Meta-Econometrics. Tom Stanley, Professor of Economics, Hendrix College USA:

Transcription

1 Meta-Econometrics Tom Stanley, Professor of Economics, Hendrix College USA: Explaining and Communicating Economics Research to Economists

2 The Problem: The house of social science research is sadly dilapidated. It is strewn among the scree of a hundred journals and lies about in the unsightly rubble of a million dissertations Glass, McGaw and Smith (1981:11) The Answer: Meta-Regression Analysis Stanley, T.D., Jarrell, S.B. (1989) Meta-regression analysis: A quantitative method of literature surveys. Journal of Economic Surveys 3:

3 Meta-Analysis of Economics Research Network (MAER-Net) MAER-Net Colloquia Sonderborg, Denmark (07) Nancy, France (08) Corvallis, Oregon 09 Hendrix College, 2010 Cambridge University, UK, 2011 Perth Australia, 2012.

4 Outline of Talk Research Overload Filtering & Selection Priors & Ideology Filtering & Selection Priors & Ideology Bias & Error Meta-Analysis breaks this cycle of Selection Bias by: treating all research equally employing replicable, objective methods to collect and analyze all relevant research.

5 The Tower of Babel Research This is but the start of their undertakings! There will be nothing too hard for them to do. Come, let us go down and confuse their language on the spot so that they can no longer understand one another. Genesis 11: 6-7

6 Meta-analysis refers to the statistical analysis of a large collection of results from individual studies for the purpose of integrating the findings. It connotes a rigorous alternative to the casual, narrative discussions of research studies that typify our attempt to make sense of the rapidly expanding research literature. Glass (1976)

7 Communicating Research Opportunity: accurate and unbiased summaries could do much to advance scientific knowledge Risk: there is a natural tendency for reviewers to select and interpret results from their preferred theory or ideology; thus finding what they expect. When reviewing an empirical literature, many economists seem to become Bayesian, holding strong priors formed on the basis of theoretical considerations and raising high barriers to any contrary empirical evidence Stanley (2001)

8 Believing is seeing Demsetz (1974, p. 164). In Psychology, this is called: confirmation (or experimenter) bias For example, when asked Are you happy with your social life? Y(73%); N(22%); Undec(5%) When asked: Are you unhappy with your social life? Y(65%); N(27%); Undec(8%)

9 Believing is Seeing Publication Selection Bias Reviewers and editors are predisposed to accept papers consistent with the conventional view. Researchers may treat the conventionally expected result as a model selection test. Everyone may have a preference for statistically significant findings. Card and Krueger (1995)

10 Warning! Research findings may appear larger than they actually are. 5-fold exaggeration of the Value of a Statistical life (VSL) fold distortion of the adverse employment effect of minimum wage Doubling of the effect of Antidepressants Register or Perish! All top Medical Journals now require the prior registration of clinical trials.

11 Is Meta-Regression Analysis the Answer? MRA: Uses replicable methods to identify and code all relevant research. Employs rigorous statistical techniques to summarize and explain all research results. Accommodates and corrects selection and misspecification biases routinely found among reported economic findings.

12 1/Se Picturing Publication Selection: Funnel Graphs A funnel graph is a scatter diagram of precision (1/Se) vs. estimated effect. In the absence of publication selection, estimates will vary randomly and symmetrically around true effect (b). The expected inverted funnel 100 shape is dictated by 90 predictable heteroskedasticity. 80 Picture this: A simple graph that reveals much ado about research. (Stanley and Doucouliagos, 2010, Journal of Economic Surveys) r

13 Figure 1: Funnel Plot of Union-Productivity Partial Correlations Funnel Graph is a scatter diagram of all estimates and their precisions (precision= 1/SE) Data: 73 partial correlation coeff s of union membership and worker productivity source: Doucouliagos, C. and Laroche, P. (2003)

14 Figure 2: Value of Statistical Life (in millions of 2000 US $s) Data: 39 estimates of VSL from hedonic wage equations in millions of US $s, 2000 base year Source: Bellavance, Dionne, and Lebeau (2009)

15 Precision Figure 3: Antidepressant Trials Published FDA Data: 74 effect sizes, g, from the US-FDA registry, the gold standard. 50 were published g Source: Turner et al. (2008).

16 1/SE 500 Figure 4: Funnel Graph: Minimum Wage Employment Elasticities {trimmed by Elasticity <1} Data: 1,474 US minimum-wage elasticities of employment elasticity Source: Doucouliagos and Stanley (2009) Brit. J. of Ind. Relations

17 Testing & Correcting Publication Selection FAT-PET-MRA effect j = b + b 0 Se j + e j (j=1, 2, L) t j = b 0 + b (1/Se j ) + u j (1) {OLS} (2) {WLS} {Or use 1/Se 2 in (2) as the weight in a WLS stat package} FAT:t-test of b 0 is a test for publication bias (Egger et al., 1997) PET:t-test of b is a test of authentic effect, beyond publication bias (Stanley, 2005&2008)

18 FAT-PET-MRA-WLS Eq. (2) Table 1: MRA Tests of PB & Genuine Effect Variable Union- Productivity VSL Min Wage Anti- Depressants Se ( b 0 ).65 (1.72) * 3.20(6.67) * -1.6(-4.49) * 1.84 (5.47) * Intercept b (-1.06) * 0.81(3.56) * -.01(-1.09) *.13 (2.50) * Ave. Effect.03 $9.5 mil Top $2.00 mil PEESE $1.67mil n , *Robust t-values in parentheses

19 Putting it all together with Multiple MRA effect i =(b +Sa k Z ik )+(b 0 Se i +Sg j K ij Se i )+e i (3) {Heterogeneity} {Publication Bias}

