I study country-level predictors of the electoral performance of radical right-wing populist (RRWP) parties in 29 European states over a twenty-year period (1990-2010). The principal innovation of this study is the use of a two-part random effects model instead of a conventional fixed-effects Tobit regression to address the problem of excessive zeros in the electoral records of RRWP parties. The Tobit regression treats zeros as indicators for the unobserved part of the distribution of the dependent variable but not as actual values. The two-part regression allows for modelling zeros as actual observations, because it assumes that the data are generated by two distinct processes. The first process accounts for the probability of non-zero outcomes (i.e., that a RRWP party exist in a given country in a given year) and the other process determines actual share of votes for the party if it does actually exist. These two processes are modeled separately with using of a logistic regression in the first part of the model and a generalized gamma regression in the second part. The two-part model also allows for the cross-part correlation between random effects which gives it additional advantages over the Tobit model. Using this statistic technique, novel for the field, I tested my own hypothesis on the effect of value polarization on the electoral support for the RRWP parties, as well as revised some crucial findings of the previous research. I found a particular evidence of the positive association between country-level value polarization and far-right voting in the second part of the model. The other main results suggest that the higher actual electoral outcomes of RRWP parties can be attributed to the higher number of non-EU immigrants and the higher effective number of political parties in a country, whereas the probability of non-zero share of votes for RRWP parties is positively associated with the amount of social security funds and negatively associated with the unemployment rate. Finally, robustness checks demonstrate that the substantive results of the modelling of far-right voting are sensible to the selection of RRWP parties, but not for the imputation of missing data. Thus, selection issues should be addressed more closely in the further studies of far-right electoral successes.