Behavioral Responses to Risk: Evidence from Personal Ads

• In Event: Responses to Risk and Marginalization among LGBTQ+ Populations

Abstract

Estimating behavioral responses to changes in the cost of engaging in risky behavior is challenging because risky behaviors are private. I explore behavioral responses to risk in the context of the early years of the HIV/AIDS epidemic by digitizing over 170,000 men's personal ads posted in the oldest and largest national LGBT publication in the United States from 1975 to 1992. The vast majority of these ads represent men seeking other men for sexual or romantic relationships. Personal ads hold particular relevance for sexual minorities, who had limited opportunities to meet potential partners in traditional public spaces. I use Optical Character Recognition (OCR) and machine learning to digitize and extract relevant information from these ads about behavioral responses to HIV/AIDS in the early years of the epidemic.

Since HIV infections produce only minor symptoms in their early stages and AIDS is diagnosed at later stages, AIDS cases are exogenous to current behavioral patterns. This meant that a city's first AIDS case was unrelated to current patterns of sexual behavior. This study exploits variation in the timing of a city's first reported AIDS case to estimate the causal effect of an increase in STI risk on behavior.

I show that the first reported case marked a probable salient increase in the perceived cost of risky sexual behaviors by analyzing local newspaper reports. I construct a database of HIV/AIDS news reports published in leading local newspapers using the ProQuest Historical Newspaper database and Text and Data Mining (TDM) Studio. The first reported case of AIDS is associated with discontinuous increase in local newspaper reporting on HIV/AIDS. 

Thereafter, I use an event study difference-in-differences design to estimate the effect of the first AIDS case on personal ads. While the first AIDS case appears to have little impact on the total number of ads posted, I find significant evidence that individuals respond to the first reported AIDS case by expressing a preference for safe-sex. I find no significant evidence that the first AIDS case affected the number of personal ads seeking long-term relationships, explicitly sexual ads, or ads involving specific high-risk sexual activities, such as being the receptive anal partner. Taken together, this suggests that despite the increasing costs of risky sex, individuals continued to use the ads to seek out similar kinds of partnerships and encounters, but with a greater emphasis on safety measures to reduce the risk of HIV/AIDS transmission. 

Overall, my results provide novel evidence of behavioral responses to risk during the earliest years of the HIV/AIDS epidemic. They highlight behaviors responsive to changes in perceived risk, such as adopting safe-sex language, and behaviors unresponsive to such changes, such as overall ad usage. Additionally, this paper illustrates how novel sources of data, such as personal ads, can yield meaningful insights about minority populations typically underrepresented in conventional data sources like surveys.

Author

• Hasan Shahid, University of Maryland, Baltimore County
  Presenting Author