20 Table 2: Minimum-Wage Moderator Variables K & Z Variable Definition Mean (standard deviation) t-statistic the dependent variable in the FAT-PET regressions (2.83) 1/Se is the elasticity s precision; it is used to test for a genuine effect, PET (28.32) Panel =1, if estimate relates to panel data with time-series as the base 0.45 (0.50) Cross =1, if estimate relates to cross-sectional data with time-series as the base 0.13 (0.34) Adults =1, if estimate relates to young adults (20-24) rather than teenagers (16-19) 0.14 (0.35) 22 Z-var s + 22 K-var s = 44 Var s Male =1, if estimate relates to male employees 0.07 (0.26) Non-white =1, if estimate relates to non-white employees 0.05 (0.22) Region =1, if estimate relates to region specific data 0.10 (0.30) Lag =1, if estimate relates to a lagged minimum wage effect 0.13 (0.34) Hours =1, if the dependent variable is hours worked 0.07 (0.25) Double =1, if estimate comes from a double log specification 0.42 (0.49) AveYear is the average year of the data used, with 2000 as the base year (11.90) Agriculture =1, if estimates are for the agriculture industry 0.01 (0.11) Retail =1, if estimates are for the retail industry 0.08 (0.27) Food =1, if estimates are for the food industry 0.13 (0.34) Time =1, if time trend is included 0.37 (0.48) Yeareffect =1, if year specific fixed effects are used 0.30 (0.46) Regioneffect =1, if region/state fixed effects are used 0.34 (0.47) Un =1, if a model includes unemployment 0.56 (0.50) School =1, if model includes a schooling variable 0.15 (0.35) Kaitz =1, if the Kaitz measure of the minimum wage is used 0.40 (0.49) Dummy =1, if a dummy variable measure of the minimum wage is used 0.17 (0.38) Published =1, if the estimate comes from a published study 0.85 (0.35)

21 Table 3: Generalto-Specific MRA Minimum-Wage Elasticity Variables: Cluster-Robust FE-Panel Genuine empirical effects (Z-variables) Intercept ( b ) (4.39) (6.04) Panel (-4.72) (-10.5) Double (3.20) (5.42) Region (0.92) (6.34) Heterogeneity Adult (2.68) (3.72) Lag (1.60) (1.59) AveYear (4.34) (6.38) Un (-3.04) (-5.79) Doucouliagos and Stanley (2009), British Journal of Industrial Relations. Publication Selection Kaitz (3.06) (3.88) Yeareffect (1.98) (7.44) Published (-2.69) (-4.89) Time (-2.08) (-2.46) Publication bias (K-variables) Se ( b 0 ) (-0.11) (-5.94) Double x Se (-3.23) (-3.90) Un x Se (-1.87) (3.08)

22 Implications? Once publication selection is accommodated, no evidence of an adverse employment effect remains. Substituting any defensible notion of Best Practice Research into these estimated MRA coefficients finds no support for a practically significant adverse employment effect!

23 Minimum Wage Literature Theory and textbooks unanimously agree that the minimum wage reduces employment. It s the law of demand, and market interference leads to unintended negative consequences. Early evidence supported the conventional view. Card and Krueger s (1995) Myth and Measurement created a schism in economics by reporting experimental evidence that MW had no employment effect. There has been much controversy, since.

24 Meta-Regression Analysis: Identifying and Correcting Misspecification and Selection Biases in Economics In our view, the central task of meta-regression analysis is to filter out systematic biases, largely due to misspecification and selection, already contained in economics research. Stanley & Doucouliagos (2012, p.16, Meta- Regression Analysis in Economics and Business.

25 References Bellavance, F. Dionne, G. Lebeau, M The value of a statistical life: A meta-analysis with a mixed effects regression model. Journal of Health Economics 28: Card, D. and Krueger, A.B Time-series minimum-wage studies: A meta-analysis, American Economic Review 85: Doucouliagos, C. (H) and Laroche, P What do unions do to productivity: A meta-analysis, Industrial Relations, 4: Doucouliagos, C.(H) and Stanley, T.D Publication selection bias in minimum-wage research? A meta-regression analysis, British Journal of Industrial Relations 47: Doucouliagos, C. and Stanley, T.D Theory competition and selectivity: Are all economic facts greatly exaggerated? Journal of Economic Surveys, forthcoming. Egger, M., Smith, G.D., Scheider, M., and Minder, C Bias in meta-analysis detected by a simple, graphical test, British Medical Journal 316: Glass, G.V Primary, secondary, and meta-analysis of research. Educational Researcher 5: 3-8. Glass, G.V., McGaw, B. and Smith, M.L Meta-Analysis in Social Research. Beverly Hills: Sage. Heckman, J. J Micro data, heterogeneity, and the evaluation of public policy: Nobel lecture. Journal of Political Economy 109: Stanley, T.D Wheat from chaff: Meta-analysis as quantitative literature review. Journal of Economic Perspectives 15: Stanley, T.D., Beyond publication bias, Journal of Economic Surveys 19: Stanley, T.D Meta-regression methods for detecting and estimating empirical effect in the presence of publication bias. Oxford Bulletin of Economics and Statistics 70: Stanley, T.D., Jarrell, S.B., Meta-regression analysis: A quantitative method of literature surveys. Journal of Economic Surveys 3: Stanley, T.D., and Doucouliagos, H(C) Picture this: A simple graph that reveals much ado about research. Journal of Economic Surveys 24: Stanley, T.D., Jarrell, S. B. and Doucouliagos, H(C) Could it be better to discard 90% of the data? A statistical paradox, American Statistician 64:70-77.