Does Competition Eliminate Discrimination. Evidence from the Commercial Sex Market in Singapore (Appendix)

Appendix A: Survey Design and Data Collection Before Collection Survey Design The early versions of the questionnaire were drafted after conversations and interviews with sex workers and industry insiders. Face-to-face conversations with insiders from di¤erent networks enabled us to verify the information they provided. We then pre-tested the questionnaire by having each enumerator interview his/her close contacts, and brief us on the interviewees’responses and how the interviews were actually conducted. We then reworded the questions and modi…ed the set of responses. We also coached enumerators to rephrase questions to help the sex workers understand, when necessary. The …nal survey is composed of two parts. Part I is designed to collect sex workers’ individual information and Part II is designed to collect detailed information regarding each sex worker’s recent business transactions. The …rst three sections of Part I gather sex workers’personal characteristics (e.g. ethnicity), work-related information (e.g. earnings) as well as information about the nature of their clients including questions designed to capture their attitudes and preferences with respect to client ethnicity. For example, they are asked to rank the factors they use to determine whether they will approach a client and to price the service. They are asked how much they like/dislike clients of di¤erent ethnicities and why. The fourth section in Part I asks enumerators to rate the interviewees with respect to beauty, …gure, English skills, etc., in the eyes of a potential client. Rating the sex workers in the eyes of the clients is a common practice widely adopted in surveys with sex workers. In Part II, we ask the sex workers a series of questions regarding their most recent 4 to 7 transactions. These questions are designed to obtain information on client’s characteristics, sex worker’s perception of client’s wealth and other transaction-speci…c information. For client’s characteristics, we ask about such items as client’s age, ethnicity, whether he is a tourist, whether he is a regular customer, his looks and his dress. The transaction-speci…c information includes which party made the …rst move, who initiated the price, what the asking price and contract price of the transaction were, whether the price o¤er was rejected, venue, type and duration of the service. Both English and Chinese versions of the survey forms were prepared for use. Enumerators and Administration of the Survey Four enumerators were hired to conduct the surveys. One was a boyfriend of a female pimp in Orchard Tower and has frequented Geylang as a client for over 10 years. The second enumerator is female, a former Geylang Gang member and worked in a non-sex worker capacity in a night club. The third enumerator was a close friend of Geylang pimps, and the fourth, who did only a small number of interviews, was a pimp. Each was personally involved in the sex market and had in-depth knowledge of this market. During the qualitative stage of the project, we were able to conduct phone and personal interviews with sex workers and pimps via their connections. They also introduced regular patrons to us which allowed us to learn about client’s preferences. This is where we learned that clients see sex workers in Geylang more or less substitutable. Our knowledge on the institutional background of Geylang street was acquired through these interactions with insiders. Each enumerator was trained independently. The training started with an overview of our project, but we also tested them on their knowledge of the milieu. We then went through each question with them to explain why we asked those questions, how they should explain the questions to the interviewees if necessary and the recording format we desired. We encouraged them to write notes on the survey if the sex worker o¤ered additional feedback. This training was essential because a) the subject matter is sensitive and b) many of the sex workers speak neither English nor Mandarin well. Enumerators had to be free to rephrase questions, change the order of questions, and even simply skip questions if a sex worker proved too reluctant to discuss some topic. When asked to rate Chinese clients on a predetermined scale, the sex worker provided comparisons with other ethnicities, the enumerator might jump to the comparison section. Thus although the questionnaire was structured, the interviews often more closely resembled guided rather than fully structured interviews. Data Collection It is not feasible to fully randomize the sample in this market. Due to its underground nature, we do not have complete information on the composition of the population and exact geographic distribution of each type of sex worker. Geylang is one the key illegal segments of the market and sex workers from this venue represent a considerable proportion of the illegal sex worker population. Some of the sex workers from Geylang are controlled by pimps. Thus, they do not have autonomy to accept the interview invitation, and we needed to invite pimps for drinks or pass the gift vouchers to the pimps to obtain their consent for the interview. Due to social stigma and concerns about being exposed to the authorities, some sex workers are not open about their job details to outsiders, especially when their responses are being recorded. Not surprisingly, we sometimes faced refusals. Given the limits on our ability to randomize the sample, we sought to obtain a good mixture of di¤erent types of sex workers by visiting di¤erent lanes, and visiting them at di¤erent times of the night and di¤erent days of the week. During each trip, enumerators interviewed whoever was available at the time and made arrangements with others to talk during their break or after work. Di¤erent streets feature sex workers from di¤erent countries, and sex workers from di¤erent countries sell their services at di¤erent prices. Popular sex workers may start work later than the less popular ones and may be more likely to take a weekday o¤ from the street because they are pre-booked for the whole day or they have earned substantial pro…t on other days. Our using of insiders, tipping and random visits ensures a fair mixture of the sex workers with di¤erent characteristics. As a result, we claim our data collection practice grants us a moderately randomized sample but make no claim that it is fully random. We recognize that the enumerators have a higher chance of interviewing less attractive sex workers, as they have a longer unemployment period during a shift and are more tempted by the gift voucher we o¤er. We found subjects who have longer experience are more open to our interview request. We are often asked whether the sex workers were controlled by their pimps through violence or other threats. We did not ask this question, but have little doubt that such sex workers, whom we know from the qualitative research exist, are underrepresented in our sample. In this sense, even if we believed that the sex workers would answer such questions honestly, our sample would not be ideal for addressing this issue. However, there is no reason to believe that our sample overor under-represents sex workers who engage in price discrimination.