Purpose: To assess the use of capture-recapture methods for estimating the size of sex worker populations in sub-Saharan Africa.

Methods: We used a capture-mark-recapture method to estimate the size of the bar-based female sex worker population in Bulawayo, Zimbabwe and compared this estimate with an estimate obtained by counting sex workers.

Results: Enumerators counted 6,997 women entering 56 bars known for sex worker activity. For the capture-recapture estimate, we interviewed 1,381 sex workers at 15 bars one Saturday night and 1,469 sex workers at the same bars a week later. Of these 1,469 women, 521 reported being interviewed the previous Saturday. The capture-recapture estimate of 3,894 (1381 x 1469 / 521) was considerably lower than the number counted. When we assumed that half the women returned to the same bar (rather than randomly mix among bars) and based the estimate on bars where the proportion recaptured was more than 20 percent, the estimate (7,855) of the 56-bar population was closer to the estimate obtained by counting.

Conclusions: Estimating the size of populations at risk for transmitting HIV is critical for AIDS prevention. The capture-recapture method may prove useful but requires collecting data to assess the direction and extent of bias in estimates.

Key Words: AIDS, population size, population surveys, epidemiological methods, prostitution, Africa, sampling