The Impact of Dark Net Marketplace Shutdowns on Street Crime

Abstract (Revised)

The Dark Net has transformed global drug markets by shifting transactions from the streets to online platforms. This paper examines whether shutdowns of major Dark Net marketplaces affect street-level drug trade and drug-related crime. Using daily U.S. crime data and exploiting unexpected marketplace shutdowns, I employ a regression discontinuity design comparing crime immediately before and after shutdown events. The results show that marketplace shutdowns cause a short-term increase in street drug activity. Marijuana-related crimes rise by approximately five percent in the days following shutdowns but return to pre-shutdown levels within eighteen days. No significant effects are found for theft, assault, homicide, or prostitution. These findings suggest that online and street drug markets are substitutes, but only in the short run.

1. Introduction

Drug overdose deaths in the United States nearly quadrupled between 1999 and 2017, coinciding with the rapid growth of online drug markets. Enabled by cryptocurrencies and anonymizing technologies, Dark Net marketplaces have altered traditional drug distribution by reducing information asymmetries, improving product quality, and limiting exposure to violence.

Street drug markets are often associated with violence and acquisitive crime, both from dealers protecting territory and from users financing addiction. By contrast, online drug trade may reduce these risks by shifting transactions away from physical interactions. This paper investigates whether online and offline drug markets act as substitutes or complements, and whether Dark Net activity reduces street-level crime.

To identify causal effects, I exploit multiple unexpected shutdowns of large Dark Net marketplaces. Using a regression discontinuity design based on time, I compare crime outcomes immediately before and after shutdowns. The analysis shows that shutdowns temporarily increase street drug trade, particularly marijuana-related offenses, but the effect dissipates within two to three weeks as users migrate to new platforms. No evidence is found of spillovers into violent crime or other drug-associated offenses.

These results imply that shutting down individual marketplaces has limited long-term effectiveness and may briefly increase street drug activity. Policymakers should therefore consider the short-term displacement effects of enforcement actions and the rapid adaptability of online drug markets.

2. Institutional Context

2.1 The Dark Net

The Dark Net is a concealed segment of the internet accessible through anonymizing tools such as The Onion Router (TOR). While it has legitimate uses for privacy and political expression, it also hosts marketplaces for illicit goods, most notably drugs.

The modern era of online drug trade began in 2011 with Silk Road. Although its 2013 FBI shutdown was a landmark enforcement action, it failed to halt online drug trade. In the eight years following Silk Road’s closure, at least 87 new Dark Net marketplaces emerged. By 2017, the share of U.S. drug users purchasing drugs online had nearly doubled.

Illicit drugs account for roughly 85% of Dark Net marketplace revenues. Cannabis, stimulants, and ecstasy dominate sales, and both retail and wholesale transactions occur. Approximately one-quarter of online drug revenue comes from wholesale quantities.

Users of Dark Net marketplaces are typically young, male, and already drug consumers. Sellers are often former street dealers who migrate online to reduce violence exposure, lower arrest risk, and increase profits.

2.2 Drug Shipping

Most drugs sold online are shipped through standard postal services. Although shipments face seizure risk, detection rarely leads to identification of buyers or sellers. Sellers frequently consider national postal systems relatively safe, reporting high delivery success rates. If a package is seized, buyers can plausibly deny intent, limiting enforcement effectiveness.

2.3 Why Dark Net Markets Succeed

Dark Net marketplaces reduce key frictions of street drug trade. Feedback systems allow buyers to assess seller reliability and drug quality, reducing scams and incentivizing higher purity. Escrow systems mitigate seller moral hazard by releasing payment only after delivery, further increasing trust.

These mechanisms mirror those of legal online platforms and explain why online drug markets are more efficient and competitive than street markets.

2.4 Marketplace Shutdowns

Dark Net marketplaces can shut down due to voluntary closures, exit scams, hacking, or law enforcement raids. Most shutdowns are unexpected and represent temporary disruptions. Between 2011 and 2016, the vast majority of marketplaces closed, yet users and sellers quickly adapted by migrating to new platforms.

These shutdowns provide plausibly exogenous shocks that allow identification of the causal relationship between online drug trade and street crime.

3. Theoretical Considerations

This section outlines the mechanisms through which Dark Net marketplaces affect street-level drug trade and crime. Following Galenianos et al. (2012), illicit drug trade can be modeled as a search-and-matching process in which buyers incur costs to find trustworthy sellers. Trust is central because drug markets suffer from severe moral hazard: buyers only observe quality after consumption, creating incentives for sellers to adulterate products.

Dark Net marketplaces reduce these frictions through two key innovations. First, feedback systems allow buyers to observe vendors’ past performance, enabling sellers to credibly signal reliability. Second, escrow systems delay payment until delivery is confirmed, reducing the risk of seller opportunism. Together, these mechanisms stabilize buyer–seller matches and improve market efficiency.

Unexpected shutdowns disrupt these established online matches and temporarily undermine trust in online trade. This shock provides a setting to analyze how drug transactions reallocate between online and street markets.

