Abstract
Understanding the determinants of prostitution is key to regulate it. This paper studies the hypothesis that marriage conditions are an economic determinant of female prostitution. I exploit differences in the timing of entry into force of divorce laws across U.S. states to explore the effect of such laws on arrests of female prostitutes. Using a difference-in-difference design, I find that unilateral divorce leads to a reduction of female arrested prostitutes between 5–10%. Results are consistent with the notion that improving marriage opportunities can be a powerful force to deter entry into prostitution for a subset of the population who is inframarginal. Lack of alternatives are key to explain the choice to conduct this activity. To this extent, this work is part of a broader research agenda hinting at improvements in gender equality as a mean for tapering off female prostitution.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
The two variables are bound to move together if the arrest intensity for prostitutes is fairly constant over time, an assumption that I cannot directly test but that I regard as plausible. Moreover, insofar as my identifying variation—introduction of unilateral divorce laws—does not covary with changes in the arrest intensity for prostitutes, my results are unaffected by changes to this intensity. In Section 7.2 I find no evidence supporting this hypothesis: unilateral divorce does not affect either the number of hired officers nor arrests for different sorts of crimes.
Both these checks are run in Appendix Section H.
In this sentence the word exit denotes either not entering the market or leaving it once they are already prostitutes. At first sight it might seem that Edlund and Korn (2002) would only predict the former since, according to their model, prostitutes compromise their marriage prospects because of their activity. However, this does not need to be the case. Edlund and Korn (2002) assert that entering prostitution reduces a woman’s ability to marry and that prostitutes earn high wages as a compensation for forgone marriage opportunities. Yet, they do not model how such prospects would worsen once a prostitute leaves the market. As a consequence, it might be that if unilateral divorce improves wives’ welfare, some prostitutes may prefer to leave the market to recover at least part of their marriage prospects.
See Section 9.1 in the Appendix for references and further information on this discussion.
This claim, as the authors write, “rests on the assumption that men prefer their wives to be faithful (for instance, from a desire to raise biological children)”.
For further information on the introduction of unilateral divorce across U.S. states, Appendix Section 9.3 discusses the legislative context that led to the enactment of such laws.
Assuming that a husband’s earnings are higher than his wife’s, under a mutual consent divorce regime, if a husband wished to divorce, he could “bribe” his wife. However, a wife could not afford to do so. Under unilateral divorce, a husband could still compensate his wife financially to avoid divorce. However, the wife would need to consent.
A substantially different question is whether this mechanism occurs because prostitutes in a certain age group exit prostitution (i.e., a stock effect) or because “potential” prostitutes, in a younger age group, prefer not to enter prostitution (i.e., an inflow effect). I investigate this issue in Appendix Section D.
In Appendix Section 9.5 I provide a complete list of offenses recorded in this database.
Note that using such data at the agency level does not affect the results.
There can be a lag of at most one year between the enactment date and the effective date. Furthermore, the effective date might be postponed, rendering the enactment date even less important. For further details about using effective dates instead of enactment dates, see Vlosky and Monroe (2002). It is important to use an objective criterion to classify these laws since it could impact my identification assumption and findings, although in this setting, intuitively, it does not appear plausible that the effect is immediate; thus, using either of the two dates should not considerably affect the results.
Coding these laws involves the problems discussed in Appendix Section 9.3.
Appendix Section 9.7 presents further information about the classification followed to code unilateral divorce laws across U.S. states.
The year 1972 is missing, although there is no reason to believe it is missing due to any special pattern of hired officers.
Namely, the survey inquires into topics respondents were “least proud of”.
The survey question is as follows: “What are the things you are least proud of as an American”? The answer connected to prostitution states: “Immorality in general; low morals; deterioration in moral standards; also specific actions--e.g., drinking, gambling, overexposure; lewdness in behavior or in mass media or literature; pornography, prostitution”.
The CPS data used in this paper are drawn from the Uniform Extracts of the CPS ORG. Center for Economic and Policy Research. 2017. CPS ORG Uniform Extracts, Version 2.2.1. Washington, DC.
Currently, the only state in the U.S. that has legalized prostitution is Nevada. Nevada introduced unilateral divorce laws and legalized prostitution in different years: unilateral divorce law became effective in 1967, while prostitution was legalized in 1971.
Since this paper finds that unilateral divorce decreases prostitution by improving prostitutes’ outside options, a possible concern could be that the entry into force of unilateral divorce could cause prostitutes from surrounding states to move to that state to exit prostitution. However, I did not find any evidence supporting this hypothesis.
Arrests of female prostitutes per 1,000,000 inhabitants are computed as the number of arrested female prostitutes divided by population and multiplied by 1,000,000. The same computations are made for data on other crimes in the rest of this paper.
This issue is further addressed in Section 5.
These computations simply take into account the structure of my dependent variable to compare it to a standard log-level specification. Precisely, \(\frac{\partial \log \left(y\right)}{\partial x}=\frac{\partial \log \left(1+y\right)}{\partial x}\frac{\partial \log \left(y\right)}{\partial \log \left(1+y\right)}=\beta \frac{1+y}{y}\simeq \hat{\beta }\frac{1+\bar{y}}{\bar{y}}=-6.8 \% \frac{1+1.9}{1.9}=-10.4 \%\)
Such results might be compared to those of Table A.6 since they are also based on the county-year level regression.
Figure 9 presents the data used for human trafficking laws.
Precisely, \(\frac{\partial \log \left(y\right)}{\partial x}=\frac{\partial IHS\left(y\right)}{\partial x}\frac{\partial \log \left(y\right)}{\partial IHS\left(y\right)}=\beta \frac{\sqrt{1+{y}^{2}}}{y}\simeq \widehat{\beta }\frac{\sqrt{1+{\bar{y}}^{2}}}{\bar{y}}=8.1 \% \frac{\sqrt{1+{\left(1.9\right)}^{2}}}{1.9}=-9.2 \%\)
The mean of the dependent variable collapsed at county-year level is 22.5.
Exactly as the treatment variable (i.e., unilateral divorce law).
The point estimate is even slightly larger in absolute value than that from the main specification.
Appendix Section 9.13 explores the hypothesis that unilateral divorce laws might displace women to states where such laws are effective. As expected, I find no empirical evidence supporting this hypothesis.
In their model, there are equal numbers of women and men, and since marriage is monogamous, the number of single men and single women is the same.
An alternative mechanism, not supported by the literature, is that unilateral divorce increases women’s wages (not only wives’ wages). This increase could in turn decrease prostitution insofar as legal jobs become more attractive to women and deter them from prostitution. I also explored this hypothesis and found no evidence in its favor. The relevant tables are available upon request.
Note that considering the impact of unilateral divorce on the labor force participation of wives would be uninformative about this (i.e., the wives’ wage) mechanism. The labor force participation of wives might rise after the introduction of unilateral divorce due to an improvement in wives’ bargaining position within the household.
I computed the median age between 1980 and 2014 for women at first marriage from the U.S. Census Bureau. The median is 24.8 years, and the average is 24.5 years.
The relative size of the two samples is fairly balanced since approximately 60% of my sample falls within the marrying-fertile age range (Table A.3). Moreover, it is important to note that only having data on prostitutes’ prices would not be informative to assess the marriage compensation mechanism. A potential threat to this approach is that since according to Edlund, Engelberg, and Parsons (2009), prostitutes’ prices are higher for women between 21 and 40 years old, if unilateral divorce law decreases the number of prostitutes of marrying-fertile age due to a rise in pm, I might find an ebb in average prostitutes’ prices simply because some of the prostitutes with the highest prices are exiting the market.
The p-values are available upon request.
In addition, Appendix Section I.2 replicates this analysis for indoor prostitution. The results do not change. It could be argued that the model developed in Edlund and Korn (2002) is better suited to indoor prostitution than street prostitution. Thus, finding empirical evidence in favor of the same mechanism for indoor prostitution is reassuring.
