Appendix B: Methodological Report Contents Introduction .............................................................................................................................................. 1 Model Specifications and Model Results ................................................................................................. 1 Model 1: Intercept-Only ...................................................................................................................... 1 Model 2: Population Model & Survey Weights ................................................................................... 4 Comparison of Survey-Specific and Cross-survey Estimates ............................................................ 11 Model 3: Inclusion of Year of Survey ............................................................................................... 15 Model 4: Based to Single year: 2004 ................................................................................................ 17 Model 5: Based to subset of years: 2004-2008 ................................................................................ 19 Population Counts and Proportions by State: 2010 ........................................................................... 21 Weighting: Survey of American Jews 2010 ........................................................................................... 23 Sample Composition .......................................................................................................................... 23 State-Metropolitan Clusters................................................................................................................ 23 Knowledge Network Sampling Weight ............................................................................................. 26 Adjustments for non-response ............................................................................................................ 28 Adjustments for Representativeness of the Sample ........................................................................... 30 Relation of Post-stratification Variables to Outcomes ....................................................................... 30 Post-Stratification 1: Raking to Demos & State-Metro Clusters....................................................... 36 Post-Stratification 2: Raking to Demos & Region x Metropolitan Status......................................... 39 Post-Stratification 3: Full Post-Stratification .................................................................................... 40 Post-Stratification 4: Next Steps Model-based Estimation ............................................................... 42 Additional Considerations .................................................................................................................. 43 Handling of JBO Sample ............................................................................................................... 44 Trimming ....................................................................................................................................... 45 Weight summary ................................................................................................................................ 45 Table of Tables Table B1: Model Results, Jewish likelihood as a function of random intercepts for survey. .......................................3 Table B2: Results from Hierarchical Bayes. .................................................................................................................6 Table B3: results of regression Jewish likelihood on survey weights. ..........................................................................8 Table B4: Unweighted, Weighted and Cross-survey estimate for each sample survey, 2000-2008. ......................... 11 Table B5: Results from multilevel analysis with laplace approximation, Jewish likelihood on demographics and year of survey. ............................................................................................................................................................. 15 Table B6: Jewish population estimates using single year (2004) model, post-stratified to 2004 CPS........................ 17 Table B7: Results from Hierarchical Bayes population model for 2004-2008 ........................................................... 19 Table B8: Total Jewish Population by State: 2010. .................................................................................................. 21 Table B9: SAJ 2010 sample size within state-metropolitan clusters. ......................................................................... 23 Table B10: Comparison of SSRI estimates with Knowledge Network’s Jewish Panel and the SAJ 2010 sample. ... 24 Table 11: Demographic characteristics of SAJ 2010 sample weighted by sample weight. ........................................ 27 Table B12: Demographic characteristics of SAJ 2010 sample weighted by nonresponse weight .............................. 29 Table B13: Bivariate relationships of having a bar/bat mitzvah, years of Jewish education and intermarriage with age, education, race, metropolitan status and geographic region. ................................................................................ 32 Table B14: Bivariate relationships of attending Jewish life cycle events with age, education, race, metropolitan status and geographic region. ...................................................................................................................................... 33 Table B15: Bivariate relationships of Reading Hebrew and synagogue membership with age, education, race, metropolitan status and geographic region. ................................................................................................................. 34 Table B16: Bivariate relationships of synagogue honors with age, education, race, metropolitan status and geographic region. ....................................................................................................................................................... 35 Table B17: Demographic characteristics of NJPS 2010 sample with demographic and state groups raking adjustments. ................................................................................................................................................................. 36 Table B18: Comparison of the KN Sampling weight estimates and post-stratification weight estimates. ................. 41 Table B19: Logistic regression JBO vs. JBRI Secular on age, race, education, state and metropolitan area. ........... 44 Table B20: Comparison of estimates of survey outcomes for different weighting scenarios. .................................... 46 Table of Figures Figure B1: Trace plots for the intercept and sigma. ......................................................................................................3 Figure B2: State level estimates of Jewish population based to 2004 single year model. ......................................... 18 Figure B3: Distribution of KN Panel and SAJ 2010 respondents by state groups...................................................... 25 Figure B4: Distribution of sample weights for SAJ 2010........................................................................................... 26 Figure B5: sample weights adjusted for non-response for SAJ 2010. ....................................................................... 28 Figure B6: Comparison of KN sampling weight to raked weight by age group. ........................................................ 38 Figure B7: Comparison of MSE and variance of unweighted estimates across distribution of basic demographics, estimates with KN sampling weight and estimates after raking. ................................................................................. 39 Figure B8: Comparison of MSE and variance of unweighted estimates, KN sampling weight, 8 state cluster raking, and 9 region raking. ..................................................................................................................................................... 40 Figure B9: Comparison of MSE and variance of unweighted estimates, KN sampling estimates, 8 state cluster raking, 9 region raking, and full post-stratification estimates. ..................................................................................... 42 Methodological Report Elizabeth Tighe Raquel Magidin de Kramer Begli Nursahedov Janet Aronson October 2011 Introduction This report provides the methodological and analytic details associated with the crosssurvey analysis of the US Jewish population 2000-2008. This includes specification of the statistical models and model results. Also included are details associated with the use of cross-survey results in the analysis of the Survey of American Jews 2010 (Saxe, 2010). Several alternatives for the use of Jewish population totals in post-stratification adjustments are examined. Model Specifications and Model Results Model 1: Intercept-Only Prior to fitting a post-stratification model, an intercept-only model was fit to examine the degree of variance across surveys in estimates of the Jewish population before taking into account sampling and other variables. Three models were fit. The first was run as a classical multilevel using LMER (Bates, 2011). LMER applies a LaPlace approximation, which performs well under some conditions and not so well under others -- such as ours where there is near zero variance and large discrepancies in cluster sizes. The LMER model with LaPlace approximation was compared to two Hierarchical Bayes models.1 One model used non-informative priors (explain). The other incorporated the LMER estimates as informative priors.2 All analyses were run on Linux/Ubuntu calling JAGS (Just Another Gibbs Sampler) from R. 1 Bayesian models were run using JAGS (Plummer, 2010) with rjags (Plummer, 2011) and R2Jags (Su & Yajima, 2011) on Ubuntu Linux. 2 This was done given the estimate of the Jewish population was expected to be low – between less than 1% to no higher than 6% based on analysis of the individual surveys (See Table B4) and the variance across surveys was also estimated to be low – with estimates varying within this restricted range. The Markov chain Monte Carlo algorithms sample from the entire solution space (0% to 100%) and can be slow to converge if the proposed variance is either too large or too small (see Roberts and Rosenthal, 2001). One approach to speed convergence is to use the variance estimate from a maximum likelihood model, or in this case the LaPlace approximation from the LMER model, scaled by a factor from 2 to 10 (Congdon, 2009). 