Effects on Street Drug Trade

Hypothesis 1: Shutdowns of Dark Net marketplaces increase street drug trade.

This effect depends on the degree of substitutability between online and offline markets and persists until new online matches are formed. I assume that online and offline markets are imperfect substitutes and that online participants are relatively more risk-averse, choosing online trade to avoid violence and arrest.

When a marketplace shuts down, online buyers and sellers become unmatched. Buyers temporarily turn to street markets, while sellers redirect supply offline. These relationships remain unstable, as buyers continue searching for safer online options. Once a new marketplace gains credibility, online matches are gradually re-established and street trade returns to its pre-shutdown equilibrium.

4. Data

This study combines crime data, Dark Net marketplace shutdowns, and scraped marketplace listings. Additional socioeconomic data are used to assess external validity.

4.1 Dark Net Marketplace Openings and Shutdowns

Data on marketplace lifecycles come from DNM-Archives, which provides comprehensive records of Dark Net marketplace openings, closures, and closure reasons. To focus on economically relevant platforms, I restrict attention to marketplaces with lifespans exceeding the average (232 days). I exclude marketplaces with unknown closure dates, closures after 2017, and platforms operating primarily outside the U.S.

To avoid overlapping treatment windows, I require shutdowns to be at least 60 days apart. The final sample includes 13 marketplaces and 10 distinct shutdown dates. These marketplaces operated for an average of 21 months and closed due to law enforcement raids, exit scams, voluntary shutdowns, or hacking.

Table 1 lists the marketplaces included in the analysis along with their operating periods and closure types.

4.2 Crime Data

Crime data come from the National Incident-Based Reporting System (NIBRS), which provides incident-level records of reported offenses and substance involvement. To ensure consistency, I restrict the sample to local police agencies reporting continuously from 2010 to 2019 and exclude federal agencies to avoid conflating street crime with enforcement actions targeting Dark Net platforms.

The resulting dataset includes daily arrest counts from 3,309 agencies across 1,229 counties in 35 states, covering approximately 24% of the U.S. population. I aggregate incidents to the state-by-day level for the main analysis and also estimate models using national daily aggregates, as shutdowns represent nationwide shocks.

Drug-related outcomes include arrests involving ecstasy, crack cocaine, heroin, and marijuana. Additional outcomes capture theft, assault, homicide, and prostitution. Summary statistics are reported in Table 2.

4.3 Dark Net Listings Data

To provide additional evidence on substitution across online platforms, I use the Grams dataset from DNM-Archives. Grams was a Dark Net search engine that scraped listings from more than 20 marketplaces. The data include daily information on individual listings, vendor identities, and shipping locations from 2014 to 2016, with occasional gaps due to scraping limitations.

4.4 Socioeconomic Controls

State-level data on labor force participation, unemployment, income, and cash assistance are drawn from the American Community Survey (2010–2019). These data are used to verify that states included in the NIBRS sample do not differ systematically from excluded states.

5. Empirical Strategy

5.1 Empirical Setting

The objective is to identify the causal effect of Dark Net marketplace shutdowns on street crime. An ideal experiment would randomly remove access to online drug markets across locations. However, because Dark Net access is not geographically restricted, spatial variation cannot be used to define treatment and control groups.

To overcome this limitation, I exploit exogenous time variation in the availability of online drug markets using a regression discontinuity in time (RDiT) design (Hausman and Rapson, 2018). Unexpected marketplace shutdowns generate sharp, nationwide shocks. With high-frequency crime data, crime levels immediately before shutdowns provide a credible counterfactual for crime in the absence of the shutdown.

5.2 Research Design

I estimate the effect of shutdowns using a sharp regression discontinuity design, where time relative to the shutdown date determines treatment status. The baseline specification is estimated on a 30-day window around each shutdown date. The coefficient of interest captures the average change in street crime immediately following a shutdown.

The model includes state fixed effects to control for time-invariant heterogeneity and time-varying controls such as unemployment and median income. I estimate both a state-by-day panel specification and a national time-series specification, consistent with shutdowns being national-level shocks.

Smooth functions of time are allowed on both sides of the cutoff and are estimated using local polynomial regressions. Identification relies on the assumption that, absent shutdowns, crime would evolve smoothly through the cutoff. Under this assumption, the discontinuity at the shutdown date identifies the causal effect. Robustness checks include alternative bandwidths and linear, quadratic, and cubic specifications.

A key threat to identification is confounding events occurring near shutdown dates. This concern is mitigated by the use of daily data and multiple shutdowns. Moreover, if shutdowns coincided with other crime-relevant shocks, non-drug-related crimes would also display discontinuities. I find no such effects.

Another concern is temporal seasonality, particularly around weekends. I address this by including day-of-week fixed effects, ensuring identification relies on within–day-of-week variation.

Because time is the running variable, manipulation tests such as McCrary density tests are infeasible. However, advance knowledge of shutdown dates by market participants is unlikely. To further address anticipation concerns, I estimate models separately by shutdown type and show that results are driven by unexpected closures.