Appendix Section 9.9.3 explores the correlation between three free time activities drawn for the ATUS database and unilateral divorce. This section suggests that unilateral divorce is positively associated to increments in wives’ free time activities, supporting the marriage compensation mechanism.
A possible explanation is that states where unilateral divorce law becomes effective also reduce police budgets. Note that this event would threaten my identification assumption if it occurs contemporaneously with the entry into force of unilateral divorce.
There are alternative potential mechanisms involving police officers to explain the findings of the paper. For instance, it could be that, contemporaneously with the introduction of unilateral divorce in a certain state, police officers become less strict in arresting criminals or decrease their working hours. Even if implausible, these mechanisms would be able to explain the findings of this paper.
Appendix Section 9.11.1 presents the results of the same analysis using the yearly change in (i.e., first difference of) the number of officers per 1000 inhabitants and the growth rate of the number of officers per 1000 inhabitants as the dependent variables. Again, I find no evidence supporting this mechanism.
This regression analysis has two main features. First, it uses crimes committed only by women since unilateral divorce might change men’s behavior. Indeed, assuming that, on average, male incarceration decreases the likelihood that women marry (Charles and Luoh, 2010) and that, on average, women (i.e., wives) used to own less resources than men (i.e., husbands) implies that the introduction of unilateral divorce should decrease crimes committed by men by increasing wives’ bargaining power (w.r.t. mutual consent divorce). As a consequence, using crimes committed by men would be uninformative for studying the aforementioned mechanism. Second, this analysis makes use only of crimes not connected to prostitution since crimes related to prostitution (e.g., rape, sexual offenses, loitering, homicides) could be affected by unilateral divorce and not via a general decrease in arrests (Cunningham, DeAngelo & Tripp, 2017, HG.org, 2017, Urban Justice Center, 2005).
Therefore, there is no reason to believe that a lack of statistical significance in the regression results might be due to low precision of the estimates, as this was not the case for a much rarer crime such as prostitution. In addition, Appendix Section 9.11.2 presents results for each of the main categories of offenses recorded by UCR. The findings do not change.
Each estimate corresponds to the lower bound of a confidence interval. The sample mean of the dependent variable used in column (2) (i.e., average men’s real wage) is 13.2.
I estimate both regressions since it could be argued that the number of unmarried men does not vary substantially across months.
Note that this result does not contradict the marriage compensation mechanism since according to this mechanism, unilateral divorce improves wives’ welfare. First, the effect of unilateral divorce law on the marriage market is a composite effect depending on the effect of the law on other sub-populations (not only on prostitutes). Second, it might be that prostitutes do not enter or exit prostitution in the hope of being married but are not ultimately married.
The first “reverse sting” operation to catch prostitutes’ clients took place in Nashville, Tennessee, in 1964. Ten years later, considerable financial resources were devoted to arresting male customers in St. Petersburg, Florida, based on some of the main principles that were later used in the so-called “Nordic Model” (i.e., criminalizing the purchase of prostitution). In the same year, the first shaming campaign was started in Eugene, Oregon, in which names and/or photos of prostitutes’ clients were publicized. Similarly, in 1995, the first school to re-educate arrested sex buyers opened in San Francisco. The vast majority of these policies were intended to combat prostitution by reducing its demand.
The eight cities in the study were Denver, CO; Washington, DC; San Diego, CA; Miami, FL; Seattle, WA; Dallas, TX; Kansas City, MO; and Atlanta, GA.
For further details on the stratification of the prostitution market in the U.S., see Shively et al. (2012).
It is worth mentioning that this feature does not imply the authors pin down the sign of this relationship. Indeed, in their model with endogenous norms they conclude reputation has an ambiguous effect on the price of prostitution.
Age groups are classified according to UCR database as in Table A.3. Starting at 25 years old, ages are grouped into five-year blocks: 25 to 29 years old, 30 to 34 years old, and so on and so forth.
There could be the concern that there is no effect in the 17-24 age group since data are not pooled. However, Section I.1 presents the results of running a regression pooling together with arrests of female prostitutes between 17 and 24 years old and the results do not change.
Arrests of female prostitutes per 1,000,000 inhabitants is computed as the number of arrested female prostitutes divided by population and multiplied by 1,000,000. Same computations are made for data on other crimes.
There might be the concern that the parallel trend assumption does not hold since means differ for the two control groups and pre-treatment period of the treated. Yet, this has nothing to do with the parallel trend assumption. Comparing means of this sort is not tantamount to checking the plausibility of the parallel trend assumption. First, one should compare values of the outcome variable for both control groups and pre-treatment treated units in the same time period (i.e., pre-treatment period). Second, one should compare trends (i.e., at least two values). Third, since timing of the treatment varies across treated units one should take into account time fixed effects. As for the first and second points, parallel trend graphical checks are available upon request. While, to take into account the three issues the most convincing methodology to the best of my knowledge is an event study (such as Fig. 1).
Age groups are defined according to the UCR database.
Cunningham and Kendall (2011c) hypothesized that “the Internet and other modern technologies are drawing prime-aged (street) prostitutes into indoor work”. There could be the concern that this hypothesis is driving my findings. For this to occur, internet would need to be introduced simultaneously to unilateral divorce laws. Using data on indoor prostitutes would shed light on this mechanism too.
Appendix Section 9.10 provides the exact list of the occupational codes used.
An example of the SIC code classification is https://www.osha.gov/pls/imis/sic_manual.display?id=267&tab=description
All the categories are reported in Appendix Section 9.5.
Two crimes in Panel A could have been in Panel B. First, for “total drug abuse” (i.e., drugs crimes/use), there is evidence in the literature that both prostitutes and prostitutes’ clients make use of drugs. However, their relative percentage with respect to the whole “drugs market” is unclear. This is why such regressions’ results also appear in Section 7. Second, “vagrancy” there is evidence in the literature that prostitutes’ arrests are seldom reported as “loitering” (for example, the New York State Division of Criminal Justice Services classifies “loitering” as including “loitering for prostitution”). Given the close connection between “vagrancy” and “loitering”, the former could also be considered as an offense connected to prostitution.
Dank et al. (2014) also highlight the expansion of internet use to match clients and prostitutes.
Namely, “The Erotic Review”, “Erotic Review” (easier and faster version to search on Google), “Craigslist”, “Backpage” and “Backpage erotic”. I cannot consider “Craigslist erotic” since it was not searched in Google enough times (i.e., it was searched so rarely that Google does not index the number of searches).
This website has been used in the literature to collect data on prostitutes and customers (see, among others, Cunningham & Shah, 2017).
The sample size varies across columns since Google Trends data are available only for states where the number of searches is not close to zero. Searches for certain words were close to zero in some states. However, this was not the case for any treated state. A list of missing state/s for each word is available upon request.
This data set has been used by Edlund and Pande (2002) to show that, after being divorced, women are more likely to support left-wing parties.
Note that since the YPSS considers the marital status of respondents in wave w, if this were the case, it would bias my results.
This supports the hypothesis that it is the first divorce (separation) that changes males’ attitudes toward prostitution.
This mechanism seems unlikely because spouses can file a divorce in a different state from the one where they were married as long as one of the spouses meets the residency requirements of that state, so there would not be any incentive to move to a state to get married due to their unilateral divorce law.
References
Agan, A., & Starr, S. (2017). Ban the box, criminal records, and racial discrimination: A field experiment. The Quarterly Journal of Economics, 133, 191–235.
Agan, A. Y., & Makowsky M. D. (2018). The minimum wage, EITC, and criminal recidivism. Technical report, National Bureau of Economic Research.
Alesina, A. F., & Giuliano, P. (2007). Divorce, fertility and the value of marriage. (Mimeo).
Allard, P., & Herbon, G. (2003). Prostitution enforcement in Chicago: An examination of costs and alternatives. Washington, DC: The Sentencing Project.