1 The LMER model fit a basic multilevel model for the dichotomous outcome (Jewish/Not Jewish) with random intercepts for survey: mod1 <- lmer(curreljw ~ (1|survf), family=binomial(link="logit")) The Bayesian model with uninformative priors was specified as: model{ for (i in 1:N){ curreljw[i] ~ dbin(p.bound[i], 1) p.bound[i] <- max(0, min(1,p[i])) logit(p[i]) <- b.cons + b.surv[survs[i]] } b.cons ~ dnorm(0,.01) for (j in 1:n.surv){ b.surv[j] ~ dnorm(0, tau.surv) } tau.surv <- pow(sigma.surv, -2) sigma.surv ~ dunif(0,10) } With proposal distributions based to the LMER estimates, this model was modified as: model{ for (i in 1:Ndm){ curreljw[i] ~ dbin(p.bound[i], 1) p.bound[i] <- max(0, min(1,p[i])) logit(p[i]) <- b.cons + b.surv[survs[i]] } b.cons ~ dnorm(-3.8,155) for (j in 1:n.surv){ b.surv[j] ~ dnorm(0, tau.surv) } tau.surv <- pow(sigma.surv, -2) sigma.surv ~ dunif(0,.7) } Both of these models yielded similar results to the LMER model (see Table B1). Trace plots for the third model, displayed in Figure B1, indicate that the proposal variances 2 perform well. The figure displays plots for the intercept and sigma, which were the slowest parameters to converge. TABLE B1: MODEL RESULTS, JEWISH LIKELIHOOD AS A FUNCTION OF RANDOM INTERCEPTS FOR SURVEY. JAGS w/noninformative priors LMER Intercept JAGS w/ LMER priors Est. SE Est. SE Est. SE -3.84 0.026 -3.84 0.025 -3.84 0.025 Survey Variance Tau 21.696 22.034 Variance 0.046 0.046 0.045 ICC 0.014 0.014 0.014 MOR 1.225 1.226 1.224 Deviance 68977 68867 68867 FIGURE B1: TRACE PLOTS FOR THE INTERCEPT AND SIGMA. 3 Model 2: Population Model & Survey Weights The full post-stratification model specification submitted to JAGS – using results from an initial fit using LMER to specify proposal distributions included random intercepts for survey along with fixed effects for demographic variables and fixed effects for state and metropolitan status3: model{ for (i in 1:Ndm){ curreljw[i] ~ dbin(p.bound[i], 1) p.bound[i] <- max(0, min(1,p[i])) XBSt1[i] <- b.st1*st1[i] + b.st2*st2[i] + b.st3*st3[i] + b.st5*st5[i] + b.st6*st6[i] + b.st7*st7[i] + b.st8*st8[i] + b.st9*st9[i] + b.st10*st10[i] + b.st11*st11[i] + b.st12*st12[i] + b.st13*st13[i] + b.st14*st14[i] + b.st15*st15[i] + b.st16*st16[i] + b.st17*st17[i] + b.st18*st18[i] + b.st19*st19[i] + b.st20*st20[i] + b.st21*st21[i] + b.st22*st22[i] + b.st23*st23[i] + b.st24*st24[i] + b.st25*st25[i] XBSt2[i] <- b.st26*st26[i] + b.st27*st27[i] + b.st28*st28[i] + b.st29*st29[i] + b.st30*st30[i] + b.st31*st31[i] + b.st32*st32[i] + b.st33*st33[i] + b.st34*st34[i] + b.st35*st35[i] + b.st36*st36[i] + b.st37*st37[i] + b.st38*st38[i] + b.st39*st39[i] + b.st40*st40[i] + b.st41*st41[i] + b.st42*st42[i] + b.st43*st43[i] + b.st44*st44[i] + b.st45*st45[i] + b.st46*st46[i] + b.st47*st47[i] + b.st48*st48[i] + b.st49*st49[i] XBDem[i] <- b.cons + b.racb*racb[i] + b.rach*rach[i] + b.raco*raco[i] + b.edu*eduncg[i] + b.age1*age1[i] + b.age2*age2[i] + b.age3*age3[i] + b.age5*age5[i] + b.age6*age6[i] + b.met*metstat[i] + b.exa1*eduncg[i]*age1[i] + b.exa2*eduncg[i]*age2[i] + b.exa3*eduncg[i]*age3[i] + b.exa5*eduncg[i]*age5[i] + b.exa6*eduncg[i]*age6[i] logit(p[i]) <- XBDem[i] + XBSt1[i] + XBSt2[i] + b.surv[survs[i]] } 3 States were included as fixed rather than random effects to reduce the influence of outlier states, particularly states such as Utah or the Dakotas where the estimated size of the population is so small (recommendation from David Rindskopf). 4 b.cons ~ dnorm(-3.74,35) b.racb ~ dnorm(-2.39,7) b.rach ~ dnorm(-1.66,13) b.raco ~ dnorm(-1.19,16) b.age1 ~ dnorm(-0.12,11) b.age2 ~ dnorm(-0.35,37) b.age3 ~ dnorm(-0.27,47) b.age5 ~ dnorm(0.11,50) b.age6 ~ dnorm(0.22,50) b.edu ~ dnorm(-1.49,25) b.met ~ dnorm(1.24,32) b.exa1 ~ dnorm(0.88,7) b.exa2 ~ dnorm(0.24,9) b.exa3 ~ dnorm(0.18,11) b.exa5 ~ dnorm(0.18,13) b.exa6 ~ dnorm(0.41,16) b.st1 ~ dnorm(-1.88,3) b.st2 ~ dnorm(-0.57,11) b.st3 ~ dnorm(-1.91,2) b.st5 ~ dnorm(-0.73,9) b.st6 ~ dnorm(-0.07,11) b.st7 ~ dnorm(-0.82,2) b.st8 ~ dnorm(1.04,5) b.st9 ~ dnorm(0.28,41) b.st10 ~ dnorm(-0.79,11) b.st11 ~ dnorm(-1.72,2) b.st12 ~ dnorm(-0.37,21) b.st13 ~ dnorm(-1.8,4) b.st14 ~ dnorm(-1.95,2) b.st15 ~ dnorm(-1.7,2) b.st16 ~ dnorm(-1.87,2) b.st17 ~ dnorm(-1.86,3) b.st18 ~ dnorm(-1,3) b.st19 ~ dnorm(0.41,21) b.st20 ~ dnorm(0.13,24) b.st21 ~ dnorm(-0.93,12) b.st22 ~ dnorm(-1.39,6) b.st23 ~ dnorm(-2.13,1) b.st24 ~ dnorm(-1.13,7) b.st25 ~ dnorm(-2.14,1) b.st26 ~ dnorm(-1.97,1) b.st27 ~ dnorm(-0.27,5) b.st28 ~ dnorm(-0.86,3) b.st29 ~ dnorm(0.53,32) b.st30 ~ dnorm(-0.7,3) b.st31 ~ dnorm(0.89,55) b.st32 ~ dnorm(-1.2,8) b.st33 ~ dnorm(-3.12,0.3) b.st34 ~ dnorm(-0.91,16) b.st35 ~ dnorm(-1.83,2) b.st36 ~ dnorm(-1.12,5) b.st37 ~ dnorm(-0.35,28) b.st38 ~ dnorm(-0.62,3) b.st39 ~ dnorm(-1.57,3) b.st40 ~ dnorm(-3.28,0.25) b.st41 ~ dnorm(-1.82,5) b.st42 ~ dnorm(-1.32,15) b.st43 ~ dnorm(-2.49,1) b.st44 ~ dnorm(-0.15,3) b.st45 ~ dnorm(-1.02,10) b.st46 ~ dnorm(-1.12,9) b.st47 ~ dnorm(-1.62,2) b.st48 ~ dnorm(-1.65,5) b.st49 ~ dnorm(-1.02,1) for (j in 1:n.surv){ b.surv[j] ~ dnorm(0, tau.surv) } tau.surv <- pow(sigma.surv, -2) sigma.surv ~ dunif(0,.4) 5 Results from this model are displayed in Table B2 (column 1). Column 2 displays results after including the original survey weights as a covariate in the population model. Column 3 displays the same for the subset of surveys where the original survey weight was significantly related to the likelihood of respondents identifying as Jewish. See Table B3 for results of regressions of Jewish likelihood on survey weights. TABLE B2: RESULTS FROM HIERARCHICAL BAYES. Cross-Survey Estimates Est. SD Inclusion of Survey Weight Est. SD Subset Sig. Survey Weight Est. SD Random Intercept for Survey Tau Variance ICC MOR Deviance 89.15 88.42 278.10 206.72 962.45 0.01 0.004 0.001 0.003 0.001 0.0003 1.11 1.06 1.03 58256 33042 19899 2560.39 Fixed Effects Final Weight -- -- <.001 <.001 <.001 <.001 -3.75 0.06 -3.83 0.09 -3.68 0.11 18-24 years -0.12 0.09 -0.10 0.12 -0.15 0.17 25-34 years -0.35 0.05 -0.31 0.07 -0.41 0.09 35-44 years -0.27 0.04 -0.25 0.06 -0.34 0.07 55-64 years 0.11 0.04 0.13 0.05 0.11 0.07 65 years and older 0.22 0.04 0.20 0.05 0.20 0.07 -1.49 0.05 -1.41 0.07 -1.45 0.10 Black non-Hispanic -2.40 0.11 -2.55 0.16 -2.55 0.22 Hispanic -1.66 0.09 -1.67 0.11 -1.75 0.16 Other non-Hispanic -1.19 0.08 -1.25 0.10 -1.21 0.13 18-24 years 0.88 0.11 0.88 0.15 1.02 0.20 25-34 years 0.23 0.10 0.23 0.13 0.15 0.18 35-44 years 0.18 0.08 0.16 0.11 0.16 0.15 55-64 years 0.17 0.08 0.10 0.10 0.13 0.13 65 years and older 0.41 0.07 0.35 0.09 0.29 0.12 1.24 0.05 1.29 0.08 1.23 0.10 Alabama -1.88 0.19 -2.04 0.26 -2.10 0.37 Arizona -0.79 0.09 -0.77 0.12 -0.98 0.17 Arkansas -1.76 0.27 -1.66 0.34 -1.69 0.44 Colorado -0.37 0.06 -0.32 0.08 -0.28 0.10 Connecticut -1.82 0.16 -1.88 0.20 -1.79 0.25 Intercept Age Education Less than College Race/Ethnicity Age x Education Metropolitan Status In Metro State 6 Delaware -1.97 0.23 -2.12 0.35 -2.10 0.45 District of Columbia -1.72 0.21 -1.99 0.31 -2.07 0.42 Florida -1.89 0.20 -1.96 0.29 -2.50 0.48 Georgia -1.87 0.21 -1.65 0.26 -1.64 0.32 Idaho -1.01 0.20 -0.98 0.25 -1.13 0.38 Illinois 0.41 0.06 0.41 0.08 0.45 0.11 Indiana -0.58 0.09 -0.53 0.11 -0.42 0.15 0.13 0.06 0.10 0.07 0.05 0.11 Kansas -0.93 0.08 -0.94 0.11 -0.93 0.15 Kentucky -1.39 0.13 -1.28 0.17 -1.43 0.22 Louisiana -2.17 0.36 -2.46 0.53 -2.53 0.68 Maine -1.13 0.12 -1.12 0.16 -1.14 0.19 Maryland -2.22 0.46 -2.07 0.60 -2.00 0.68 Massachusetts -2.02 0.33 -2.25 0.49 -2.18 0.58 Michigan -0.27 0.14 -0.15 0.17 -0.08 0.21 Minnesota -0.88 0.20 -0.59 0.23 -0.60 0.31 Mississippi 0.53 0.05 0.53 0.07 0.56 0.09 Missouri -1.93 0.28 -1.94 0.36 -2.11 0.49 Montana -0.71 0.18 -0.67 0.23 -0.73 0.31 Nebraska 0.90 0.04 0.85 0.05 0.80 0.06 Nevada -1.20 0.10 -1.41 0.15 -1.44 0.18 New Hampshire -3.42 0.95 -2.94 0.99 -2.36 0.87 New Jersey -0.90 0.07 -0.94 0.10 -0.90 0.12 New Mexico -1.84 0.21 -1.70 0.27 -1.68 0.36 New York -1.11 0.14 -1.25 0.18 -1.33 0.25 North Carolina -0.35 0.06 -0.35 0.07 -0.37 0.09 North Dakota -0.63 0.20 -0.83 0.27 -1.07 0.42 Ohio -1.57 0.17 -1.30 0.20 -1.07 0.23 Oklahoma -3.62 0.94 -3.07 1.01 -2.47 0.90 Oregon -1.83 0.15 -2.16 0.23 -2.49 0.36 Pennsylvania -1.32 0.08 -1.21 0.09 -1.15 0.12 Rhode Island -2.53 0.34 -2.75 0.49 -2.62 0.56 South Carolina -0.16 0.20 -0.39 0.29 -0.08 0.32 South Dakota -1.02 -1.12 -1.67 -1.65 -1.07 -0.73 -0.07 -0.85 1.03 0.28 0.10 0.10 0.28 0.14 0.38 0.10 0.09 0.26 0.15 0.05 -0.97 -1.33 -2.03 -1.74 -1.02 -0.78 -0.07 -0.91 1.08 0.32 0.13 0.14 0.41 0.19 0.48 0.13 0.10 0.35 0.19 0.06 -0.98 -1.52 -2.12 -2.00 -0.49 -0.63 -0.12 -1.46 1.17 0.31 0.16 0.20 0.59 0.27 0.46 0.15 0.15 0.55 0.24 0.07 Iowa Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming Note: The reference group are those age 45-54, white non-Hispanic, college graduates, residing in nonmetropolitan California. 7 TABLE B3: RESULTS OF REGRESSION JEWISH LIKELIHOOD ON SURVEY WEIGHTS. Survey ID S113 S40 S103 S112 S116 S10 S106 S130 S124 S94 S58 S90 S12 S18 S134 S97 S6 S65 S107 S37 S41 S84 S129 S2 S11 S31 S16 S75 S81 S125 S117 S137 S4 S14 S82 S32 S9 S42 S96 S62 S83 S71 S63 S72 S132 S59 B -1.328 -0.205 -1.717 -1.290 -1.653 -1.138 -2.370 -1.184 -1.112 -3.382 -4.311 -2.677 -0.751 -1.093 -2.261 -4.748 -1.666 -1.556 -1.773 -1.620 -0.462 -1.486 -1.184 -1.019 -.502 -.519 -.799 -2.598 -1.376 2.305 -.995 -1.369 .324 -.181 -1.317 -.739 .508 -.503 -1.378 -1.913 -1.557 -1.873 -1.018 -1.610 -.001 -2.283 8 S.E. .212 .036 .358 .328 .443 .311 .721 .374 .366 1.146 1.502 .962 .279 .407 .843 1.772 .622 .596 .686 .631 .180 .586 .474 .415 .205 .214 .335 1.102 .587 .984 .430 .592 .141 .079 .583 .328 .228 .228 .629 .885 .730 .882 .486 .772 .001 1.171 Sig. .000 .000 .000 .000 .000 .000 .001 .002 .002 .003 .004 .005 .007 .007 .007 .007 .007 .009 .010 .010 .010 .011 .012 .014 .014 .015 .017 .018 .019 .019 .021 .021 .022 .022 .024 .024 .026 .027 .028 .031 .033 .034 .036 .037 .047 .051 S33 S92 S85 S24 S80 S105 S93 S51 S43 S69 S39 S86 S88 S127 S34 S22 S60 S64 S115 S89 S118 S47 S76 S74 S66 S73 S15 S61 S29 S70 S126 S79 S101 S100 S36 S28 S52 S57 S111 S67 S30 S5 S109 S131 S19 S114 S13 S8 S44 S139 -.235 -1.292 -.762 -.443 -1.047 -1.277 -1.447 -.218 -.725 -1.181 -.428 -.896 -1.016 -1.159 -.163 .528 -1.229 -1.274 -.517 -.793 -.701 -.122 -1.652 -.975 -1.052 -1.256 -.004 -.966 -.187 -1.005 .256 -1.593 1.206 -.340 -.186 .175 -.356 -.475 .336 -.647 -.306 -.366 -.497 .000 -.218 -.268 .136 -.455 .156 -.266 9 .121 .673 .403 .236 .563 .692 .789 .119 .396 .647 .235 .494 .583 .667 .094 .309 .735 .772 .315 .489 .433 .075 1.030 .610 .658 .806 .002 .630 .122 .690 .181 1.127 .881 .255 .142 .137 .281 .377 .283 .561 .266 .319 .435 .000 .200 .246 .125 .424 .147 .263 .052 .055 .059 .060 .063 .065 .067 .067 .067 .068 .069 .070 .081 .082 .084 .087 .094 .099 .100 .105 .105 .107 .109 .110 .110 .119 .123 .125 .126 .145 .157 .158 .171 .183 .190 .201 .206 .208 .236 .248 .249 .251 .254 .262 .275 .277 .277 .283 .286 .311 S87 S20 S128 S98 S91 S53 S38 S7 S121 S17 S3 S77 S45 S49 S136 S138 S120 S55 S140 S35 S99 S95 S50 S78 S102 S104 S133 S25 S21 S1 S54 S119 S135 S27 -.372 .288 -.546 -.195 -.420 -.106 -.057 -.460 -.520 -.183 .166 -.369 -.079 -.076 -.366 -.102 -.503 -.036 -.180 .038 -.235 -.187 -.060 -.181 -.079 .063 -.072 -.023 .042 -.039 -.004 .005 .006 -.001 10 .376 .291 .578 .210 .460 .120 .072 .578 .675 .251 .229 .521 .118 .121 .590 .182 .912 .066 .349 .076 .488 .451 .148 .486 .259 .208 .261 .097 .261 .312 .091 .150 .185 .163 .322 .322 .345 .352 .361 .376 .424 .426 .442 .466 .469 .479 .502 .529 .535 .575 .581 .588 .606 .617 .630 .678 .687 .709 .760 .762 .782 .813 .873 .901 .969 .973 .974 .997 Comparison of Survey-Specific and Cross-survey Estimates TABLE B4: UNWEIGHTED, WEIGHTED AND CROSS-SURVEY ESTIMATE FOR EACH SAMPLE SURVEY, 2000-2008. Survey ID 1 2 3 4 7 9 Survey Shop Sample Size ISR/NORC 1006 ISR/NORC 801 ISR/NORC 1212 ISR/NORC 2320 PSRAI PSRAI 10 PSRAI 11 PSRAI 12 14 PSRAI SRBI 15 U-Indiana SurvRes 17 ISR/NORC 18 19 ISR/NORC ISR/NORC 20 ISR/NORC 21 ISR/NORC 23 26 PSRAI PSRAI 27 PSRAI 28 PSRAI 29 30 PSRAI TNS 32 UConn SurvRes 33 UConn SurvRes 34 36 UConn SurvRes NewEng SurvRes 42 SRBI 44 Gallup 45 46 Gallup Gallup 47 PSRAI 48 OpResearch 49 50 ICR PSRAI 51 CBS/NYTimes 52 TNS Unweighted Prop. Survey Specific Lower Upper Prop. 95% CI 95% CI Cross-Survey Model-Base Lower Upper Prop. 95% CI 95% CI 0.027 0.020 0.013 0.031 0.020 0.016 0.024 0.020 0.018 0.011 0.031 0.018 0.015 0.022 0.029 0.021 0.014 0.030 0.020 0.017 0.025 0.017 0.012 0.011 0.006 0.027 0.022 0.017 0.014 0.021 1000 0.012 0.013 0.017 0.013 0.020 3142 0.022 0.019 0.015 0.024 0.019 0.015 0.022 3002 0.020 0.018 0.014 0.023 0.018 0.015 0.021 3000 0.018 0.016 0.012 0.021 0.018 0.015 0.021 3204 0.021 0.022 0.017 0.029 0.018 0.015 0.021 2988 0.017 0.015 0.010 0.020 0.017 0.014 0.020 1018 0.029 0.031 0.022 0.044 0.019 0.016 0.023 0.022 0.022 0.014 0.036 0.020 0.017 0.024 0.018 0.015 0.007 0.030 0.018 0.015 0.021 0.020 0.021 0.012 0.035 0.019 0.016 0.022 0.017 0.019 0.011 0.032 0.019 0.016 0.022 0.017 0.022 0.010 0.016 0.029 0.030 0.019 0.015 0.022 1500 0.019 0.025 0.020 0.016 0.024 2041 0.013 0.012 0.008 0.018 0.017 0.013 0.020 2002 0.016 0.013 0.009 0.019 0.017 0.014 0.020 2002 0.020 0.019 0.013 0.026 0.018 0.015 0.021 3002 0.022 0.018 0.014 0.024 0.018 0.015 0.021 3003 0.013 0.014 0.010 0.019 0.017 0.014 0.020 1012 0.030 0.019 0.013 0.027 0.019 0.016 0.023 1000 0.024 0.016 0.010 0.026 0.019 0.015 0.023 1000 0.030 0.030 0.021 0.043 0.019 0.015 0.023 1003 0.056 0.025 0.020 0.033 2000 0.016 0.017 0.011 0.025 0.017 0.014 0.020 1642 0.024 0.020 0.014 0.029 0.019 0.015 0.022 882 0.032 0.025 0.016 0.037 0.019 0.016 0.024 876 0.041 0.031 0.022 0.046 0.020 0.017 0.025 1501 0.013 0.015 0.009 0.025 0.016 0.013 0.019 1029 0.030 0.020 0.013 0.030 0.020 0.015 0.025 1617 0.025 0.014 0.010 0.020 0.019 0.016 0.023 1503 0.021 0.018 0.012 0.026 0.018 0.015 0.022 1134 0.021 0.017 0.010 0.030 0.018 0.014 0.021 1204 0.028 0.023 0.016 0.034 0.020 0.016 0.024 2813 2747 2791 4475 2023 11 53 54 55 56 SRBI Gallup ICR SRBI 57 Gallup 58 Gallup 59 60 PSRAI Gallup 61 Gallup 62 CBS/NYTimes 63 64 Gallup ICR 65 PSRAI 66 Gallup 67 68 TNS PSRAI 69 Gallup 70 TNS 72 73 PSRAI Gallup 74 Gallup 75 TNS 76 77 TNS TNS 78 TNS 79 TNS 80 81 TNS TNS 82 TNS 83 TNS 84 85 TNS TNS 86 TNS 87 TNS 88 91 TNS PSRAI 92 PSRAI 93 PSRAI 94 95 SRBI CBS/NYTimes 96 CBS/NYTimes 97 CBS/NYTimes 98 CBS/NYTimes 1001 0.028 0.025 0.017 0.036 1003 0.026 0.016 0.011 0.024 1010 0.029 0.018 0.012 0.028 990 0.024 0.023 0.015 0.034 1002 0.015 0.014 0.008 0.025 1002 0.030 0.021 0.014 0.032 1405 0.018 0.016 0.011 0.025 1006 0.027 0.022 0.014 0.035 1003 0.023 0.016 0.010 0.026 1145 0.019 0.013 0.008 0.021 1040 0.029 0.018 0.013 0.027 1026 0.023 0.016 0.010 0.026 2558 0.011 0.011 0.006 0.019 1015 0.020 0.018 0.011 0.029 969 0.034 0.025 0.017 0.035 2009 0.023 0.019 0.014 0.026 1000 0.034 0.026 0.018 0.037 1005 0.018 0.014 0.009 0.023 1703 0.017 0.015 0.010 0.021 1005 0.021 0.016 0.010 0.027 1029 0.017 0.013 0.007 0.021 1007 0.018 0.014 0.009 0.024 1202 0.024 0.020 0.013 0.029 1201 0.018 0.014 0.009 0.022 1202 0.027 0.022 0.015 0.032 1004 0.016 0.013 0.008 0.022 1004 0.035 0.026 0.018 0.037 1007 0.020 0.016 0.010 0.026 1036 0.023 0.020 0.013 0.032 1229 0.029 1003 0.020 0.015 0.009 0.025 1002 0.021 0.018 0.011 0.028 1001 0.024 0.018 0.012 0.028 1001 0.024 0.017 0.011 0.026 1006 0.020 0.016 0.010 0.026 1502 0.022 0.018 0.012 0.026 1110 0.027 0.021 0.014 0.032 1130 0.019 0.021 0.012 0.036 1251 0.024 0.019 0.012 0.030 1104 0.020 0.013 0.008 0.022 1126 0.028 0.024 0.014 0.041 989 0.025 0.024 0.013 0.042 829 0.023 0.012 0.007 0.021 12 0.020 0.016 0.024 0.019 0.016 0.022 0.020 0.016 0.024 0.019 0.015 0.023 0.017 0.013 0.020 0.019 0.016 0.024 0.017 0.014 0.021 0.019 0.016 0.023 0.018 0.014 0.022 0.018 0.015 0.022 0.019 0.016 0.023 0.018 0.015 0.022 0.017 0.014 0.020 0.018 0.014 0.021 0.020 0.016 0.025 0.019 0.016 0.023 0.021 0.017 0.025 0.018 0.014 0.022 0.017 0.014 0.020 0.018 0.015 0.022 0.017 0.013 0.021 0.018 0.015 0.022 0.019 0.015 0.023 0.018 0.014 0.022 0.020 0.016 0.025 0.018 0.014 0.021 0.022 0.018 0.027 0.018 0.014 0.022 0.019 0.015 0.023 0.020 0.016 0.024 0.018 0.014 0.022 0.019 0.015 0.023 0.019 0.015 0.023 0.019 0.016 0.023 0.018 0.015 0.022 0.019 0.016 0.023 0.018 0.015 0.022 0.019 0.016 0.023 0.018 0.014 0.021 0.020 0.016 0.024 0.018 0.014 0.021 0.019 0.016 0.023 0.019 0.015 0.023 99 100 101 102 CBS/NYTimes CBS/NYTimes CBS/NYTimes CBS/NYTimes 103 Gallup 104 Gallup 105 106 Gallup Gallup 107 Gallup 108 Gallup 110 111 GreenbergRosner PSRAI 112 TNS 114 TNS 115 116 TNS TNS 117 TNS 118 SRBI 119 120 SRBI TNS 121 PSRAI 122 PSRAI 123 124 SRBI RTI 125 PSRAI 126 PSRAI 127 128 PSRAI PSRAI 129 SRBI 130 SRBI 131 132 SRBI SRBI 133 PSRAI 134 PSRAI 135 136 PSRAI PSRAI 137 PSRAI 138 PSRAI 139 140 PSRAI PSRAI 141 PSRAI 142 PSRAI 143 PSRAI 894 0.025 0.020 0.012 0.033 1125 0.015 0.014 0.008 0.025 1276 0.024 0.025 0.013 0.044 992 0.009 0.008 0.004 0.016 1007 0.027 0.018 0.012 0.029 1010 0.033 0.025 0.017 0.036 1014 0.024 0.017 0.011 0.029 1027 0.022 0.018 0.011 0.032 883 0.021 0.013 0.008 0.020 885 0.028 0.022 0.014 0.034 897 0.017 0.022 0.013 0.036 1512 0.020 0.016 0.011 0.023 1028 0.017 0.014 0.008 0.024 1208 0.018 0.013 0.009 0.020 1207 0.018 0.015 0.009 0.023 1008 0.027 0.026 0.017 0.038 1004 0.016 0.015 0.008 0.028 58373 0.021 81422 0.022 1004 0.021 0.018 0.011 0.028 2528 0.018 0.017 0.012 0.024 2007 0.020 0.017 0.012 0.024 35009 0.019 0.017 0.016 0.019 2610 0.013 0.015 0.006 0.035 2802 0.024 0.020 0.015 0.026 813 0.012 962 0.013 0.014 0.008 0.026 1038 0.016 0.017 0.010 0.028 960 0.017 0.022 0.012 0.038 1043 0.021 0.022 0.013 0.037 2254 0.016 0.016 0.011 0.023 651 0.021 3620 0.018 0.016 0.012 0.020 1733 0.024 0.020 0.014 0.028 1488 0.011 0.008 0.004 0.017 1516 0.020 0.023 0.013 0.040 1501 0.017 0.016 0.010 0.025 503 0.012 2551 0.031 850 0.020 1515 0.027 0.021 0.037 0.020 0.019 0.013 0.028 1508 0.017 0.017 0.011 0.026 1503 0.024 0.019 0.013 0.028 13 0.018 0.015 0.023 0.018 0.015 0.022 0.017 0.014 0.021 0.018 0.015 0.022 0.016 0.013 0.020 0.018 0.015 0.022 0.019 0.016 0.023 0.018 0.015 0.022 0.018 0.015 0.022 0.018 0.014 0.022 0.019 0.016 0.023 0.018 0.014 0.022 0.018 0.015 0.022 0.017 0.014 0.021 0.017 0.014 0.021 0.018 0.014 0.021 0.020 0.016 0.024 0.017 0.013 0.021 0.019 0.018 0.020 0.018 0.017 0.019 0.018 0.015 0.022 0.019 0.015 0.022 0.018 0.015 0.021 0.017 0.016 0.019 0.018 0.014 0.021 0.019 0.016 0.023 0.017 0.014 0.021 0.017 0.014 0.021 0.017 0.014 0.021 0.018 0.014 0.021 0.018 0.015 0.022 0.017 0.014 0.020 0.019 0.015 0.023 0.018 0.015 0.021 0.019 0.016 0.023 0.017 0.013 0.020 0.019 0.015 0.023 0.017 0.014 0.020 0.017 0.013 0.021 0.021 0.018 0.025 0.019 0.015 0.023 0.018 0.014 0.022 0.018 0.014 0.021 144 145 146 147 148 PSRAI PSRAI PSRAI PSRAI PSRAI 1502 0.029 0.025 0.017 0.035 1505 0.016 0.014 0.009 0.023 1089 0.020 0.020 0.012 0.033 2264 0.024 0.023 0.017 0.031 752 0.027 Samples not include in Cross-survey estimate due to Missing Post-stratification Variables 6 16 Harris Mitofsky 25 SRBI 31 UConn SurvRes 35 UConn SurvRes 39 SRC UMD 40 Harris 41 MarketFacts 43 Gallup 90 ICR 89 KnowledgeNetworks 109 SRBI 71 LA Times 3048 0.016 0.015 0.011 0.021 2002 0.019 0.019 0.013 0.028 5603 0.019 0.016 0.013 0.019 1015 0.019 0.016 0.010 0.027 1002 0.021 0.018 0.010 0.031 1824 0.011 0.013 0.008 0.021 2846 0.024 0.014 0.011 0.019 10204 0.017 0.012 0.010 0.015 1721 0.028 0.025 0.017 0.035 1001 0.008 0.007 0.004 0.015 3596 0.020 0.021 0.010 0.044 2910 0.023 0.020 0.016 0.026 1603 0.032 0.014 0.009 0.021 14 0.018 0.015 0.022 0.020 0.016 0.024 0.017 0.014 0.020 0.018 0.015 0.022 0.019 0.016 0.022 Model 3: Inclusion of Year of Survey LMER model: mod <- lmer(curreljw ~ racecat + agecat + edu + eduncg:agecat + metstat + state + year + (1|survs), family=binomial(link="logit")) All categorical variables (racecat, agecat, education, metropolitan status, state and year) were entered as “factors” in the model with the largest group within each factor as the reference group. For year, the reference group was the year 2004. TABLE B5: RESULTS FROM MULTILEVEL ANALYSIS WITH LAPLACE APPROXIMATION, JEWISH LIKELIHOOD ON DEMOGRAPHICS AND YEAR OF SURVEY. Cross-Survey Estimates Est. SD z value Pr(>|z|) Tau -- -- -- -- Variance 0 0 -- -- ICC -- -- -- -- MOR -- -- -- -- 58241 -- -- -- -3.74 0.08 -49.53 <.001 2001 0.06 0.04 1.59 0.11 2002 0.05 0.10 0.50 0.62 2003 -0.08 0.06 -1.26 0.21 2005 0.03 0.05 0.49 0.62 2006 0.03 0.05 0.58 0.56 2007 0.01 0.05 0.19 0.85 2008 -0.08 0.04 -2.01 0.04 2009 -0.01 0.05 -0.18 0.86 18-24 years -0.12 0.09 -1.27 0.21 25-34 years -0.36 0.05 -6.88 0.00 35-44 years -0.28 0.05 -6.01 0.00 55-64 years 0.11 0.04 2.49 0.01 65 years and older 0.22 0.04 4.91 0.00 -1.49 0.06 -23.63 <.001 Random Intercept for Survey Deviance Fixed Effects Intercept Year Age Education Less than College 15 Race/Ethnicity Black non-Hispanic -2.39 0.12 -20.41 <.001 Hispanic -1.66 0.09 -18.67 <.001 Other non-Hispanic -1.19 0.08 -15.09 <.001 18-24 years 0.88 0.12 7.17 0.00 25-34 years 0.24 0.11 2.24 0.03 35-44 years 0.18 0.10 1.88 0.06 55-64 years 0.17 0.09 1.97 0.05 65 years and older 0.41 0.08 5.17 0.00 1.24 0.06 20.59 <.001 Alabama -1.88 0.21 -8.85 <.001 Arizona -0.57 0.10 -5.82 0.00 Arkansas -1.91 0.29 -6.55 0.00 Colorado -0.73 0.11 -6.84 0.00 Connecticut -0.07 0.09 -0.74 0.46 Delaware -0.82 0.27 -3.01 0.00 District of Columbia 1.03 0.15 6.75 0.00 Florida 0.28 0.05 5.64 0.00 Georgia -0.79 0.10 -7.97 0.00 Idaho -1.72 0.30 -5.64 0.00 Illinois -0.37 0.07 -5.28 0.00 Indiana -1.80 0.16 -11.13 <.001 Iowa -1.95 0.25 -7.69 0.00 Kansas -1.70 0.23 -7.48 0.00 Kentucky -1.87 0.23 -8.23 <.001 Louisiana -1.85 