spain 18. sweden 19. uk Total 20 40 60 80 0 20 40 60 80 0 20 40 60 80 0 Percent 0 20 40 60 80 1. bulgaria 0 .5 1 0 .5 1 0 .5 1 0 .5 1 0 .5 1 comprehensiveness of consultation Graphs by cname 21 Lead indicators: Sophistication versus consultation 22 Lead indicators: accountability • Degree of accountability = ( overarching goals + operational objectives + implementation plan + monitoring progress + ex-post evaluation) / 7. 23 Typology of IAs across Europe: our approach • Theory driven, IA focused (no countries, no years) sophisticscope ation consultation alt. cost benefit account policy figures figures options impact on state/ budget rudimentary IA low low low low no some some no rat.-instr. IA: cost effectiveness style low high low low no many no yes rat.-instr. IA: CBA style (shallow) high low low low yes many many yes rat.-instr. IA: CBA style (sophistic.) high high low low yes many many yes participatory model of IA high low high high yes/no some some no symbiotic model high high high high yes many many yes24 Typology of IAs across Europe: results • Statistically robust findings: 6 major clusters+other • Partial confirmation of hypotheses alt. bene impact polic cost fit on y figur figur state/bud optio es es get ns 0.64 0.23 0.36 0.39 sco pe sophi stic consulta tion acco unt other IAs 0.25 0.18 0.32 0.38 rudimentary IA rat.-instr. IA: cost effectiveness style rat.-instr. IA:CBA style (shallow) participation/acco untability oriented IA (no alt. p. o.) participation/acco untability oriented IA (with alt. p. o.) 0.23 0.07 0.19 0.14 0 0 0 0.37 0.12 0.11 0.18 0 1 0.33 0.18 0.41 0.23 0 0.19 0.10 0.73 0.68 0.18 0.22 0.78 0.67 symbiotic model 0.43 0.47 0.50 0.45 weighted Freq. % 1845 12.15 0.35 1304 8.59 0 0.89 1627 10.71 1 1 0.79 3551 23.38 0 1 0 0.32 1944 12.80 1 1 0 0.38 1322 8.71 3594 25 23.67 1 1 1 0.71 Regressions: theory • Logic map of causal links Institutional environment Macro-level Micro-level Impact assessment characteristics Impact assessment quality 26 Regressions: variables • Dependent variable: pseudo-quality • IA characteristics: 1. 2. 3. 4. 5. area of regulation, type of legal text, preparing institution, consultation score, accountability score. • Country-level institutional environment characteristics-IA-specific institutions: 1. 2. 3. 4. existence of a central IA body checking IA quality, existence of a national IA research institute, legal requirements to include certain impacts such as competition, environment, legal requirements to include at least one alternative option. • The country-level institutional environment characteristics-general quality of government variables: 1. World Economic Forum wastefulness of government spending (0-very wasteful, 6not wasteful at all), 2. length of time government is in power. 27 Regressions: Main effects Level-1: IA Level-2: yearcountry Effect heterogeneity Main effect consultation index x ++ consultation index (squared) x (--) accountability index x +++ central IA bod (=1) x national IA research inst. (=1) x required scope index x mandatory alternative opt (=1) x ++ govt. years in power x + WEF: wasteful gov. spending x - WEF: wasteful gov. spending (sq.) x (+) Controls* x x ++ ++ x - +/28 Regressions-selected predictors: fixed effects: WEF wastefulness of gov spending 29 Regressions-selected predictors: random effects year-country predictions EU BG UK FR NL CZ 30 Regressions-selected predictors: random effects Heterogeneous effect of IA Board-type institutions 31 Detailed results optimization method: maximum likelihood, constrained covariance structure dep var: pseudoq model nr.: Regression coefficeints (fixed effect) Constant Level-1 factor: individual IA controls consultation index consultation index (squared) accountability index IA IA IA characteristics characteristics empty' characteristics +IA institutions +IA institutions model +IA institutions (random (random (random slope) intercept) slope)+pol inst 1 0.0346 *** -3.95 2 3 4 0.0304* 0.0329** 0.0364*** -2.29 -2.84 -3.55 0.0320*** -4.07 0.0083 -0.28 0.0754*** -9.96 0.0321*** -4.15 0.00587 -0.2 0.0747*** -9.84 0.0223** -3.12 -0.0232 (-0.83) 0.0821*** -8.71 32 Detailed results Level-2 factor: year-country cells central IA bod (=1) national IA research inst. (=1) required scope index mandatory alternative opt (=1) government cycles first 4 years second 4 years third 4 years WEF: wasteful gov spending 0.0139 -1.85 0.0326* -2.21 -0.00506 (-1.69) 0.0245* 0.00899 -1.04 0.0288* -2.24 -0.00467 (-1.69) 0.0222* 0.00224 -0.3 0.0261* -2.57 -0.00594** (-2.64) 0.0186* -2.03 -1.98 -2.11 ref.cat 0.00830* -2.28 -0.00927 (-0.95) -0.00999* (-2.03) 33 Detailed results Variance components (random effects) Level-3: country var(required scope index) 0.00001 0.00000 var(centraly IA bod (=1)) 0.00024 0.00014 0.00047 0.00045 0.00001 0.00000 var(_cons) 0.00132 0.00087 Level-2: year-country cells var(required scope index) var(_cons) 0.00039 0.00023 0.00000 0.00000 var(Residual) 0.00244 0.00225 0.00224 0.00109 N 1947 1864 1864 980 R2 n.a. 23.4% 23.5% 23.9% -6012.9 -5937.2 -5966.4 -3829.8 Model summary Deviance t statistics in parentheses * p<0.05, ** p<0.01, *** p<0.001 34