Arunachalam, R., & Shah, M. (2008). Prostitutes and brides? The American Economic Review, 98, 516–522.
Beauchamp, A., & Chan, S. (2014). The minimum wage and crime. The BE Journal of Economic Analysis & Policy, 14, 1213–1235.
Becker, G. S. (1962). Irrational behavior and economic theory. Journal of political economy, 70, 1–13.
Becker, G. S. (1968). Crime and punishment: An economic approach. In The economic dimensions of crime, pp. 13–68. Springer.
Becker, G. S. (1981). Altruism in the family and selfishness in the market place. Economica, 48, 1–15.
Bisschop, P., Kastoryano, S. & van der Klaauw, B. (2017). Street prostitution zones and crime. American Economic Journal: Economic Policy.
Borusyak, K., & Jaravel, X. (2017). Revisiting event study designs. Available at SSRN 2826228.
Cáceres-Delpiano, J., & Giolito, E. (2012). The impact of unilateral divorce on crime. Journal of Labor Economics, 30, 215–248.
Charles, K. K., & Luoh, M. C. (2010). Male incarceration, the marriage market, and female outcomes. The Review of Economics and Statistics, 92, 614–627.
Cho, S.-Y., Dreher, A., & Neumayer, E. (2013). Does legalized prostitution increase human trafficking? World Development, 41, 67–82.
Ciacci, R., & Sviatschi, M. M. (2021, 07). The Effect of Adult Entertainment Establishments on Sex Crime: Evidence from New York City. The Economic Journal. ueab042.
Cook, P. J., Kang, S., Braga, A. A., & Ludwig, J. (2015). An experimental evaluation of a comprehensive employment-oriented prisoner re-entry program. Journal of Quantitative Criminology, 31, 355–382.
Cunningham, S., DeAngelo, G., & Tripp, J. (2017). Craigslist’s Effect on Violence Against Women.
Cunningham, S., & Kendall, T. D. (2010). Risk behaviours among internet-facilitated sex workers: evidence from two new datasets. Sexually Transmitted Infections, 86, iii100–iii105.
Cunningham, S., & Kendall, T. D. (2011a). Men in transit and prostitution: Using political conventions as a natural experiment. The BE Journal of Economic Analysis & Policy11(1).
Cunningham, S., & Kendall, T. D. (2011b). Prostitution 2.0: The changing face of sex work. Journal of Urban Economics, 69, 273–287.
Cunningham, S., & Kendall, T. D. (2011c). Prostitution, technology, and the law: new data and directions. Research Handbook on the Economics of Family Law, 221.
Cunningham, S., & Kendall, T. D. (2013). Prostitution 3.0: A comment. Iowa L. Rev. Bull., 98, 131.
Cunningham, S., & Shah, M. (2017). Decriminalizing indoor prostitution: Implications for sexual violence and public health. The Review of Economic Studies, 85, 1683–1715.
Cunningham, S., & Shah, M. (2021). The economics of sex markets: Regulation, online technology and sex trafficking. Handbook of Labor, Human Resources and Population Economics (forthcoming).
Dank, M. L., Khan, B., Downey, P. M., Kotonias, C., Mayer, D., Owens, C., Pacifici, L. & Yu, L. (2014). Estimating the size and structure of the underground commercial sex economy in eight major US cities. Urban Institute Washington, DC.
De Chaisemartin, C., & D'Haultfœuille, X. (2020). Two-way fixed effects estimators with heterogeneous treatment effects. American Economic Review, 110, 2964–96.
Della Giusta, M. (2016). emanding sex: critical reflections on the regulation of prostitution. Routledge.
Della Giusta, M., Di Tommaso, M. L., & Strøm, S. (2009). Who is watching? the market for prostitution services. Journal of Population Economics, 22, 501–516.
Doleac, J. (2016). Increasing employment for individuals with criminal records. Policy Proposal, The Hamilton Project, Washington, DC.
Doleac, J. L., & Hansen, B. (2016). Does “ban the box” help or hurt low-skilled workers? statistical discrimination and employment outcomes when criminal histories are hidden. Technical report, National Bureau of Economic Research.
Edlund, L. (2013). The role of paternity presumption and custodial rights for understanding marriage patterns. Economica, 80, 650–669.
Edlund, L., Engelberg, J., & Parsons, C. A. (2009).The wages of sin.
Edlund, L., & Korn, E. (2002). A theory of prostitution. Journal of political Economy, 110, 181–214.
Edlund, L., & Pande, R. (2002). Why have women become left-wing? the political gender gap and the decline in marriage. The Quarterly Journal of Economics, 117, 917–961.
Elrod, L. D., & Spector, R. G. (1997). A review in the year of family law: A search for definitions and policy. Fam. LQ, 31, 613.
Farley, M. (2003). Prostitution, trafficking and traumatic stress, Volume 2. Psychology Press.
Farley, M. (2004a). Bad for the body, bad for the heart: prostitution harms women even if legalized or decriminalized. Violence against women, 10, 1087–1125.
Farley, M. (2004b). Prostitution is sexual violence. Psychiatric Times, 21, 7–8.
Farley, M., & Butler, E. (2012). Prostitution and trafficking-quick facts. Prostitution Research & Education.
Farley, M., Cotton, A., Lynne, J., Zumbeck, S., Spiwak, F., Reyes, M. E., Alvarez, D., & Sezgin, U. (2004). Prostitution and trafficking in nine countries: An update on violence and posttraumatic stress disorder. Journal of Trauma Practice, 2, 33–74.
Farrell, A., & Cronin, S. (2015). Policing prostitution in an era of human trafficking enforcement. Crime, Law and Social Change, 64, 211–228.
Fondation-Scelles. (2012). Rapport mondial sur l’exploitation sexuelle: la prostitution au coeur du crime organisé. Economica.
Friedberg, L. (1998). Did unilateral divorce raise divorce rates? evidence from panel data. Technical report, National Bureau of Economic Research.
Gertler, P., Shah, M., & Bertozzi, S. M. (2005). Risky business: the market for unprotected commercial sex. Journal of Political Economy, 113, 518–550.
Goodman-Bacon, A. (2021). Difference-in-differences with variation in treatment timing. Journal of Econometrics.
Gray, J. S. (1998). Divorce-law changes, household bargaining, and married women’s labor supply. The American Economic Review, 88, 628–642.
Grossbard, A. (1976). An economic analysis of polygyny: The case of Maiduguri. Current anthropology, 17, 701–707.
Grossbard, S. (1993). On the economics of marriage-a theory of marriage, labor and divorce.
Grossbard, S. (2005). Women’s labor supply, marriage, and welfare dependency. Labour, 19, 211–241.
Grossbard, S. (2014). The marriage motive: a price theory of marriage: how marriage markets affect employment, consumption, and savings. Springer.
Grossbard-Shechtman, A. (1982). A theory of marriage formality: the case of Guatemala. Economic Development and Cultural Change, 30, 813–830.
Grossbard-Shechtman, A. (1984). A theory of allocation of time in markets for labour and marriage. The Economic Journal, 94, 863–882.
Gruber, J. (2004). Is making divorce easier bad for children? the long-run implications of unilateral divorce. Journal of Labor Economics, 22, 799–833.
Hannagan, R. J., Pimlott, J. P., & Littvay, L. (2010). Does an Emily’s list endorsement predict electoral success, or does Emily pick the winners? PS: Political Science & Politics, 43, 503–508.
Heer, D. M., & Grossbard-Shechtman, A. (1981). The impact of the female marriage squeeze and the contraceptive revolution on sex roles and the women’s liberation movement in the united states, 1960 to 1975. Journal of Marriage and the Family, 49–65.
HG.org. (2017). Prostitution in the united states. https://www.hg.org/article.asp?id=30997. Accessed: 2017−08-30.
Immordino, G., & Russo, F. F. (2015). Laws and stigma: the case of prostitution. European Journal of Law and Economics, 40, 209–223.
