The abortion level in Russia has been exceptionally high for several decades, yet during the last ten years it experienced a considerable decline. There is a concern that this favourable change could be largely due to a deterioration of statistical registration in the 1990s. In this paper, we use three reproductive and health surveys for a crosscheck with provider statistics, and analyse patterns and determinants of abortions. Each survey includes questions about the history of abortions. Our data indicate that survey estimates of the crude, total, and age-specific abortion rates emerge to be very close to respective figures from provider statistics for about two years preceding each survey. Survey estimates progressively deviate downwards from provider statistics when moving further back in time due to growing underreporting. This finding suggests that provider statistics on abortion in Russia are a true reflection of the situation they monitor, that the observed declining trend in abortion is a real one, and that analyses of survey data on abortions are justified for up to two years preceding the survey. Logistic regression using the data from the survey carried out in 2000 for the period of 1998–2000 reveals that the odds for an induced abortion are lower in case of a higher educational degree, that the odds increase with the number of children and decrease with the use of more reliable contraceptive methods, and that married women are more likely to have an abortion than never-married ones but less likely when compared to cohabiting women.