0.22 -8.37 <.001 Maine Age x Education Metropolitan Status In Metro State -1.00 0.21 -4.68 0.00 Maryland 0.41 0.07 5.91 0.00 Massachusetts 0.13 0.06 2.05 0.04 Michigan -0.93 0.09 -10.14 <.001 Minnesota -1.39 0.14 -10.20 <.001 Mississippi -2.13 0.38 -5.60 0.00 Missouri -1.13 0.12 -9.17 <.001 Montana -2.14 0.50 -4.26 0.00 Nebraska -1.97 0.34 -5.87 0.00 Nevada -0.27 0.15 -1.87 0.06 New Hampshire -0.86 0.21 -4.01 0.00 New Jersey 0.53 0.06 9.52 <.001 New Mexico -0.71 0.19 -3.63 0.00 0.89 0.04 21.04 <.001 New York 16 North Carolina -1.20 0.11 -10.87 <.001 North Dakota -3.11 1.00 -3.11 0.00 Ohio -0.91 0.08 -11.30 <.001 Oklahoma -1.83 0.23 -7.86 0.00 Oregon -1.12 0.14 -7.97 0.00 Pennsylvania -0.35 0.06 -5.81 0.00 Rhode Island -0.61 0.21 -2.99 0.00 South Carolina -1.58 0.18 -8.58 <.001 South Dakota -3.28 1.00 -3.27 0.00 Tennessee -1.83 0.17 -10.85 <.001 Texas -1.32 0.08 -15.92 <.001 Utah -2.49 0.36 -7.00 0.00 Vermont -0.14 0.22 -0.65 0.51 Virginia -1.02 0.10 -10.20 <.001 Washington -1.12 0.11 -10.27 <.001 West Virginia -1.62 0.29 -5.56 0.00 Wisconsin -1.65 0.15 -10.84 <.001 Wyoming -1.02 0.41 -2.46 0.01 Note: The reference group are those age 45-54, white non-Hispanic, college graduates, residing in non-metropolitan California and survey year 2004. Model 4: Based to Single year: 2004 Estimates were similar to those obtained across the set of nine years (compare Table B6 to Table 2 of main report). The primary difference is that the smaller cell sizes associated with estimation within particular states lead to less reliable (wider confidence intervals) associated with state-level estimates (see Figure B2). TABLE B6: JEWISH POPULATION ESTIMATES USING SINGLE YEAR (2004) MODEL, POST-STRATIFIED TO 2004 CPS. Total US Adultsa Total Jewish Population Pct of Total US adults Pop Total All Groups Age 18-24 years 25-34 years 35-44 years 45-54 years 55-64 years 65+ years Pct 213,308,957 27,643,318 38,948,429 43,300,765 40,786,805 28,186,364 34,443,276 13.0 18.3 20.3 19.1 13.2 16.1 Total US Jewish Adults Pct(CI) Pop. Lower 95% Upper 95% CI CI 1.9 (1.7,2.1) 4,023,000 3,715,873 4,393,809 1.9 1.3 1.4 2.0 2.3 2.7 (1.6,2.2) (1.1,1.5) (1.3,1.6) (1.8,2.2) (2.0,2.5) (2.4,3.1) 520,898 504,386 617,964 796,997 638,603 943,842 431,193 618,416 439,740 579,687 544,098 703,396 716,353 888,963 571,055 717,319 833,900 1,056,509 17 Pct of Total US Jewish adults Pct. (CI) 12.9 12.5 15.4 19.8 15.9 23.5 (11.1,14.9) (11.2,14.0) (13.9,16.9) (18.4,21.3) (14.5,17.2) (21.7,25.4) Education Non-College College Grad Plus 159,546,537 53,762,420 74.8 25.2 1.1 4.3 (1.0,1.2) (3.9,4.6) 1,733,954 2,288,736 1,556,352 1,935,457 2,114,055 2,493,855 43.1 56.9 (40.7,45.2) (54.8,59.3) Race White, non-Hisp Black, non-Hisp. Hispanic Other non-Hisp. 150,951,210 23,825,461 26,496,449 12,035,838 70.8 11.2 12.4 5.6 2.5 0.2 0.3 1.2 (2.3,2.7) (0.1,0.3) (0.2,0.5) (1.0,1.6) 3,736,654 3,450,691 4,081,138 51,074 31,557 75,528 86,977 57,824 120,902 147,984 115,980 189,798 92.9 1.3 2.2 3.7 (91.6,94.0) (0.8,1.9) (1.5,2.9) (2.9,4.6) 173,843,153 81.5 2.2 (2.1,2.5) 3,898,138 96.9 (96.3,97.4) 39,465,804 18.5 0.3 (0.3,0.4) 124,552 3.1 (2.6,3.7) Metropolitan Metro Non-Metro 3,601,446 4,261,066 101,712 151,602 Notes: Source: 2004 Current Population Survey, March Supplement. FIGURE B2: STATE LEVEL ESTIMATES OF JEWISH POPULATION BASED TO 2004 SINGLE YEAR MODEL. 18 Model 5: Based to subset of years: 2004-2008 Parameter estimates are displayed in Table B7and mirror those observed in the full set of surveys from 2000 to 2008. TABLE B7: RESULTS FROM HIERARCHICAL BAYES POPULATION MODEL FOR 2004-2008 Cross-Survey Estimates Est. SD R-hat Random Intercepts for Survey Tau 69 Variance 0.014 ICC 0.004 MOR Deviance 1.07 1.12 41863.18 Fixed Effects Intercept -4.11 0.06 1.05 0.92 0.08 1.02 25-34 years 0.03 0.10 1.03 35-44 years -0.02 0.09 1.02 55-64 years 0.39 0.07 1.03 65 years and older 0.71 0.06 1.03 1.56 0.06 1.04 Black non-Hispanic -2.41 0.13 1.01 Hispanic -1.73 0.11 1.01 Other non-Hispanic -1.15 0.09 1.00 18-24 years -0.99 0.14 1.01 25-34 years -0.39 0.12 1.03 35-44 years -0.26 0.10 1.01 55-64 years -0.26 0.09 1.02 65 years and older -0.47 0.08 1.02 -1.34 0.07 1.01 Alabama -1.88 0.25 1.00 Arizona -0.55 0.11 1.01 Arkansas -2.00 0.34 1.00 Colorado -0.61 0.12 1.00 0.02 0.10 1.00 -0.54 0.28 1.00 Age 18-24 years Education College Grad Race/Ethnicity Age x Education Metropolitan Status Non Metro State Connecticut Delaware 19 District of Columbia 1.25 0.16 1.00 Florida 0.33 0.05 1.00 Georgia -0.76 0.11 1.00 Idaho -1.59 0.31 1.00 Illinois -0.33 0.07 1.00 Indiana -1.65 0.17 1.00 Iowa -1.84 0.27 1.00 Kansas -1.85 0.28 1.00 Kentucky -2.20 0.30 1.00 Louisiana -1.95 0.27 1.00 Maine -0.91 0.23 1.00 0.43 0.08 1.00 Maryland Massachusetts 0.20 0.07 1.00 Michigan -0.81 0.10 1.00 Minnesota -1.31 0.14 1.00 Mississippi -2.35 0.51 1.01 Missouri -1.13 0.14 1.00 Montana -1.90 0.50 1.00 Nebraska -1.71 0.33 1.00 Nevada -0.18 0.16 1.00 New Hampshire -0.66 0.23 1.00 New Jersey 0.54 0.06 1.00 New Mexico -0.53 0.19 1.00 0.90 0.04 1.00 North Carolina -1.10 0.11 1.00 North Dakota -3.15 1.06 1.00 Ohio -0.91 0.09 1.00 Oklahoma -1.93 0.29 1.00 Oregon -1.21 0.16 1.00 Pennsylvania -0.29 0.06 1.00 Rhode Island -0.71 0.25 1.00 South Carolina -1.51 0.21 1.00 South Dakota -3.22 1.06 1.00 Tennessee -1.89 0.21 1.00 Texas -1.25 0.09 1.00 Utah -2.55 0.41 1.01 Vermont 0.01 0.24 1.00 Virginia -0.99 0.12 1.00 Washington -1.28 0.13 1.00 West Virginia -1.49 0.32 1.00 Wisconsin -1.79 0.19 1.00 New York Wyoming -0.95 0.43 1.00 Note: Reference group, age 45-54, white non-Hispanic, college graduates , residing in metropolitan California. 20 Population Counts and Proportions by State: 2010 TABLE B8: TOTAL JEWISH POPULATION BY STATE: 2010. Total US Adults Prop. Jewish of All Adults Est. Pct(CI) Alabama 3,579,052 0.3 (0.2, 0.5) Arizona 4,800,864 1.5 (1.2, 1.8) Arkansas 2,144,490 0.3 (0.1, 0.5) California 27,322,887 2.3 (2.2, 2.5) Colorado 3,727,626 1.8 (1.4, 2.2) Connecticut 2,667,563 3.4 (2.8, 4.0) 672,703 1.4 (0.8, 2.2) Delaware 482,035 7.4 (5.6, 9.4) Florida 14,368,065 3.5 (3.2, 3.8) Georgia 7,078,286 1.1 (0.9, 1.3) Idaho 1,104,481 0.5 (0.3, 0.8) Washington, DC Illinois 9,570,396 1.9 (1.6, 2.1) Indiana 4,721,604 0.5 (0.3, 0.7) Iowa 2,280,184 0.4 (0.2, 0.6) Kansas 2,032,547 0.4 (0.2, 0.7) Kentucky 3,265,707 0.2 (0.1, 0.4) Louisiana 3,292,002 0.3 (0.2, 0.5) Maine 1,027,470 1.0 (0.6, 1.5) Maryland 4,318,837 3.9 (3.4, 4.5) Massachusetts 5,175,248 4.2 (3.7, 4.7) Michigan 7,454,419 1.2 (1.0, 1.4) Minnesota 3,954,543 0.8 (0.6, 1.0) Mississippi 2,064,766 0.2 (0.1, 0.4) Missouri 4,535,180 0.8 (0.6, 1.1) Montana 754,160 0.3 (0.1, 0.7) Nebraska 1,321,501 0.5 (0.2, 0.8) Nevada 1,964,403 1.9 (1.4, 2.6) New Hampshire 1,029,031 1.5 (0.9, 2.1) New Jersey 6,608,420 4.7 (4.2, 5.2) New Mexico 1,460,277 1.2 (0.8, 1.7) 14,719,227 5.8 (5.4, 6.2) New York 7,012,892 0.7 (0.6, 0.9) 488,495 0.2 (0.0, 0.6) Ohio 8,746,658 1.0 (0.8, 1.2) Oklahoma 2,708,166 0.4 (0.2, 0.6) Oregon 2,970,906 0.8 (0.6, 1.1) North Carolina North Dakota 21 Total Jewish Adults Population 12,200 95% lower 7,200 95% Upper 18,400 72,000 59,100 88,200 6,300 3,000 10,900 629,000 590,400 672,900 67,000 52,800 81,300 90,200 73,800 107,400 9,500 5,500 15,100 35,600 26,800 45,500 503,100 460,200 547,800 75,200 60,700 92,100 5,700 2,900 9,400 180,900 156,400 205,300 22,900 16,200 31,400 8,400 4,700 12,900 8,500 4,600 13,600 7,800 4,000 12,800 11,200 6,100 17,900 10,400 6,300 15,200 170,000 148,600 193,800 216,900 190,300 244,200 90,500 74,300 107,800 30,800 22,600 39,600 3,400 1,100 7,500 37,900 27,600 48,600 2,400 800 5,300 6,300 3,000 10,800 38,000 27,500 50,200 15,100 9,500 22,000 309,100 278,200 343,900 17,500 11,700 24,300 856,500 799,300 912,200 51,700 40,900 63,800 800 < 50 2,700 86,500 72,200 101,900 9,600 5,000 15,600 24,400 17,700 31,900 Pennsylvania Rhode Island South Carolina South Dakota Tennessee 9,591,280 2.0 (1.8, 2.3) 806,505 1.7 (1, 2.5) 3,394,958 0.5 (0.3, 0.7) 598,678 0.1 (0.0, 0.4) 4,761,287 0.4 (0.2, 0.5) Texas 17,691,183 0.6 (0.5, 0.7) Utah 1,922,109 0.2 (0.1, 0.5) Vermont 494,178 2.0 (1.2, 3.0) Virginia 5,859,015 1.1 (0.9, 1.3) Washington 5,156,507 0.9 (0.7, 1.1) West Virginia 1,412,776 0.5 (0.3, 0.9) Wisconsin 4,266,034 0.5 (0.3, 0.7) Wyoming 404,406 0.7 (0.3, 1.3) 22 194,100 172,900 217,700 13,400 7,900 20,400 16,800 10,900 24,400 800 < 50 2,700 18,000 11,600 25,600 103,500 86,400 120,400 4,600 1,800 8,700 9,800 6,000 15,000 62,600 50,500 77,100 45,800 35,300 57,800 7,500 3,700 12,600 20,000 13,800 28,000 2,800 1,000 5,400 Weighting: Survey of American Jews 2010 To identify factors necessary to include in weighting, the demographic composition of the SAJ 2010 sample was compared to the US adult Jewish population. In addition, the relationship of demographic variables to a representative sample of survey outcomes were examined. Post-stratification is done for those demographic variables that are both over- or under-represented relative to the total population and are related to survey outcomes. Sample Composition Basic demographic information is collected as part of the KnowledgePanel. To compare the KN sample to cross-survey estimates of the Jewish population, all KN demographic variables were recoded into standard format. Age (in years) was recoded into 6 categories (18-24/25-34/35-44/45-54/55-64/ 65+). Race/Ethnicity, originally recorded as 5 categories (White non-Hispanic/Black non-Hispanic, Other non-Hispanic, Hispanic, and More than Two Races non-Hispanic) was recoded into 4 categories of race, with the 2 or more category (n=22) included as Other non-Hispanic. Educational attainment was recoded from 4 categories (Less than high school/High school/Some college/Bachelor’s degree or higher) into two categories (Less than Bachelor’s/Bachelor’s degree or higher). Also included was metropolitan status (metro/non-metro) and state of residence.4 State-Metropolitan Clusters Given the sparseness of data at the state level, respondents were grouped into eight statemetropolitan clusters. These clusters were defined based on a combination of distributions of the Jewish population as estimated in the most recent years of crosssurvey data and based on the number of cases in the SAJ 2010 sample, such that low incidence regions would not be combined with higher incidence regions in a way that could result in biasing estimates either upward or downward. Population models were refit using these state-metro clusters rather than state and metropolitan areas to obtain corresponding population estimates for analysis of the SAJ 2010 sample. Table B9: SAJ 2010 sample size within state-metropolitan clusters. State-metro Clusters 4 Cluster 1 New York state metro & DC Cluster 2 Jewish Population Incidence SAJ 2010 Sample above 6% 150 New Jersey just under 5% 60 Cluster 3 MD & MA metro Just over 4% 