Jakobsson, N., & Kotsadam, A. (2013). The law and economics of international sex slavery: prostitution laws and trafficking for sexual exploitation. European Journal of Law and Economics, 35, 87–107.
Kempadoo, K. (1999). Slavery or work? reconceptualizing third-world prostitution. positions, 7, 225–237.
Kempadoo, K. (2007). The war on human trafficking in the Caribbean. Race & Class, 49, 79–85.
Kempadoo, K., Sanghera, J. & Pattanaik, B. (2015). Trafficking and prostitution reconsidered: New perspectives on migration, sex work, and human rights. Routledge.
Lee, S., & Persson. P. (2015). Human trafficking and regulating prostitution.
Marvell, T. B. (1989). Divorce rates and the fault requirement. Law and Society Review, 543–567.
Mazur-Hart, S. F., & Berman, J. J. (1977). Changing from fault to no-fault divorce: An interrupted time series analysis. Journal of Applied Social Psychology, 7, 300–312.
Pearl, J. (1986). The highest paying customers: America’s cities and the costs of prostitution control. Hastings LJ, 38, 769.
Pimlott, J. P. (2010). Women and the democratic party: The evolution of Emily’s list. Cambria Press.
Polaris A. (2017). look back: Building a human trafficking legal framework. https://polarisproject.org/wp-content/uploads/2019/09/2014-Look-Back.pdf. Accessed: 2022-12−07.
Raphael, S. (2010). Improving employment prospects for former prison inmates: Challenges and policy. In Controlling crime: Strategies and tradeoffs, pp. 521–565. University of Chicago Press.
Raphael, S., & Weiman, D. F. (2007). The impact of local labor market conditions on the likelihood that parolees are returned to custody. Barriers to Reentry? The Labor Market for Released Prisoners in Post-Industrial America, 304–332.
Rasul, I. (2004). The impact of divorce laws on marriage. Unpublished manuscript, University of Chicago.
Rasul, I. (2005). Marriage markets and divorce laws. Journal of Law, Economics, and organization, 22, 30–69.
Rheinstein, M. (1955). The law of divorce and the problem of marriage stability. Vanderbilt Law Review, 9, 633.
Rheinstein, M. (1972). Marriage stability, divorce, and the law.
Schnepel, K. T. (2017). Good jobs and recidivism. The Economic Journal, 128, 447–469.
Shively, M., Kliorys, K., Wheeler, K., Hunt, D., et al. (2012). A national overview of prostitution and sex trafficking demand reduction efforts. Washington, DC: The National Institute of Justice.
Stevenson, B. (2008). Divorce law and women’s labor supply. Journal of Empirical Legal Studies, 5, 853–873.
Stevenson, B., & Wolfers, J. (2006). Bargaining in the shadow of the law: Divorce laws and family distress. The Quarterly Journal of Economics, 121, 267–288.
Stevenson, B., & Wolfers, J. (2007). Marriage and divorce: Changes and their driving forces. The Journal of Economic Perspectives, 21, 27–52.
Sun, L., & Abraham, S. (2020). Estimating dynamic treatment effects in event studies with heterogeneous treatment effects. Journal of Econometrics.
The Economist. (2004). Sex is their business. http://www.economist.com/node/3151258. Accessed: 2017-08-30.
Tuttle, C. (2019). Snapping back: Food stamp bans and criminal recidivism. American Economic Journal: Economic Policy, 11, 301–27.
Uggen, C., & Shannon, S. K. (2014). Productive addicts and harm reduction: How work reduces crime-but not drug use. Social Problems, 61, 105–130.
Urban Justice Center. (2005). Behind Closed Doors: An Analysis of Indoor Sex Work in NYC.
Vlosky, D. A., & Monroe, P. A. (2002). The effective dates of no-fault divorce laws in the 50 states. Family Relations, 51, 317–324.
Voena, A. (2015). Yours, mine, and ours: Do divorce laws affect the intertemporal behavior of married couples? The American Economic Review, 105, 2295–2332.
Weitzman, L. J. (1985). The divorce revolution: The unexpected social and economic consequences for women and children in America.
Wickelgren, A. L. (2007). Why divorce laws matter: Incentives for non-contractible marital investments under unilateral and consent divorce. The Journal of Law, Economics, & Organization, 25, 80–106.
Wolfers, J. (2006). Did unilateral divorce laws raise divorce rates? a reconciliation and new results. The American Economic Review, 96, 1802–1820.
Yang, C. S. (2017). Local labor markets and criminal recidivism. Journal of Public Economics, 147, 16–29.
Yttergren, Å, & Westerstrand, J. (2016). The Swedish legal approach to prostitution. trends and tendencies in the prostitution debate. NORA-Nordic Journal of Feminist and Gender Research, 24, 45–55.
Acknowledgements
I would like to thank Juan J. Dolado for his generous advice and support during the different stages of this project. I am also grateful to Andrea Ichino, Dominik Sachs and Francesco Fasani for valuable suggestions that helped improve this paper as well as to Inés Berniell, Raúl Sánchez de la Sierra, Gabriela Galassi and Gabriel Facchini for useful comments. Financial support from the Spanish Ministry of Science (PGC2018−093506-B-I00) is gratefully acknowledged.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author declares no competing interests.
Additional information
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Online Appendix
Online Appendix
1.1 Background on the U.S. prostitution market
Prostitution is one of the most unsafe occupations in the U.S., worse than being an Alaskan fisherman, logger, or oil rig worker. As reported by HG.org (2017), the death rate for prostitutes in the U.S. is 204 out of every 100,000; that for Alaskan fishermen is 129 out of every 100,000. Moreover, statistics on prostitutes are conservative since prostitution is illegal in the U.S. (it is only allowed in Nevada in brothels and certain areas of the state). Prostitutes facing violence have nowhere to go without risking arrest themselves. Since the 1960s, combating prostitution has been a key target of many American policy interventions (Shively, Kliorys, Wheeler & Hunt, 2012).Footnote 48
Dank et al. (2014) found that, in 2007, in eight major U.S. cities, prostitution generated a market value ranging from $39.9 to $290 million.Footnote 49 Furthermore, Pearl (1986) estimated that 16 U.S. cities spent on average $15.3 million each year on prostitution control. More recently, Allard and Herbon (2003) found that prostitution arrests caused an expense of $10.3 million in the city of Chicago alone. According to HG.org (2017), the annual average of approximately 70,000-80,000 arrests for prostitution costs American taxpayers $200 million. Unsurprisingly, prostitution moves huge amounts of money in the form of both generated income and crime prevention.
The large amounts of money that prostitution moves around might originate from the lack of agreement on prostitution law. Opponents of prostitution contend that prostitution is dehumanizing (e.g., Farley, 2003, 2004a, Farley & Butler, 2012, Farley et al., 2004). According to this line of thought, prostitutes are victims of physical and psychological violence. For example, Farley (2004b) estimated that approximately 85% to 95% of prostitutes wish to escape from prostitution but have no other options for survival. By contrast, those supporting legalization of prostitution argue that prostitutes chose to exchange their time and services for money as in any other job (e.g., Kempadoo,1999, 2007, Kempadoo, Sanghera & Pattanaik, 2015, TheEconomist, 2004). Hence, it is the criminalization of prostitution that worsens prostitutes’ standard of living. They claim that since prostitution cannot be stopped, legalizing it would be the only way to tax and “protect” prostitutes.
