Here are results for heteroscadasticy tests of the models in the paper. The test is Breusch-Pagan test of heteroscedasticity. The null hypothesis is that the regression is homoscadastic and the results show whether the null has to be rejected. The results below show that none of the test statistics are statistically significant, not even close. Thus the conclusion is that there are no heteroscedasticity problems in the models.
Model 1a: chi-squared = 11.2759, with 9 df, p=0.26 Model 1b: chi-squared = 13.6045, with 15 df, p=0.56 Model 2a: chi-squared = 8.1058, with 9 df, p=0.52 Model 2b: chi-squared = 11.5762, with 14 df, p=0.64 Model 3a: chi-squared = 9.0840, with 10 df, p=0.52 Model 3b: chi-squared = 13.3010, with 15 df, p=0.58