Ordinal Logit producing NaN values

I am studying topics on preferences for redistribution. When I run an ordinal logit model like this:

polr(as.factor(Preferences_for_redistribution) ~ income + age + age_squared + other controls + as.factor(year), data = dat, Hess = TRUE)

all other variables are fine but the age or sometime age_squared variables are producing NaN values. Why might this be? I've seen tons of other papers that use ordinal logits with models like above.