This ideological problem regarding how to regulate prostitution is all the more important because the U.S. prostitution market is highly stratified. Thus, the effects of any given regulation of the prostitution market might differ across market segments. The prostitution market in the U.S. can be divided into three segments. On the lowest tier, there are street prostitutes. Street prostitutes are usually controlled by pimps and thus make the least money. Further, they lack control over their choice of clients and are more likely to be victims of violence and to be arrested. Operating at the medium level are those working indoors in brothels, massage parlors, gentlemen’s clubs and strip clubs. They usually enjoy better conditions than street prostitutes. Finally, escorts comprise the highest level of prostitutes. In this market segment, prostitutes have control over their choice of clients and “careers”; usually, they are not controlled by a pimp, earn high wages and are less likely to be victims of violence. This group is the one that best fits the image of prostitutes depicted by supporters of legalized prostitution. Prostitution in the medium and high tiers of this stratification takes place indoors: this is why it is also known as indoor prostitution, while street prostitution is also known as outdoor prostitution.Footnote 50
This study makes use of data on female prostitution arrests, which are more likely to represent outdoor prostitution than indoor prostitution. However, I build a proxy variable for indoor prostitution when analyzing the mechanisms linking unilateral divorce and prostitution.
1.2 Literature review
This paper contributes to three different lines of research. First, the empirical findings of this paper complement scholarship on the determinants of prostitution and on the relevance of several mechanisms at play in economic models of prostitution (Cunningham & Shah, 2021).
In particular, the literature has analyzed what is known as the prostitution wage premium puzzle: prostitution is low-skilled, labor-intensive, and female dominated but well paid. Scholars have explained this puzzle with supply-side hypotheses. On the one hand, Gertler et al. (2005) document and explain why riskier sex acts are priced higher than less risky sex acts. This link might be viewed as a wage premium due to sale of unprotected sex. According to this hypothesis, prostitutes are willing to face the risk of contracting sexually transmitted infections since customers are willing to pay more to avoid using condoms. On the other hand, Della Giusta et al. (2009) claim that this wage premium can be explained by the low reputation that prostitution has and the social stigma it faces.Footnote 51 Finally, Edlund and Korn (2002) suggest that marriage compensation is key to understanding the prostitution wage premium puzzle: marriage market prospects are an important source of income for women, but by entering into prostitution, women compromise such prospects. The present paper tests this third hypothesis and finds evidence in its favor.Footnote 52 In addition, a strand of the literature has focused on analyzing how policy interventions connected to prostitution regulation affect other crimes. For example, Cho, Dreher, and Neumayer (2013), Jakobsson and Kotsadam (2013), Lee and Persson (2015) study the link between human trafficking and prostitution, while Bisschop, Kastoryano, and van der Klaauw (2017), Ciacci and Sviatschi (2021), Cunningham and Shah (2017) analyze how changes in prostitution policies or business establishments connected to prostitution affect sex crimes. However, to the best of my knowledge, this is the first paper that examines how a policy intervention outside the prostitution market affects the latter.
Second, this paper contributes to a stream of research in sociology, law and economics that evaluates the impact of unilateral divorce laws on various outcomes (see, e.g., Alesina and Giuliano, 2007, Edlund & Pande, 2002, Friedberg, 1998, Gray, 1998, Gruber, 2004, Rasul, 2004, 2005, Stevenson, 2008, Stevenson & Wolfers, 2006, 2007, Voena, 2015, Weitzman, 1985, Wickelgren, 2007). However, none on these papers addresses the effects of these laws on prostitution.
Finally, the results of this paper also contribute to a growing line of the literature in sociology, criminology and economics that studies the effect of changing the opportunity cost of criminals on crime (see, e.g., Agan & Starr, 2017, Agan & Makowsky, 2018, Beauchamp & Chan, 2014, Becker, 1962, 1968, Cook, Kang, Braga & Ludwig, 2015, Doleac, 2016, Doleac & Hansen, 2016, Raphael, 2010, Raphael & Weiman, 2007, Schnepel, 2017, Tuttle, 2019, Uggen & Shannon, 2014, Yang, 2017).
1.3 Legislative background: The Divorce Revolution
Traditionally, in the U.S., divorce was permitted only on grounds of demonstrating guilt of misconduct by one of the two spouses and had to be agreed on mutually by both spouses (i.e., consent of the innocent party was required before a divorce was granted). Generally, such grounds were abandonment, cruelty, incurable mental illness, or adultery. The law was regarded as inadequate due to the major emotional and financial transaction costs involved in the verification of guilt of wrongdoing during the divorce process.
Thus, dissolution of marriages that were broken for mundane reasons (i.e., without misconduct by any spouse) was only possible if one of the two parties declared herself or himself guilty. In addition, since divorce had to be mutually agreed, the belief was that whenever husbands wished to divorce, they would bribe their wives to obtain their consent, while if wives wished to divorce, they could not afford to bribe their partners.
However, since divorce was considered to be against the public interest, civil courts formerly denied a divorce if there was evidence of cooperation between the two spouses or if they attempted to counterfeit the grounds for divorce. In fact, divorce could be barred even if one of the two spouses was found guilty. The three main reasons for refusing a divorce petition were as follows: recrimination, the suing spouse also being found guilty; condonation, forgiving the misconduct explicitly or implicitly by continuing to live with the partner after knowing of it; and connivance, participating in the fault, such as organizing an act of adultery.
This law not only required marital wrongdoing to file the divorce petition but also punished spouses for such misbehavior. Indeed, both husband and wife could be punished if they were found guilty of wrongdoing. If the husband was at fault, he usually suffered the loss of child custody and the imposition of economic responsibilities; if the wife was found at fault, she might suffer the loss of alimony and child custody.
There was the tacit perception that the abolition of fault grounds and mutual consent would eliminate the hypocrisy that incited the use of perjury and the forgery of evidence to surmount strict legal hurdles (Marvell, 1989, Mazur-Hart & Berman, 1977, Rheinstein, 1955, 1972). On the one hand, the guilt or innocence of the spouses would be irrelevant if no-fault divorce were available. On the other hand, consent of the partner would be useless if unilateral divorce were available.
In 1969, the California Family Law Act completely removed the requirements of fault as the basis for divorce and allowed spouses to file for divorce without the consent of their partner. This act established only two grounds for divorce: (i) irreconcilable differences; (ii) incurable insanity. Following Weitzman (1985), researchers have viewed this reform as the basis for both no-fault and unilateral divorce.
The focus of the reform was gender neutral: it assumed that the divorcee was economically independent and employable. Consequently, this law established two major bases for alimony awards: the divorcees’ employability and the length of the marriage. If either of the divorcees were not economically independent, this law also helped her/him to garner new-skills or to improve existing skills to become self-sufficient.
The California Family Law Act started a movement to reform divorce laws in the U.S. known as “The Divorce Revolution”, and various states followed suit. The movement gathered an apolitical consensus. Right-wingers viewed it as an expansion of personal rights and freedom. Left-wingers promoted it to prevent women from being locked into unfortunate marriages.
Unlike the case of California, “The Divorce Revolution” consisted of two steps: no-fault divorce and unilateral divorce. First, states moved to no-fault divorce regimes, which were already effective (to different degrees) in various states prior to 1950 while retaining mutual agreement. Next, states moved to unilateral divorce, requiring the consent of only one spouse to legally dissolve the marriage. This second step, which was uncommon before the 1960s, started in 1969 immediately after the passage of the California Family Law Act.
No-fault divorce does not change the bargaining structure within a marriage relationship. It solely reduces transaction costs by decreasing bargaining costs and eliminating financial penalties that could no longer be inflicted on at-fault spouses. Indeed, a no-fault divorce law eliminates the requirement of proof of guilt or innocence of either spouse. After the introduction of no-fault divorce, marriage dissolution could be lodged on grounds such as “incompatibility” or “irreconcilable differences”. However, it has to be agreed to mutually by both partners. It was formulated simply to make marriage dissolution less dolorous and mournful.
Unilateral divorce goes a step further. It removes the property rights that mutual consent divorce grants either to the innocent spouse (for fault divorces) or to the spouse who does not wish to get divorced (for no-fault divorces). Namely, unilateral divorce could change spouses’ behavior in two different ways. First, it allows spouses who are unable to prove the guilt of their partner or cannot afford to bribe their partner to file for divorce. Second, it changes the bargaining power between the members of the couple.
