Recombination has been conjectured to play an important role for the evolution of drug resistance in HIV. In theoretical models it has been shown that recombination can have both an accelerating and a decelerating effect on the evolution of drug-resistance. Whether the models predict an accelerating or decelerating effect, depends on the specific assumptions. In particular, previous models have shown that both stochastic and population dynamic effects are likely to affect the expected impact of recombination. Here, we investigate the effect of recombination in a model that combines stochasticity with population dynamics. This approach allows assessing the effect of recombination not only on the time to resistance but also on the probability that the virus evolves resistance at all. Previous models had to neglect this latter aspect, because they assumed by construction that the virus always adapts successfully. We find that recombination hardly affects the probability that resistance evolves at all. Furthermore, our results show that recombination decelerates in most cases the emergence of drug resistance. These results contrast previous population genetic studies, which argued that, in a stochastic regime, recombination is expected to accelerate the evolution of drug resistance. The results also highlight that population dynamic effects can strongly influence the population genetics of recombination.