72 Cluster 4 CT, VT, FL metro 3% - 4% 102 Cluster 5 PA & CA metro, NY non-metro 2% - 3% 194 All analyses limited to Continental U.S.. 23 Cluster 6 AZ, CO, NH, RI, NV, IL metro 1.5% - 2% 97 Cluster 7 NC, MN, MO, WA, VA, WY, OR, MI, DE, GA, OH, NM, ME metro & CT, CA, VT, FL, MA, MD non-met 1% - 1.5% 144 Cluster 8 All other metro & non-metro areasa Below 1% 75 Notes: a) Excludes Alaska & Hawaii. Demographic composition of the sample is displayed in Table B10. Included are distributions of each variable for the total US adult population and the US adult Jewish population (based on cross-survey estimates) in comparison to the full KN Jewish panel and the subset of panel members who responded to the SAJ 2010 survey. Since the crosssurvey analysis is based to those who identify as Jewish in response to questions about religious affiliation, the sample is further separated into those who identified as Jewish by religion and those who were added to the sample based on responses to the two follow-up questions. TABLE B10: COMPARISON OF SSRI ESTIMATES WITH KNOWLEDGE NETWORK’S JEWISH PANEL AND THE SAJ 2010 SAMPLE. US Adults: CPS 2010 Population % Jewish Population: SSRI 2010 Est. & Lower Upper KN All Jewish Panel Members N % SAJ 2010 Respondents SAJ 2010 Jewish by Religion SAJ 2010 Other Jewish N % N % N % 23 2.1 17 1.9 6 3.2 Age Age 18-24 years 12.8 12.6 11.4 13.8 34 Age 25-34 years 17.9 13.3 12.3 14.3 108 8.2 73 6.8 53 5.9 20 10.7 Age 35-44 years 17.6 13.5 12.7 14.4 167 12.6 128 11.8 92 10.3 36 19.3 Age 45-54 years 19.4 18.2 17.3 19.1 254 19.2 209 19.3 176 19.7 33 17.6 Age 55-64 years 15.4 18.9 17.9 19.9 374 28.3 317 29.2 273 30.5 44 23.5 Age 65 years plus 16.8 23.5 22.3 24.6 385 29.1 331 30.7 283 31.7 48 25.7 27.3 72.7 59.7 40.3 58.2 38.8 61.2 41.8 960 362 72.6 27.4 801 280 74.1 25.9 658 236 73.6 26.4 143 44 76.5 23.5 68.3 11.8 13.9 5.9 91.3 1.5 3.4 3.9 90.3 1.1 2.7 3.3 92.3 1.9 4.1 4.5 1234 8 36 44 93.3 0.6 2.7 3.3 1015 3 26 37 93.9 0.3 2.4 3.4 843 2 21 28 94.3 0.2 2.3 3.1 172 1 5 9 92 0.5 2.7 4.8 6.1 2.9 4.0 7.3 15.8 8.4 20.2 36.3 20.9 7.4 9.2 13.9 19.6 8.8 12.7 7.7 19.8 6.7 8.4 13 18.5 8 11.8 7.1 22 8.1 9.9 14.8 20.6 9.5 13.5 8.4 217 81 101 151 296 134 225 115 16.4 6.1 7.7 11.4 22.4 10.2 17 8.7 171 69 83 118 250 108 189 91 15.8 6.4 7.7 10.9 23.2 10 17.5 8.4 150 60 71 102 194 97 144 74 16.8 6.7 8 11.4 21.7 10.9 16.1 8.3 21 9 12 16 56 11 45 17 11.2 4.8 6.4 8.6 29.9 5.9 24.1 9.1 Education College Non-College Race/Ethnicity White-NonHispanic Black-NonHispanic Hispanic Other-NonHispanic State Group Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Cluster 6 Cluster 7 Cluster 8 24 2.6 Even though RDD and other general population surveys tend to under-represent younger age groups and over-represent older age groups (cf., Keeter, et al., 2000), the KN Jewish panel, along with the subset who responded to the SAJ 2010 survey, does so to an even greater degree. For typical RDD surveys, percentages aged 55 years and older tend to be just a few percentages higher than the US as a whole, whereas the KN Jewish panel is over 60% aged 55 years and older. The KN Jewish panel also over-represents college graduates and under-represents -- to a lesser degree -- racial/ethnic minorities. These differences are true of the sample overall as well as for the subgroups who identify by religion and those who do not. There are also disparities in the representation of the panel and sample by geographic dispersion. Clusters 1 through 4 are under-represented (see Figure B3). These are the geographic areas with the highest Jewish population incidence, New York metropolitan, New Jersey, Maryland and Massachusetts metropolitan areas, and Connecticut, Vermont and Florida metropolitan areas. State clusters 5, 6, and to the largest degree 7 are overrepresented. These include the metropolitan areas of Pennsylvania and California, the non-metropolitan areas of New York, and metropolitan areas in states such as North Carolina, Georgia, Ohio, Minnesota, Missouri, Wyoming, and Michigan. FIGURE B3: DISTRIBUTION OF KN PANEL AND SAJ 2010 RESPONDENTS BY STATE GROUPS. 25 Knowledge Network Sampling Weight KN provides a base sampling weight to account for the idiosyncrasies of the panel design, and potential over- or under-representation of particular groups. These adjustments include under-sampling of telephone numbers unmatched to valid mailing addresses, RDD selection proportional to the number of telephone landlines in the household, oversampling of Chicago and Los Angeles, oversampling of the four largest states and central region states, under-sampling of households not covered by the MSN TV network service, oversampling of African-American and Hispanic telephone exchanges, address-based sample phone matching, and ABS oversample stratification adjustment which included oversampling of high density minority communities (Knowledge Networks, 2010). The average weight was .89, corresponding to an effective sample size of 958.5 Weights ranged from .22 to 4.74, a ratio of 21.5 of lowest to highest. Weights were below 1 for 75% of the sample (see Figure B4). FIGURE B4: DISTRIBUTION OF SAMPLE WEIGHTS FOR SAJ 2010. The distribution of sample demographics after weighting for these factors is displayed in Table 11. The weighting adjustments have little effect on the differences between the sample and the US Jewish population as a whole. Disproportionate representation of older adults and college graduates remain. Over- and under-representation of metropolitan areas within states also remain. For respondents who identified as Jewish 5 The KN sampling weight was not standardized to the observed sample size. Thus, sample counts sum to 958 rather than the observed 1,079 cases. 26 but not by religion, the sampling weights alter the distributions by metropolitan status substantially, particularly with far greater weight given to Jews in non-metropolitan areas of Texas, Ohio and Washington State. TABLE 11: DEMOGRAPHIC CHARACTERISTICS OF SAJ 2010 SAMPLE WEIGHTED BY SAMPLE WEIGHT. SAJ 2010 Respondents N Total All Groups % 958 SAJ 2010 Jewish by Religion N % SAJ 2010 Other Jewish N 787 % 171 Age Age 18-24 years 23 2.4 18 2.2 6 3.3 Age 25-34 years 75 7.9 55 7.0 20 12.0 Age 35-44 years 115 12.0 84 10.7 31 18.0 Age 45-54 years 191 19.9 155 19.7 36 21.1 Age 55-64 years 271 28.3 227 28.9 44 25.8 Age 65 years plus 283 29.5 249 31.6 34 19.9 College 697 72.8 572 72.6 125 73.3 Non-College 261 27.2 215 27.4 46 26.7 901 94.1 742 94.3 159 93.2 . Education Race/Ethnicity White-NonHispanic Black-NonHispanic 4 0.4 3 0.4 1 0.5 Hispanic 21 2.2 18 2.2 4 2.1 Other-NonHispanic 32 3.3 25 3.1 7 4.2 Cluster 1 150 15.6 133 16.9 17 9.8 Cluster 2 75 7.9 63 8.0 12 7.1 Cluster 3 86 8.9 72 9.1 14 8.2 Cluster 4 115 12.0 96 12.2 19 11.0 Cluster 5 205 21.4 161 20.4 44 25.8 Cluster 6 90 9.4 81 10.3 9 5.5 Cluster 7 163 17.0 124 15.8 39 22.7 Cluster 8 75 7.8 58 7.4 17 9.8 State Group 27 Adjustments for non-response Knowledge Networks made available sociodemographic information on all Jewish panel members. Analyses were conducted to examine differences between the 1,082 panel members who responded to the SAJ 2010 survey and the 267 who did not (Phillips, 2010). Analyses consisted of logistic regressions (weighted by the KN sampling weight) predicting the likelihood of response (yes/no) as a function of demographic variables. It was determined that age, household size and geographic region were significantly related to the likelihood of responding. The KN sampling weight was, therefore, adjusted to reflect the distributions of these variables in the full KN Jewish panel. Adjustments were done using QBAL (Werner, 2004) and included 6 categories of age, 4 categories of household size (1/2/3/4+) and nine categories of geographic region (Massachusetts, Other New England, New York, Mid-Atlantic, Midwest, South Atlantic, Florida, South, Mountain & Pacific). The adjustments appeared to have very little effect on the overall sampling weights, with a mean similarly of .89, range .19 to 5.14 and just under 75% of the sample with weights below 1 (see Figure B5).6 FIGURE B5: SAMPLE WEIGHTS ADJUSTED FOR NON-RESPONSE FOR SAJ 2010. 6 Typically adjustments for non-response are based to a sample frame. The adjustments here treat the knowledge network sample as if it were a frame and balances in terms of how the 200 or so panel members who didn’t respond differ from the 1,000 that did. Such an adjustment assumes that the KN Jewish panel is representative of all US Jewish adults, which based on comparison to Jewish population estimates, it appears not to be. Perhaps if it were a larger representative sample such adjustments might make sense, but in this context, it’s not clear adjustments of the sample achieved to the panel is useful. 28 Adjustments for non-response reduce somewhat the disparities by age, but overall appear to have little effect on sample demographics (see Table B12). TABLE B12: DEMOGRAPHIC CHARACTERISTICS OF SAJ 2010 SAMPLE WEIGHTED BY NONRESPONSE WEIGHT SAJ 2010 Respondents N % SAJ 2010 Jewish by Religion N % SAJ 2010 Other Jewish N % Age Age 18-24 years 27 2.8 20 2.6 7 4.1 Age 25-34 years 86 8.9 63 8.0 23 13.3 Age 35-44 years 125 13.0 91 11.6 33 19.2 Age 45-54 years 187 19.5 153 19.5 34 19.7 Age 55-64 years 265 27.7 222 28.2 43 25.2 Age 65 years plus 269 28.1 237 30.2 32 18.5 College 699 72.9 573 72.9 126 73.3 Non-College 259 27.1 213 27.1 46 26.7 900 94.0 740 94.2 160 93.1 Education Race/Ethnicity White-NonHispanic Black-NonHispanic 4 0.4 3 0.4 1 0.5 Hispanic 22 2.3 18 2.3 4 2.2 Other-NonHispanic 32 3.4 25 3.2 7 4.3 Cluster 1 157 16.4 139 17.7 18 10.3 Cluster 2 75 7.8 63 8.0 12 6.8 State Group Cluster 3 81 8.4 68 8.6 13 7.5 Cluster 4 122 12.8 102 12.9 21 12.1 Cluster 5 197 20.6 155 19.8 42 24.2 Cluster 6 87 9.1 78 9.9 9 5.1 Cluster 7 164 17.1 123 15.6 41 23.6 Cluster 8 76 7.9 58 7.4 18 10.2 29 Adjustments for Representativeness of the Sample To account for the disproportionate representation of demographic groups within the Jewish population, an additional stage of weighting was done to balance the sample demographics to match Jewish population estimates for age, race, education, the interaction of age and education interaction, as well as geographic region. This was done for the subset of 892 respondents who identified as Jewish in response to religious identification, which is the group most similar to those on which the population estimates are also based. Relation of Post-stratification Variables to Outcomes Prior to weighting, relationships between outcomes and potential post-stratification variables was examined. These variables included: Had a Bar or Bat Mitzvah (As a child/As an adult/No) Jewish education o Years of supplementary school (0 years/1-5 years/6-12 years) o Years of day school (0 years/Any years) o Years of camping (0 years/1-3 years/4 or more years) Intermarriage (Yes/No) Ability to read and understand Hebrew (Don’t know Hebrew alphabet at all/Can read the letters but not understand the words/Some of what I read/Most of what I read/Everything I read) Synagogue membership (Yes/Other in household/No) Attending any Jewish life cycle event (brit, baby naming ceremony, bar/bat mitzvah, wedding, shivah call) in past year (Yes any event, No) Any synagogue honor (aliyah, read Torah, chanted Haftarah/other) Relationships with post-stratification variables were examined using logistic (or ordinal/multinomial) regressions. All analyses were conducted using the SVY functions in STATA to take into account variance estimation using the sampling and non-response weights. Table B13 displays bivariate statistics for bar/bat mitzvah, Jewish education and intermarriage. Only 28 of the 892 respondents indicated that they had a bar or bat mitzvah as an adult. This variable was, therefore, dichotomized into a 0/1, yes/no. Age was significantly related to the likelihood of having a bar or bat mitzvah. All age groups under age 44 were significantly more likely to have had a bar/bat mitzvah then older adults. Age was also significantly related to Jewish education and intermarriage. Education and race were significantly related to the number of years of supplemental Jewish education, but were unrelated to the other variables. Metropolitan status was related to the likelihood of intermarriage, but otherwise unrelated to the other measures. State cluster was related to all of these outcome variables in some way. 30 Table B13 through Table B16 display results of bivariate tests of association for reading/understanding Hebrew, synagogue membership, attending Jewish life cycle event in the past year and receiving any synagogue honors in the past year. Again, age is related to all of these activities, with the exception of some of the synagogue honors. Being called to Torah for Aliyah, reading Torah and chanting Haftarah are unrelated to age. Education is related to the likelihood of having a bar/bat mitzvah and attending a wedding, but unrelated to attending a brit or making a shiva call. Race is unrelated to these activities, with the exception of a weak, marginally significant relationship with making a shiva call. Metropolitan status, and state cluster are also unrelated to these activities, with the exception of shiva calls, which are significantly more likely in metropolitan than non-metropolitan areas and more likely in the highest density Jewish population state clusters. 31 TABLE B13: BIVARIATE RELATIONSHIPS OF HAVING A BAR/BAT MITZVAH, YEARS OF JEWISH EDUCATION AND INTERMARRIAGE WITH AGE, EDUCATION, RACE, METROPOLITAN STATUS AND GEOGRAPHIC REGION. Had Bar/Bat Mitzvah (n=812) Coef. Std. Err. Jewish Ed: Supplemental (n=869) t P>t Coef. Std. Err. 0.143 -1.500 0.133 -0.678 Jewish Ed: Day School (n=864) t P>t Coef. Std. Err. 0.130 -5.230 0.000 3.837 Intermarriage (n=527 married) t P>t Coef. Std. Err. t P>t 0.420 9.130 0.000 1.152 0.211 5.460 0.000 Age Constant -0.216 Age 18-24 3.122 1.063 2.940 0.003 2.345 0.664 3.530 0.000 0.536 1.125 0.480 0.634 Age 25-34 1.341 0.393 3.420 0.001 0.490 0.297 1.650 0.099 2.205 0.595 3.710 0.000 -0.871 0.498 -1.750 0.081 Age 35-44 0.649 0.292 2.220 0.027 0.274 0.268 1.020 0.307 2.033 0.521 3.900 0.000 -0.816 0.368 -2.220 0.027 Age 45-54 0.771 0.235 3.280 0.001 0.763 0.200 3.810 0.000 1.447 0.515 2.810 0.005 -0.942 0.295 -3.190 0.001 Age 55-64 0.256 0.205 1.250 0.212 0.607 0.170 3.560 0.000 1.187 0.518 2.290 0.022 -0.793 0.272 -2.920 0.004 0.367 0.093 3.960 0.000 -1.201 0.098 -12.210 0.000 2.537 0.170 14.920 0.000 0.512 0.115 4.460 0.000 0.196 -3.220 0.001 -0.464 0.180 -2.580 0.010 -0.261 0.348 -0.750 0.454 0.154 0.243 0.640 0.526 0.083 0.011 -1.102 0.091 -12.150 0.000 2.657 0.157 16.930 0.000 0.574 0.105 5.480 0.000 -14.879 0.841 -17.690 0.000 -11.439 0.851 -13.440 0.000 Education Constant College grad -0.630 Race Constant 0.210 Black nH -- Hispanic 0.079 Other nH -0.107 0.605 2.540 0.130 0.896 -0.822 0.467 -1.760 0.079 1.175 0.692 1.700 0.090 -0.071 0.575 -0.120 0.901 0.511 -0.210 0.834 0.113 0.520 0.220 0.827 0.618 0.684 0.900 0.367 -0.694 0.555 -1.250 0.212 Metropolitan Constant 0.195 0.082 2.370 0.018 -1.090 0.091 -12.000 0.000 2.586 0.150 17.250 0.000 0.602 0.104 5.770 0.000 Non-Metro 0.229 0.486 0.470 0.638 -0.479 0.322 -1.490 0.138 -0.531 1.032 -0.510 0.607 -1.403 0.512 -2.740 0.006 Constant 0.198 0.177 1.120 0.262 -0.976 0.154 -6.330 0.000 3.025 0.355 8.530 0.000 0.180 0.222 0.810 0.420 Cluster 1 -0.088 0.270 -0.330 0.744 -0.064 0.204 -0.310 0.755 1.016 0.447 2.270 0.023 1.223 0.388 3.150 0.002 Cluster 2 0.184 0.344 0.530 0.593 -0.119 0.261 -0.460 0.649 0.141 0.734 0.190 0.848 0.599 0.422 1.420 0.156 Cluster 3 0.301 0.331 0.910 0.364 0.728 0.264 2.760 0.006 0.662 0.645 1.030 0.305 1.071 0.457 2.350 0.019 Cluster 4 -0.090 0.297 -0.300 0.762 -0.007 0.289 -0.020 0.982 0.179 0.626 0.290 0.775 1.099 0.415 2.650 0.008 Cluster 5 -0.264 0.295 -0.890 0.371 -0.049 0.257 -0.190 0.848 0.207 0.563 0.370 0.714 0.193 0.343 0.560 0.573 Cluster 6 0.116 0.265 0.440 0.662 0.604 0.255 2.370 0.018 0.341 0.495 0.690 0.491 -0.250 0.324 -0.770 0.440 Cluster 7 -0.104 0.345 -0.300 0.763 -0.218 0.276 -0.790 0.430 0.439 0.698 0.630 0.530 -0.356 0.378 -0.940 0.347 State Cluster 32 TABLE B14: BIVARIATE RELATIONSHIPS OF ATTENDING JEWISH LIFE CYCLE EVENTS WITH AGE, EDUCATION, RACE, METROPOLITAN STATUS AND GEOGRAPHIC REGION. Brit/Baby Naming (n=887) Coef. Std. Err. t Bar/Bat Mitzvah (n=884) P>t Coef. Std. Err. t Wedding (n=886) P>t Coef. Std. Err. t Shiva Call (n=884) P>t Coef. Std. Err. t P>t Age Constant -1.308 0.173 -7.550 0.000 -0.259 0.139 -1.860 0.063 -1.308 0.157 -8.320 0.000 -0.029 0.140 -0.210 0.837 Age 18-24 -1.040 1.057 -0.980 0.326 -1.177 0.718 -1.640 0.102 -0.480 0.758 -0.630 0.527 -2.240 0.883 -2.540 0.011 Age 25-34 -0.145 0.405 -0.360 0.721 -1.149 0.377 -3.050 0.002 0.817 0.351 2.330 0.020 -1.156 0.372 -3.110 0.002 Age 35-44 -0.796 0.370 -2.150 0.032 -0.071 0.275 -0.260 0.795 -0.286 0.315 -0.910 0.364 -1.177 0.315 -3.730 0.000 Age 45-54 -0.518 0.306 -1.690 0.091 -0.147 0.222 -0.660 0.508 -0.192 0.271 -0.710 0.479 -0.367 0.224 -1.640 0.102 Age 55-64 -0.171 0.247 -0.690 0.488 -0.527 0.204 -2.580 0.010 0.321 0.219 1.470 0.142 -0.168 0.199 -0.840 0.399 Constant -1.476 0.114 -12.930 0.000 -0.393 0.090 -4.380 0.000 -1.097 0.100 -10.960 0.000 -0.337 0.090 -3.770 0.000 College grad -0.313 0.242 -1.290 0.197 -0.566 0.188 -3.010 0.003 -0.477 0.205 -2.330 0.020 -0.136 0.190 -0.720 0.473 -1.542 0.104 -14.830 0.000 -0.534 0.081 -6.560 0.000 -1.210 0.091 -13.360 0.000 -0.323 0.081 -3.960 0.000 0.595 0.270 0.789 -0.255 0.479 -0.530 0.594 -0.363 0.587 -0.620 0.537 -0.773 0.554 -1.390 0.163 0.455 0.590 0.557 -1.069 0.507 -2.110 0.035 Education Race Constant Black nH Hispanic 0.159 Other nH -0.384 0.529 -0.730 0.468 0.299 0.423 0.710 0.479 0.267 Constant -1.522 0.101 -15.000 0.000 -0.517 0.080 -6.460 0.000 -1.202 0.089 -13.500 0.000 -0.331 0.080 -4.130 0.000 Non-Metro -1.695 1.029 -1.650 0.100 -0.628 0.476 -1.320 0.187 -0.392 0.538 -0.730 0.466 -1.761 0.635 -2.780 0.006 Constant -1.622 0.232 -6.990 0.000 -0.630 0.170 -3.700 0.000 -1.367 0.208 -6.570 0.000 -0.968 0.181 -5.350 0.000 Cluster 1 0.122 0.329 0.370 0.711 0.249 0.259 0.960 0.336 0.116 0.293 0.400 0.692 1.333 0.265 5.020 0.000 Cluster 2 0.181 0.409 0.440 0.658 0.327 0.326 1.000 0.317 0.272 0.368 0.740 0.459 1.527 0.350 4.360 0.000 Cluster 3 0.322 0.395 0.820 0.415 0.509 0.320 1.590 0.112 0.264 0.384 0.690 0.493 0.791 0.329 2.410 0.016 Cluster 4 -0.069 0.433 -0.160 0.873 0.215 0.297 0.720 0.469 0.391 0.326 1.200 0.231 0.961 0.294 3.270 0.001 Cluster 5 0.133 0.369 0.360 0.719 0.046 0.281 0.160 0.870 0.294 0.316 0.930 0.351 0.254 0.288 0.880 0.378 Cluster 6 0.162 0.326 0.500 0.620 -0.112 0.258 -0.430 0.665 0.227 0.297 0.770 0.444 0.213 0.266 0.800 0.425 Cluster 7 -0.543 0.470 -1.160 0.248 -0.583 0.335 -1.740 0.082 -0.544 0.392 -1.390 0.165 -0.455 0.348 -1.310 0.191 Metropolitan State Cluster 33 TABLE B15: BIVARIATE RELATIONSHIPS OF READING HEBREW AND SYNAGOGUE MEMBERSHIP WITH AGE, EDUCATION, RACE, METROPOLITAN STATUS AND GEOGRAPHIC REGION. Read & Understand Hebrew (n=890) Coef. Std. Err. T Synagogue Membership (n=894) P>t Coef. Std. Err. 0.136 -1.580 0.113 -0.208 0.141 t P>t Age Constant -0.216 -1.480 0.139 Age 18-24 0.554 0.304 1.820 0.069 0.794 0.565 1.400 0.161 Age 25-34 0.857 0.264 3.250 0.001 -0.297 0.337 -0.880 0.378 Age 35-44 0.065 0.243 0.270 0.788 -0.279 0.278 -1.000 0.316 Age 45-54 0.664 0.215 3.090 0.002 0.080 0.226 0.360 0.723 Age 55-64 0.282 0.194 1.450 0.147 -0.403 0.201 -2.010 0.045 Education Constant -0.668 0.090 -7.420 0.000 -0.320 0.089 -3.590 0.000 College grad -0.609 0.187 -3.260 0.001 -0.064 0.187 -0.340 0.733 Constant -0.502 0.082 -6.090 0.000 -0.350 0.082 -4.290 0.000 Black nH -14.787 0.840 -17.600 0.000 1.833 1.417 1.290 0.196 Hispanic -0.331 0.438 -0.750 0.451 0.264 0.474 0.560 0.578 Other nH 0.550 0.393 1.400 0.162 -0.012 0.427 -0.030 0.978 Race Metropolitan Constant -0.524 0.082 -6.420 0.000 -0.332 0.080 -4.140 0.000 Non-Metro -0.660 0.414 -1.590 0.111 -0.169 0.424 -0.400 0.691 Constant -0.289 0.156 -1.860 0.064 -0.655 0.176 -3.730 0.000 Cluster 1 0.376 0.251 1.500 0.135 0.201 0.269 0.750 0.455 Cluster 2 0.094 0.311 0.300 0.763 0.587 0.331 1.770 0.077 Cluster 3 0.656 0.276 2.380 0.018 0.440 0.321 1.370 0.170 Cluster 4 0.075 0.266 0.280 0.778 0.650 0.291 2.240 0.026 Cluster 5 -0.166 0.255 -0.650 0.514 0.099 0.282 0.350 0.725 Cluster 6 0.482 0.227 2.120 0.034 0.627 0.261 2.410 0.016 Cluster 7 0.206 0.302 0.680 0.496 0.034 0.316 0.110 0.915 State Cluster 34 TABLE B16: BIVARIATE RELATIONSHIPS OF SYNAGOGUE HONORS WITH AGE, EDUCATION, RACE, METROPOLITAN STATUS AND GEOGRAPHIC REGION. Called to Torah for Aliyah (n=889) Coef. Std. Err. t Read Torah (n=885) P>t Coef. -3.366 Std. Err. t Chanted Haftarah (n=886) P>t Coef. 0.000 -3.056 Std. Err. t Age Constant -1.092 0.167 -6.540 0.000 Age 18-24 -0.910 0.794 -1.150 0.252 Age 25-34 -0.639 0.468 -1.360 0.173 1.020 0.636 1.600 0.109 0.702 0.622 1.130 Age 35-44 -0.609 0.392 -1.550 0.120 0.516 0.736 0.700 0.483 -1.439 1.052 -1.370 Age 45-54 -0.054 0.259 -0.210 0.834 0.297 0.595 0.500 0.618 0.021 0.507 0.040 Age 55-64 -0.550 0.256 -2.150 0.032 0.426 0.463 0.920 0.358 0.104 0.433 0.240 Education constant -1.218 0.108 -11.290 0.000 -2.833 0.203 -13.990 0.000 -2.924 College grad -0.618 0.263 -2.350 0.019 -1.060 0.685 -1.550 0.122 -0.600 Race constant -1.336 0.101 -13.240 0.000 -3.122 0.204 -15.310 0.000 -3.112 4.605 1.430 3.220 0.001 0.835 0.696 1.200 0.231 Black nH 0.327 -10.310 0.301 -10.160 P>t 0.000 Other (n=855) Coef. Std. Err. t P>t -1.622 0.211 -7.680 0.000 -1.623 