Furthermore, no-fault divorces are more complex to code since the definition of what constitutes a no-fault divorce is much broader than the definition of unilateral divorce. The literature classifies no-fault divorce into four categories: (a) living separate and apart as grounds for divorce; (b) incompatibility as grounds for divorce; (c) no-fault provisions added to traditional grounds as grounds for divorce; (d) no-fault is the sole grounds for divorce (Elrod & Spector, 1997). These differences have given rise to widespread disagreement among scholars using no-fault divorce dates (Vlosky & Monroe, 2002). An important point of divergence has been how to categorize fault-based laws that added “living separate and apart” provisions as no-fault laws. Even if such settlements consent to divorce without any proof of wrongdoing, the waiting period might be so long that it renders the provision either too weak to be regarded as no-fault or tantamount to a fault divorce law. The key difference is that true no-fault divorce laws are difficult to compare to legislative changes that simply revise fault-based grounds.
Unilateral divorce laws are easier to code; the only difference is whether the provision requires a separation period. The literature has treated as unilateral divorce regimes either both provisions with and without separation requirements or only provisions without separation requirements. Following Gruber (2004), I use unilateral divorce laws without separation for two reasons. First, since I code the law using a dummy variable, the comparison of identical unilateral divorce laws seems more reasonable and accurate. Second, even if unilateral divorce laws without separation requirements usually became effective later than those with separation requirements, I observe when such laws enter into effect since my sample period spans from 1980 to 2014.
Finally, coding might differ on whether enactment dates or effective dates were used. The enactment date is the date on which a law is approved, while the effective date is the date on which a law enters into force. There can be a lag of months between the enactment and the effective date. Coding the effective date is usually more laborious than coding the enactment date since it necessitates a review of the session laws of each state. Nevertheless, I use the effective date since it is the one that is crucial in legal actions.
1.4 Nature of the effect: Inflow vs Stock
Figure 6 shows the effect of unilateral divorce on prostitution across age groups clustering variance at county level (dotted-dashed line) and state level (dashed line).Footnote 53. As in the main regression, the dependent variable is in logs, each regression includes county, year and month fixed effects and county-year trends.
Notes: This figure shows the effect of unilateral divorce on prostitution across age groups. Each coefficient and standard errors come from a regression, with the same structure as in the main specification, where the dependent variable was computed using the age group indicated. Confidence intervals are at the 90% level. These results suggest that unilateral divorce both prevents women from entering prostitution and affects women who are already prostitutes
There are two ways in which unilateral divorce could affect prostitution: either by preventing women from becoming prostitutes (i.e., inflow effect) or by affecting prostitutes who are already in the market (i.e., stock effect). If unilateral divorce decreases young (old) prostitutes’ arrests, it would support the former (latter) effect. Figure 6 shows that unilateral divorce mainly reduces prostitution in women between 25 and 29 years old and in those between 45 and 49 years old.Footnote 54 Hence, there is evidence in favor of both effects.
In addition, Fig. 6 has two features worth mentioning. First, unilateral divorce does not affect prostitutes aged 17 and 24 years old and prostitutes aged 50 and 65 years old or older. In these two age groups, the point estimate is close to zero, and it is reassuring to find that the standard errors are narrow. Second, in contrast, in the age group of women whose age falls between 25 and 49 years old, there seems to be a U-shaped curve.
There are a couple of limitations of this analysis that are worth mentioning. First, I cannot address cases of prostitutes where the pimp is their husband since in the UCR dataset I have no information on whether prostitutes are single or married. Clearly, these cases could get divorced also under a mutual consent regime, but it would be easier to do so under unilateral divorce. Second, it would be interesting to explore whether unilateral divorce redistributed prostitution towards segments where the bargaining power of the prostitute is higher with respect to that of the customer (e.g., online sex markets). However, I have no data on online prostitution either and it might seem plausible to suspect that the measures used to proxy prostitution in this paper –arrests and indoor prostitution via occupational codes– are not representative of this specific segment of the market. Thus, I cannot offer any empirical evidence on this matter.
1.5 List of crimes in UCR data set
Table A.1.
1.6 Further information on the data set
1.6.1 Descriptive statistics
Table A.2 displays summary statistics for arrests of female prostitutes per 1,000,0000 inhabitants across treated and control states.Footnote 55 Data are at the county-month level, and treated states are disaggregated at pre- and post-treatment levels.Footnote 56
Table A.3 shows summary statistics for arrests of female prostitutes per 1,000,0000 inhabitants broken out by age group. Columns (1) to (4) respectively report mean, standard deviation, minimum and maximum. Column (5) reports the share of each group, out of the total arrests of female prostitutes, without taking into account the populationFootnote 57
Figure 7 displays arrests of female prostitutes per 1,000,0000 inhabitants (in the same logarithmic transformation as the dependent variable) for the three groups of states: treated, never treated and already treated. Vertical lines represent the year in which unilateral divorce laws became effective in each of the treated states.
Notes: This figure plots arrests of female prostitutes per 1,000,0000 inhabitants, in the same logarithmic transformation as the dependent variable, for the three groups of states analyzed in the study: treated, never treated and already treated. Vertical lines represents the year in which unilateral divorce law became effective in each of the treated states
This figure cannot be used to assess whether the trends of treated and control groups are parallel since the effective dates of unilateral divorce laws differ across states. However, it shows that, as many more states adopt unilateral divorce, treated states experience a substantial decline in arrests of female prostitutes per 1,000,0000 inhabitants, in line with my findings. In other words, as treated states adopt unilateral divorce, arrests of female prostitutes decrease more severely there than in control states.
1.7 Effective date of unilateral divorce laws across U.S. states
The effective date is established using Thomson Reuters Westlaw. In the section “Statutes and Court rules”, Thomson Reuters Westlaw keeps track of different legislations and when they became effective. This procedure establishes an effective month for each state that experienced a change in divorce law during my sample period. Figure 8 maps treated and control states (i.e., never treated and already treated, respectively).
Notes: This figure maps U.S. states according to their treatment status
1.8 Further tables sensitivity to model specification changes
This section presents robustness checks to changes in the specification of the regression model.
Source: Polaris (2017)
1.9 Comment on potential mechanisms: marriage compensation mechanism
1.9.1 Comparison group
There could be the concern that the finding that unilateral divorce has a greater impact on arrested prostitutes of marrying-fertile age is due to the choice of using arrested prostitutes of other ages as the comparison group. This latter group is composed of arrested prostitutes either between 17 and 24 years old or strictly older than 49 years old since the marrying-fertile age group is formed by prostitutes between 25 and 49 years old. The potential concern is that results are driven by the inclusion of prostitutes strictly older than 49 years old that might seem less frequent than their younger counterparts.
To address this issue, this section presents the results of running equation (6) but using arrested prostitutes between 17 and 24 years old only (i.e., arrested prostitutes older than 49 years old are excluded). Using only prostitutes between 17 and 24 years old signifies using only prostitutes of fertile age but too young to get married.
Table A.7 shows the results of running the same analysis as before but for this age group. Findings are qualitatively similar: there is evidence that unilateral divorce law has a larger impact on arrested prostitutes of marrying-fertile age than on arrested prostitutes of other ages. This evidence supports the marriage compensation mechanism.
Figure 10 shows the results of running equation (2) for the “17–24 years old” sample. In line with the marriage compensation mechanism, the coefficients show that the results are not driven by this age group.
Notes: This figure plots the estimated coefficients of the event study analysis for arrested prostitutes in the “17–24 years old” sample. On the horizontal axis is the event time in years (groups of 12 months). On the vertical axis, the coefficients are measured in terms of their effect on the dependent variable. The coefficients are measured relative to the omitted coefficient (t = −1). For each coefficient, the solid line graphs the point estimate, while dashed and dotted lines graph confidence intervals at the 95% level. The dashed-dotted line plots such confidence intervals clustering variance at state-time level, while the dashed line plots the confidence intervals clustering variance at state level. The pattern of the estimated coefficients is consistent with the marriage compensation mechanism: post-treatment coefficients are statically equal to zero supporting that results are not driven by this age group. The bottom right corner displays the p-value for checking pre-trends (following suggestions by Borusyak and Jaravel (2017)). Regardless of the level of clustering there is no evidence that pre-treatment variables are jointly statistically different from zero
1.9.2 Indoor prostitution
A potential concern could be that female prostitutes of marrying and fertile age became more difficult to arrest for reasons disconnected to their opportunity cost of getting married. As far as I can determine, there is no clear plausible mechanism that could support this explanation.Footnote 58
CPS data provide information on the occupational code; this allows me to restrict the sample to potential indoor prostitutes. Using the occupational code, I can restrict the sample to female respondents working in industrial sectors connected to indoor prostitution. Hence, I obtain a reasonable proxy for potential indoor prostitutes.Footnote 59 It is worth mentioning that a limitation of this analysis is that it does not consider a group of indoor prostitutes. Specifically, in the first decade of the 2000s indoor prostitution increased, this surge was mainly due to independent prostitutes operating on ad sites such as Craiglist or Backpage (Cunningham et al., 2017). It seems plausible to believe that such prostitutes would not declare to operate in the industrial sectors linked to indoor prostitution. Thus, the analysis here presented is uninformative about such prostitutes.
Namely, I consider the following regression model similar to regression model (4):
where Indoor prostitutessmy is the number of women in occupational sectors that contain indoor prostitution businesses per 1,000,000 inhabitants in state s, month m, and year y; αm, αy and αs are respectively month, year and state fixed effects; and \(\alpha_{s} * {y}\) are state-year linear trends. As in the previous analysis, I split the sample depending on the age of female respondents. In particular, I split the sample into two groups: indoor prostitutes of marrying-fertile age and indoor prostitutes of other ages.
Columns (1) and (4) of Table A.8 show the results of running equation (8) for marrying-fertile age and other ages. The results show that unilateral divorce decreases potential indoor prostitutes of marrying-fertile age but does not affect potential indoor prostitutes of other ages. Columns (2) and (5) report results using IHS, while columns (3) and (6) report the results in levels. The results are stable across functional forms.
1.9.3 Time use
This section exploits data from the American Time Use Survey to explore whether there is evidence that unilateral divorce increases wives’ welfare. Table A.9 considers specification (A.2) for three different outcomes.
In equation (A.2) the underscript asy stands for cohort a of state s in year y; Usy is an indicator variable taking value 0 if unilateral divorce did not go into effect and value 1 once it goes into effect and afterwards. female is an indicator variable taking value 1 if sex of the cohort is female and 0 if it is male. marital is an indicator variable taking value 1 for each of the j marital status categories in the survey and αa, αy and αs are respectively cohort, year and state fixed effects. The dependent variable Y is the average number of minutes per day in three activities measuring welfare and connected to bargaining power in a couple. Namely, these three different activities are personal care, sport and leisure and relaxing activities.
Since the sample spans only from 2003 to 2014 the number of treated states decreases. However, it is reassuring to find that unilateral divorce correlates positively with an increment in such activities for married women. This evidence supports the notion that the introduction of unilateral divorce improved wives’ welfare.
1.10 Industry sectors used to measure indoor prostitution
To measure potential indoor prostitutes, I restrict CPS data to the following occupational codes in the table below. The names of the variables are drawn from the monthly extracts of the CPS Uniform database of the Centre of Economic Policy Research (CEPR).Footnote 60 In order to code such variables, it is useful to use both SIC and NAICS systems.
Specifically, I restrict my sample to women working in industry sectors composed of strip clubs and escort-girl services (i.e., sectors that comprise indoor prostitution establishments). Note that these industry sectors are composed of various occupations, among which there are strip clubs, massage parlors and escort-girls services. Hence, women in this sample might be working in other occupations too. However, this sample is more likely to be formed by prostitutes. Recall that in the U.S., the prostitution market is highly stratified. Women arrested for prostitution are very likely street prostitutes, who make up the low segment of the market. The sample I extract from CPS data is composed of strip clubs, massage parlors and escort-girls services, who form the medium and high segments of the market. According to the theory, indoor prostitutes are as likely to respond to an increase in pm as outdoor prostitutes.
For variables ind70 and ind80, strip clubs belong to an occupational sector named “Miscellaneous entertainment and recreative services”, while escort services belong to “Miscellaneous personal services”. In the last three variables, these names respectively change to “Other amusement, gambling, and recreative services” and “Other personal services”.Footnote 61 This sample spans from 1980 to 2014. Sectors for variables occ70 and occ80 are labeled as “Personal service occupations, not elsewhere classified”. Finally, Sectors for variables occ03, occ11 and occ12 are labeled as “Miscellaneous personal appearance workers” and “Personal care and service workers, all other”.
1.11 Comment on potential mechanisms: fight against crime mechanism
1.11.1 Officers
There could be the concern that hired officers do not vary considerably over years and that this lack of variation is driving the results of the police mechanism.
To address this issue, this section considers equation (2) but makes use of two different transformations of the dependent variable. First, I use the first difference of officers per 1000 inhabitants. In other words, I use the variation (i.e., increase/decrease) of hired officers normalized by a state’s population. Second, I use the growth rate of officers per 1000 inhabitants. Results are presented in the same fashion as in the police mechanism analysis.
I find no empirical evidence supporting that unilateral divorce correlates with a reduction of officers.
1.11.2 Other crimes
This section presents results of running my main specification using as dependent variable each of the main categories of offenses recorded by UCR (28 main categories of offenses excluding prostitution).Footnote 62 Such offenses are recorded in two panels depending on whether there is evidence in the literature that they are connected to prostitution. Namely, Panel A shows offenses not connected to prostitution, while Panel B displays offenses connected to prostitution.
There is evidence in the literature (Cunningham et al., 2017, Dank et al., 2014, HG.org, 2017, Urban Justice Center, 2005) that prostitution is connected to different crimes. Using such literature, I divided offenses in two groups: connected and not connected to prostitution, as shown in Table A.12.Footnote 63
Each cell in the column shows the estimated coefficient, and its standard error, associated with unilateral divorce using the corresponding offense in the row as the dependent variable transformed according to the corresponding column. In fact, each column shows the results of running the above mentioned regression with a different functional form of the dependent variable. Columns (1), (2), and (3), respectively use the dependent variable in logs, IHS, and levels. Each regression includes month and year-fixed effects, county-fixed effects, and linear trends, and variance is clustered at the state level.
1.12 Comment on demand mechanisms
1.12.1 Internet searches
The first data set used is drawn from Google Trends. Cunningham and Kendall (2010, 2011b, 2013) contend that “overall, online solicitation represents an augmentation of the prostitution market”.Footnote 64 Indeed, according to these researchers, the advent of the internet has allowed prostitutes to (i) more easily reach a larger pool of potential clients, (ii) build reputations for their services and (iii) use screening to filter out unwanted clients.
Therefore, using Google Trends, I gather data on searches for different words that might be used by prostitutes’ potential clients. The frequency with which these words are searched online might proxy for the demand for prostitution. First, I consider different synonyms of “prostitute”. Second, I consider the word “sex”. Next, I consider words connected to indoor prostitution such as “stripper”, “strip club” and “escort”. Finally, I consider words connected to websites known for matching customers and prostitutes.Footnote 65 The Erotic Review is one of the most important websites that matches prostitutes and clients in the U.S.Footnote 66 It seems plausible that if the demand for prostitution exhibited a change in those years, the searches for such words should have also changed.
Since the Google Trends data set is at the state-month level, in this case, the regression is also estimated at that level. Then, I run the following regression:
where Searchessmy stands for the number of searches for a certain word in state s, month m and year y; αm, αy and αs are month, year, and state fixed effects, respectively; and αs*y is a state-year linear trend. If unilateral divorce increases (decreases) the demand for prostitution, the estimated coefficient should be positive (negative) and significant.