1.070 -1.520 0.130 0.260 -3.192 1.035 -3.080 0.002 0.172 -1.149 0.484 -2.370 0.018 0.967 0.314 0.305 1.030 0.304 0.810 -0.499 0.318 -1.570 0.117 0.196 -14.920 0.000 -1.771 0.132 -13.420 0.000 0.469 -1.280 0.202 -0.423 0.329 -1.280 0.200 0.191 -16.330 0.000 -1.881 0.127 -14.800 0.000 1.076 0.679 1.580 0.113 0.085 0.670 0.130 0.899 0.514 0.770 0.670 0.505 0.282 0.540 0.520 0.601 Hispanic -0.490 0.653 -0.750 0.453 Other nH -0.571 0.554 -1.030 0.303 Metropolitan constant -1.355 0.100 -13.550 0.000 -3.080 0.183 -16.800 0.000 -1.846 0.123 -15.050 0.000 Non-Metro -0.344 0.532 -0.650 0.518 0.507 0.763 0.660 0.506 -1.314 1.033 -1.270 0.204 State Cluster constant -1.348 0.202 -6.660 0.000 -3.021 0.401 -7.540 0.000 -2.639 0.345 -7.650 0.000 -2.004 0.263 -7.630 0.000 Cluster 1 -0.068 0.340 -0.200 0.841 -0.372 0.620 -0.600 0.549 -0.461 0.545 -0.850 0.398 0.120 0.435 0.280 0.782 Cluster 2 0.581 0.367 1.580 0.114 0.560 0.763 0.730 0.464 -0.549 0.809 -0.680 0.498 0.169 0.469 0.360 0.718 Cluster 3 0.114 0.402 0.280 0.777 -0.061 0.716 -0.080 0.932 -0.279 0.629 -0.440 0.658 0.698 0.447 1.560 0.119 Cluster 4 0.319 0.333 0.960 0.338 -0.189 0.725 -0.260 0.795 -0.333 0.629 -0.530 0.597 0.611 0.406 1.500 0.133 Cluster 5 -0.284 0.362 -0.780 0.433 -0.059 0.716 -0.080 0.934 -0.796 0.683 -1.170 0.244 -0.356 0.430 -0.830 0.409 Cluster 6 -0.212 0.306 -0.690 0.488 0.344 0.602 0.570 0.568 -1.259 0.683 -1.840 0.065 -0.262 0.395 -0.660 0.508 Cluster 7 -1.066 0.509 -2.090 0.037 -1.003 0.839 -1.200 0.232 -0.290 0.643 -0.450 0.652 -0.124 0.510 -0.240 0.809 35 Post-Stratification 1: Raking to Demos & State-Metro Clusters Since all of the potential post-stratification variables were related to outcomes of interest in some way, all were included in post-stratification adjustments. The six categories of age, four categories of race, two categories of education across the metropolitan and nonmetropolitan areas of the 48 states (plus DC) used in the SSRI population models result in possible population estimates for over 4,700 demographic groups. Clearly with a sample size of 1,000, there are many groups (i.e., post-stratification cells) for which there are no corresponding observations in the KN sample. In such situations where there are too few observations within a survey for some of the sub-groups to estimate them reliably (or when population estimates for all of the sub-groups are not available), weights are typically “raked”, such that the marginal distributions of the weighted totals for each separate demographic variable align with the marginal distributions of the target population. Thus, the weighted sample estimates for the six categories of age are aligned to the population estimates for the same six categories of age, ignoring differences within specific age categories that there might be between states, education levels, etc.. Raking uses iterative procedures such that after adjustments to the weights for each variable are made, the process is repeated until the weights no longer change after having adjusted for previous variables in the model. With revised population estimates based to the eight state-metro clusters, raking adjustments were done using the rake procedure in the R Survey package. Table B17 displays sample composition variables after raking. Included for comparison are the SSRI 2010 population estimates from revised state-metro model, along with unweighted and weighted estimates based to the KN sampling weight. Two raked estimates are included in the table. The first represents weighting to the total Jewish population. The second represents these weights normalized to the observed sample size. TABLE B17: DEMOGRAPHIC CHARACTERISTICS OF NJPS 2010 SAMPLE WITH DEMOGRAPHIC AND STATE GROUPS RAKING ADJUSTMENTS. Age ssri cross-survey estimates Est. Lower Upper unweighted N Est. weighted with sampwt SE N Est. raked to dems & stategroup SE N Est. SE raked to dems & stategroup (normalized) N Est. SE 18-24 0.127 0.115 0.140 17 0.019 0.0046 18 0.022 0.006 531,019 0.127 0.033 113 0.127 0.033 25-34 0.132 0.122 0.143 53 0.059 0.0079 55 0.070 0.010 552,756 0.132 0.019 118 0.132 0.019 35-44 0.134 0.126 0.143 92 0.103 0.0102 84 0.107 0.012 564,386 0.135 0.015 120 0.135 0.015 45-54 0.181 0.172 0.190 175 0.196 0.0133 155 0.197 0.015 754,467 0.180 0.019 161 0.180 0.019 55-64 0.188 0.179 0.198 272 0.305 0.0154 227 0.289 0.017 789,276 0.188 0.016 168 0.188 0.016 65+ 0.238 0.226 0.249 283 0.317 0.0156 249 0.316 0.018 997,439 0.238 0.019 212 0.238 0.019 College 0.596 0.582 0.611 657 0.737 0.0148 572 0.726 0.018 2,501,024 0.597 0.030 533 0.597 533 No College 0.404 0.389 0.418 235 0.263 0.0148 215 0.274 0.018 1,688,319 0.403 0.030 360 0.403 360 White NH 0.913 0.903 0.923 842 0.944 0.0077 742 0.943 0.009 3,826,950 0.913 0.019 815 0.913 0.019 Black NH 0.015 0.011 0.019 2 0.002 0.0016 3 0.004 0.003 62,321 0.015 0.012 13 0.015 0.012 Education Race 36 Hispanic 0.033 0.027 0.040 21 0.024 0.0051 18 0.022 0.005 140,586 0.034 0.009 30 0.034 0.009 Other NH 0.038 0.032 0.045 27 0.030 0.0057 25 0.031 0.006 159,485 0.038 0.013 34 0.038 0.013 Cluster 1 0.209 0.198 0.220 150 0.168 0.0125 159 0.202 0.015 875,983 0.209 0.020 187 0.209 0.020 Cluster 2 0.074 0.067 0.081 60 0.067 0.0084 133 0.169 0.015 310,037 0.074 0.011 66 0.074 0.011 Cluster 3 0.092 0.084 0.099 71 0.080 0.0091 63 0.080 0.011 382,931 0.091 0.016 82 0.091 0.016 Cluster 4 0.139 0.130 0.148 102 0.114 0.0107 72 0.091 0.012 582,736 0.139 0.028 124 0.139 0.028 Cluster 5 0.196 0.185 0.206 194 0.217 0.0138 96 0.122 0.013 820,784 0.196 0.020 175 0.196 0.020 Cluster 6 0.088 0.080 0.095 97 0.109 0.0104 161 0.204 0.015 367,813 0.088 0.011 78 0.088 0.011 Cluster 7 0.127 0.118 0.135 144 0.161 0.0123 81 0.103 0.011 531,414 0.127 0.018 113 0.127 0.018 Cluster 8 0.077 0.071 0.084 74 0.083 0.0092 124 0.158 0.014 317,645 0.076 0.011 68 0.076 0.011 State Cluster The goal in weighting is to select variables and methods that decrease bias – the extent to which sample estimates deviate from the population -- without increasing the standard error such that the benefits of decreased bias are negated by the increase in variance. As can be seen in Table B17, raking decreased bias for all of the demographic variables. For example, the unweighted sample estimate of the proportion of the Jewish population aged 18 to 24 years of age was less than 2%. This underestimates the representation of this group by nearly 10% when compared to the 11.4% observed in the cross-survey population models. Raking to population distributions of age yielded a sample estimate of 12.7% for this group, much more closely aligned to the population estimate. The variance of the estimate, however, increases from an SE of .019 for the unweighted estimate to .033 after raking. Further, the range of weights increases dramatically, from a ratio of 22 (0.22 to 4.74) in the sampling weight to 230 (0.10 to 23) in the raked weight. Figure B6 displays comparison of the sampling weight to the raked weight, with points identified by age group to examine whether larger weights are identified with particular age groups. There are just 16 cases with weights greater than 5. Most of these clustered around 5 and less than 10 and are primarily respondents in the youngest age group, the group most underrepresented in the sample. Not visible due to the clustering on the x-axis, but 30 cases have weights below .2, primarily those aged 55 to 64 who are over-represented in the sample. 37 FIGURE B6: COMPARISON OF KN SAMPLING WEIGHT TO RAKED WEIGHT BY AGE GROUP. The MSE provides a metric by which to gauge changes in bias relative to increases in variance. The lower the MSE, the better the weighting adjustments. Figure B7 displays the MSE of unweighted estimates across distributions of basic demographics involved in weighting in comparison to the MSE of estimates weighted by the KN sampling weight and after raking. Although there is a small increase in variance after raking, as would be expected, the MSE is greatly reduced after raking. 38 0.07 0.06 0.05 0.04 Variance 0.03 MSE 0.02 0.01 0 Unweighted Sampling Weight Raked State-Met Clusters FIGURE B7: COMPARISON OF MSE AND VARIANCE OF UNWEIGHTED ESTIMATES ACROSS DISTRIBUTION OF BASIC DEMOGRAPHICS, ESTIMATES WITH KN SAMPLING WEIGHT AND ESTIMATES AFTER RAKING. Post-Stratification 2: Raking to Demos & Region x Metropolitan Status Prior to the development of state-metro clusters, the KN survey was post-stratified to demographic groups of age, race, education, metropolitan status and 9 geographic regions. These regions were based on the typical conceptual geographic clustering of US states within census divisions, which are defined independently of the distribution of the Jewish population. The exception, however, was that the highest density state of New York, along with the state of Florida were included as separate regions. The nine geographic clusters were: New England: CT, ME, MA, NH, RI & VT Mid-Atlantic: NJ & PA New York State Midwest: IL, IN, IA, KS, MI, MN, MO, NE, ND, OH, SD, WI South Atlantic: DE, DC, GA, MD, NC, SC, VA, WV Florida South: AL, AR, KY, LA, MS, OK, TN, TX Mountain: AZ, CO, ID, NV, NM, UT West: CA, OR, WA Rather than refitting the population model using these regions, state-level population counts from the cross-survey estimates for 2010 were summed to obtain estimated population counts for each of these nine regions to be used in raking. The key issue is whether the use of these nine regions with metropolitan status within each reduces bias without increasing variance to a greater degree than the use of the eight state-metro 39 clusters. Plotting overall MSE by variance across the demographic variables included in the raking , performs similarly as raking to the 8 state-metro clusters (see Figure B8). 