Google Trends data are available since 2004. Table A.13 displays the estimated coefficients after running such regressions for the largest sample I have (i.e., 2004 to 2017). While Table A.14 displays the estimated coefficients after running such regressions until 2014 to partially match the sample period of my main regression, Panels A, B and C show the results in levels, logs and IHS, respectively.Footnote 67
There is no statistical evidence supporting the notion that unilateral divorce might reduce the demand for prostitution. In fact, when considering words not connected to websites that match customers and prostitutes, the estimated coefficients change signs across regressions in both tables, and none is statistically negative. Regarding words connected to the aforementioned websites, the only coefficient statistically different from zero is that for “Backpage Erotic” . In both tables A.13 and A.14, this coefficient indicates that unilateral divorce increases searches for such websites. Overall, these findings suggest that unilateral divorce does not reduce the demand for prostitution.
1.12.2 Preferences of divorced men
Unilateral divorce law might indirectly affect the demand for prostitution. For example, it could be that it is the act of being divorced, instead of unilateral divorce law per se, that affects people’s attitudes.
To study this instance, I use data from the Youth Parent Socialization Survey (YPSS). This survey started in 1965 and had three other waves: 1973, 1982 and 1997. Since the YPSS followed individuals during these three waves, by using these data, it is possible to study how the observable characteristics of divorced people changed after their divorces.Footnote 68
In particular, to proxy for the demand for prostitution, I use changes in the opinions of male respondents about prostitution. This survey measures the dislike of their respondents toward various issues, one of which is prostitution. Consequently, I can observe whether, after being divorced, men report that they dislike prostitution more or less often than before. It seems reasonable to assume that higher levels of dislike of prostitution among male respondents might lead to reduced demand for prostitution, which could explain the findings of this paper.
I run the following regression model:
where DislikeProstitutioniw is a dummy variable taking value 1 if respondent i expresses dislike of prostitution in survey wave w, Xiw is a vector of characteristics that includes the sex of the respondent and marital status in wave w of the survey and αi,αw are individual and wave fixed effects, respectively. Finally, divorcediw is a dichotomous variable that takes value 1 if individual i was divorced in wave w of the survey. In addition, standard errors are clustered at the school code level.
This regression exploits the variation in being divorced across successive waves of the survey for a given individual to compute the correlation between divorced males and their aversion to prostitution. Namely, a positive β1 implies that marital dissolution, for both men and women, correlates with aversion to prostitution. Similarly, a positive β2 implies that divorced men are more likely to dislike prostitution.
Column (1) of Table A.15 shows the results of regression model (A.4), where divorcediw takes value 1 only for divorced respondents. Both \(\hat{{\beta }_{1}}\) and \(\hat{{\beta }_{2}}\) are not statistically significant. However, \(\hat{{\beta }_{2}}\) is positive, suggesting that divorced men might be more averse to prostitution. To check whether these findings are stable, I run three additional regressions. Column (2) of Table A.15 pools respondents whose marital status is divorced or separated (i.e., divorcediw takes value 1 for both divorced and separated respondents). In this specification, \(\hat{{\beta }_{2}}\) is negative. Furthermore, the size of the standard errors is unchanged, suggesting that the statistical nonsignificance of \(\hat{{\beta }_{2}}\) is not due to a lack of precision.
Notwithstanding, the previous regressions treat as divorced those individuals who were divorced in wave w of the survey. Hence, the same individual could be divorced in wave w but then married in wave w + 1. It is more conservative to consider as divorced (separated) individuals who were divorced (separated) at least once in the surveys. It might even be the case that it is only after the first divorce (separation) that men change their preferences toward prostitution.Footnote 69 This could explain the change in sign of \(\hat{{\beta }_{2}}\) across columns (1) and (2).
Consequently, as a further check, the last two columns of Table A.15 (i.e., namely, columns (3) and (4)) consider respondents who claimed to be divorced/separated in a previous wave of the YPSS as divorced and/or separated. As an example, suppose that individual j was divorced in wave 2 and married again in wave 3; column (1) would consider this individual to be divorced in the former and married in the latter, whereas column (3) would consider this individual to be divorced in both periods. Column (4) pools both divorced and separated individuals. Columns (3) and (4) of Table A.15 show that across both regressions, \(\hat{{\beta }_{2}}\) is negative.Footnote 70 In addition, in these columns, both \(\hat{{\beta }_{1}}\) and \(\hat{{\beta }_{2}}\) are not statistically different from zero, and the size of the standard errors is unchanged, suggesting that the lack of statistical significance is not due to imprecision. Consistently, the upper bound of the 90% confidence interval of the most conservative setting (column(4)) suggests that men who are divorced or separated for the first time are associated with an increased dislike of prostitution at most by 0.7%. This result suggests that being divorced or separated is not negatively associated with attitudes toward prostitution. Overall, these results do not support the notion that being divorced reduces the demand for prostitution.
1.13 Comment on migration and spillover effects
This section analyzes the hypothesis that police officers, clients or prostitutes might move to states where unilateral divorce goes into effect. Section 7.2 explores the notion that unilateral divorce might have an effect on the number of police officers hired in a certain state. Results indicate that unilateral divorce does not have an impact on the number of such officers. Likewise, Section 7.3.2 uses the number of unmarried men as a proxy of the number of clients of prostitution and finds that such a variable is unaffected by unilateral divorce regimes.
As for prostitutes, the main findings of this manuscript suggest that unilateral divorce reduces arrests for prostitution. Yet, this does not rule out the notion that high-end prostitutes from other states might exit prostitution and move to states where unilateral divorce goes into effect. In the sequel, I explore this hypothesis with two different approaches.
First, I use data on indoor prostitution from Appendix Section I.2. Table A.16 presents the results of running equation (A.1) but using the total prostitution of all states except state s. If entry into force of unilateral divorce in state s displaces high-end prostitutes to the same state, the estimated coefficient should be statisticaly negative.
It is reassuring to find that estimated coefficients across three functional forms (respectively, column(1) in logs, column (2) in IHS and column (3) in levels) are statistically insignificant, positive in point estimate and close to zero in size. In fact, the lower bound of the 90% confidence interval of the specification in logs suggests that unilateral divorce is associated with a decrease of indoor prostitution in other states of 0.39%. Estimates of similar magnitudes might be computed from columns (2) and (3).
Second, I use data on the sex ratio and population from CPS. This analysis might also be seen as test of whether unilateral divorce laws by making marriage more attractive to women affect the number of women living in a certain state. If this is the case the findings of this paper might be explained simply because the population in treated states increase. Moreover, this hypothesis would violate SUTVA since treatment in a certain state would affect the outcome in a different state.Footnote 71
To this aim, I use data on the sex ratio, female population and male population; specifically, since these variables are considerably stable, I use the growth rate of such variables. Below I present the results of running specification (6) using as dependent variable the growth rate of either the sex ratio or the female or male population; where, the sex ratio is defined as \(\frac{Women}{Men}\).
Columns (1), (2) and (3) of Table A.17 respectively present the results for the growth rate of the sex ratio, female population and male population. If unilateral divorce laws displaces women to states where such laws go into effect, the estimated coefficient of columns (1) and (2) will be statistically positive. Yet, this is not the case, such coefficients are statistically equal to zero, negative in point estimate and close to zero in magnitude. Indeed, the upper bound of the 90% confidence interval of the specification estimated in columns (1) and (2) suggest that unilateral divorce is associated with an increment of 0.24 and 0.05 percent points for respectively the growth rate of the sex ratio and the growth rate of the female population.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Ciacci, R. On the economic determinants of prostitution: marriage compensation and unilateral divorce in U.S. states. Rev Econ Household 21, 941–1017 (2023). https://doi.org/10.1007/s11150-022-09643-5
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11150-022-09643-5