0.07 0.06 0.05 0.04 Variance 0.03 MSE 0.02 0.01 0 Unweighted Sampling Weight Raked State-Met Clusters Raked 9 Regions FIGURE B8: COMPARISON OF MSE AND VARIANCE OF UNWEIGHTED ESTIMATES, KN SAMPLING WEIGHT, 8 STATE CLUSTER RAKING, AND 9 REGION RAKING. Post-Stratification 3: Full Post-Stratification In many contexts reliable population estimates for all possible combinations of the demographic variables involved in post-stratification simply are not available. This is one major motive for raking approaches to post-stratification. Results from the crosssurvey population model, however, can be used as this external source of data. Use of these data to do full post-stratification was compared to the raking approaches above. Although the cross-survey model yields estimates of demographic groups at the statelevel, there are simply too few cases in the KN survey sample to be able to estimate distributions within states. Thus, we compared fully post-stratified weighting estimates to the raked estimates based on the cross-survey distributions of demographic groups by the eight state-metropolitan clusters. Table B18 displays a comparison of the distributions of the weights. The fully poststratified weight has a standard deviation of 1.08 compared to 1.30 and 1.25 for raked 40 weights. Weight ratios (largest relative to smallest weight) are over 100 for all of the weights and are highest for full post-stratification. Although variance is reduced slightly under full post-stratification (see Figure B9), the MSE increases compared to the raked weights, indicating an increase in bias compared to variance. In this context, full poststratification does not appear to be as useful as raking. TABLE B18: COMPARISON OF THE KN SAMPLING WEIGHT ESTIMATES AND POST-STRATIFICATION WEIGHT ESTIMATES. Mean KN Sampling Weight Std. Dev. Minimum Maximum Sum Weight Ratio .88 .50 .22 4.74 788.62 22.02 w/ Post-stratification raking to demographics and 8 state-metro clusters 1.00 1.30 .09 23.44 894.00 254.85 w/ Post-stratification raking to demographics and 9 regions 1.00 1.25 .11 18.75 894.00 174.73 w/ Full-postratification to demographics and 8 state-metro clusters 1.00 1.08 .03 15.42 894.00 494.83 41 0.07 0.06 0.05 0.04 Variance 0.03 MSE 0.02 0.01 0 Unweighted Sampling Weight Raked State- Raked 9 Regions Full Post-Strat 8 Met Clusters Clusters FIGURE B9: COMPARISON OF MSE AND VARIANCE OF UNWEIGHTED ESTIMATES, KN SAMPLING ESTIMATES, 8 STATE CLUSTER RAKING, 9 REGION RAKING, AND FULL POST-STRATIFICATION ESTIMATES. Post-Stratification 4: Next Steps Model-based Estimation Another alternative to weighting is to employ a model-based approach (cites) which would be better suited to estimation of smaller cells and would combine the strengths of the cross-survey model-based analysis method with the development of population weights within a single survey. There are several methods yet to be explored to implement this approach. The first is to employ calibration approaches to weighting rather than the straight raking or post-stratification methods. Calibration offers a modelassisted approach that allows one to find a balance between the constraints and limitations of the marginal approach of traditional raking and full cross-classification in post-stratification. Another approach is to employ a fully Bayesian model-based approach to estimation, including sampling variables in a Bayesian model similar to the cross-survey population model and post-stratifying to Jewish population distributions derived from the cross-survey model. A third model-based approach is to incorporate the KN sample directly into the larger cross-survey data frame. This would enable one to generate survey-specific estimates of small cells by capitalizing on the borrowing of strength across all of the surveys that provide data on similar cells as those in the KN 42 sample. In addition, for all of these alternatives whether or not to trim/restrict the weights to be closer in magnitude to the original sampling weights needs to be explored. Each of these alternatives will be explored and compared to traditional raking approaches in future work. Another issue that needs to be dealt with in post-stratification is the use of fewer race categories. At the national level an estimated 1.5% of Jews identify as Black nonHispanic and 3% Hispanic and there are significant differences in distributions of the population by the four categories of race. Within the KN sample, however, only 3 of the respondents were Black non-Hispanic, far less than 1% of the sample. Pooling these few cases with the “Other non-Hispanic” group or comparing simply White non-Hispanic to all others may reduce some of the extreme values for weights as they are currently defined. In addition, what geographic identifiers might be best for Jewish population estimate needs further investigation. Our initial cross-survey models were based to the state-level. It might be useful, however, to define geographic regions differently. Two alternatives were examined in post-stratification based on the limitations of the KN sample. Other alternatives should also be explored. In particular, both the 1990 NJPS and the 2000 NJPS included sampling at the metropolitan level. NJPS 2000, in particular, included stratification of metropolitan areas based on estimated Jewish population density within 40 major metropolitan areas. For the subset of surveys that provides detailed geographic identifiers, such stratification at the county level is feasible and could be replicated to obtain data-based rather than guesstimates of Jewish population density associated with sampled geographic regions. Further, knowing how representative the KN sample of Jews is to the larger US population of Jews on social factors other than basic demographics would be useful. Although the cross-survey data set provides little data with which to examine the representativeness of the sample in terms of its Jewish identity (e.g., knowledge of Hebrew, years of Jewish day school, etc.), there is a critical mass of surveys that would enable one to examine at least a metric of religious involvement, the frequency of attending services. Many surveys also assess political ideology and philosophy and it would be useful to examine how the KN sample compares to the larger US Jewish population on these measures as well. Additional Considerations For analysis of the SAJ 2010 survey two additional factors were taken into consideration. The first was the handling of the JBO sample, Jews who indicated there religious identification was none but who identified as Jewish in response to follow-up questions. The second was the trimming of weights to reduce over influential cases of extreme low and high weights. 43 Handling of JBO Sample The JBO sample was most similar to the JBRI group who identified as Secular/Cultural/Just Jewish/No Religion in response to current denomination. These two groups were compared on variables included in the post-stratification model. This was done by regressing the dichotomous indicator (0=JBO, 1=JBRI Sec/Cul/JJ/NR) on age, race, education, and state-metropolitan cluster. Comparisons were made weighting each group by the base sampling weight, prior to making post-stratification adjustments. There were a few marginally significant differences for age, education and statemetropolitan clusters (.05 < p < .01). None of these differences would be significant if one controlled for the number of tests done with a Bonferroni or other adjustment. TABLE B19: LOGISTIC REGRESSION JBO VS. JBRI SECULAR ON AGE, RACE, EDUCATION, STATE AND METROPOLITAN AREA. Coef. Std. Err. t P>t 1.675 0.458 3.66 0 18-24 years -1.088 0.890 -1.22 0.222 25-34 years -1.141 0.842 -1.36 0.175 35-44 years -1.732 0.802 -2.16 0.031 45-54 years -1.586 0.663 -2.39 0.017 55-64 years -0.703 0.567 -1.24 0.215 Non-college -0.887 0.475 -1.87 0.062 18-24 years 1.125 1.352 0.83 0.406 25-34 years 0.655 0.925 0.71 0.479 35-44 years 1.352 0.867 1.56 0.119 45-54 years 1.731 0.746 2.32 0.021 55-64 years 0.717 0.665 1.08 0.281 Cluster 1 0.707 0.330 2.14 0.033 Cluster 2 0.447 0.535 0.84 0.404 Cluster 3 0.211 0.450 0.47 0.639 Cluster 4 0.220 0.459 0.48 0.632 Cluster 6 0.864 0.413 2.09 0.037 Cluster 7 -0.137 0.295 -0.46 0.644 Cluster 8 0.138 0.371 0.37 0.709 Intercept To reflect the greater proportion of the full Jewish population who identify as Secular/Cultural/Just Jewish/No Denomination or Religion, this sample was treated as other Sec/Cul/JJ/NR respondents and pooled with the full sample for analysis. Poststratification adjustments, raked to demographics and state group clusters as described above, were done across the full sample of 1,079 to create a full sample post-stratification weight. The weight was normalized to the sample size. Note that these analyses weights are not useful for purposes of total population estimation. They can, however, be used to 44 describe relative differences within the sample between secular and non-secularly identified Jews. Trimming To reduce the influence of extremely high or low weights, weights were trimmed such that the final weight (wif) was no more than twice and no less than half the base weight (wib), that is .5wib ≤ wif ≤ 2wib. Trimming was done using the TrimWeight function in the R Survey package (Lumley, 2011). Trimmed weights ranged from .11 to 9.5. Weight summary A summary of estimates of outcomes for different weighting scenarios are given in Table B20. 45 TABLE B20: COMPARISON OF ESTIMATES OF SURVEY OUTCOMES FOR DIFFERENT WEIGHTING SCENARIOS. Unweighted N Prop. w/ Post-stratification untrimmed Sampling Weight SE Prop SE Deff Prop SE w/ Post-stratification trimmed Deff Prop SE Deff Current Denomination Sec/Cul/JJ/None 653 0.605 0.015 0.618 0.017 1.32 0.635 0.023 2.43 0.630 0.021 2.14 Reform 218 0.202 0.012 0.190 0.013 1.22 0.172 0.016 2.02 0.175 0.016 1.92 Conservative 152 0.141 0.011 0.139 0.012 1.40 0.134 0.015 2.19 0.136 0.015 2.11 Orthodox 41 0.038 0.006 0.039 0.007 1.33 0.051 0.013 3.47 0.051 0.012 3.00 Reconstructionist 15 0.014 0.004 0.013 0.004 1.13 0.008 0.002 0.74 0.008 0.002 0.75 0 years 321 0.315 0.015 0.316 0.017 1.36 0.343 0.023 2.47 0.347 0.022 2.17 1-6 years 431 0.423 0.015 0.422 0.018 1.33 0.364 0.022 2.15 0.370 0.021 1.88 7 + years 268 0.263 0.014 0.261 0.016 1.38 0.293 0.027 3.60 0.283 0.022 2.52 962 0.935 0.008 0.937 0.008 1.22 0.925 0.011 1.92 0.924 0.011 1.89 67 0.065 0.008 0.063 0.008 1.22 0.075 0.011 1.92 0.076 0.011 1.89 0 years 595 0.656 0.016 0.670 0.018 1.36 0.634 0.029 3.28 0.646 0.024 2.31 1-3 years 199 0.219 0.014 0.209 0.016 1.34 0.246 0.030 4.26 0.232 0.022 2.55 4 + years 113 0.125 0.011 0.122 0.013 1.43 0.120 0.016 2.32 0.122 0.016 2.24 None 455 0.424 0.015 0.439 0.018 1.37 0.437 0.024 2.58 0.441 0.022 2.18 Read letters 344 0.321 0.014 0.319 0.016 1.31 0.333 0.025 3.07 0.325 0.021 2.22 Some of what read 224 0.209 0.012 0.201 0.014 1.37 0.182 0.017 2.19 0.186 0.017 2.08 50 0.047 0.006 0.041 0.006 1.04 0.048 0.008 1.70 0.048 0.008 1.68 Education Supplementary School Day School 0 years 1 + years Camping Hebrew Comprehension Most or all of what read Bar/Bat Mitvah 46 Yes 483 0.530 0.017 0.541 0.019 1.35 0.571 0.026 2.46 0.564 0.023 2.02 345 0.320 0.014 0.320 0.017 1.38 0.315 0.022 2.51 0.319 0.021 2.25 234 0.218 0.013 0.221 0.015 1.35 0.206 0.020 2.56 0.208 0.019 2.34 Brit 167 0.156 0.011 0.149 0.012 1.28 0.134 0.014 1.68 0.137 0.013 1.62 Bar Mitzvah 361 0.338 0.014 0.322 0.016 1.28 0.292 0.020 2.05 0.298 0.019 1.86 Wedding 245 0.229 0.013 0.208 0.014 1.20 0.196 0.016 1.83 0.200 0.016 1.72 Shiva call 385 0.361 0.015 0.353 0.017 1.35 0.324 0.021 2.20 0.330 0.020 1.97 360 0.572 0.02 0.560 0.023 1.31 0.575 0.025 1.65 0.575 0.025 1.62 Synagogue Membership Yes Any Synagogue Honors Yes Past Year Attend In-Marriage Jewish Spouse 47