A Tale of Two Divergent Paths

Violence, particularly gun violence, is not an insoluble problem–there are evidence-based solutions that can greatly reduce harms. In the short term, new crimes can be deterred by increasing the certainty of arrest and prosecution. In the longer term, prevention can remove risk conditions and strengthen resiliency for both people and places. In forecasting, the best predictor of what will happen tomorrow is what happened yesterday, and that implies that today’s short-term problems are (mainly) the result of a failure to improve risk conditions yesterday. And — absent evidence to the contrary — that those failures will continue. 

The upshot is that the amount of crime and violence in 2023 will be less about how much harm will be imposed on New York City, and more about how much harm the city is willing to accept.

Prediction models

Crime prediction exists at the intersection of people and policy. At the person level, it is critical to think about violence and victimization from the perspective of someone who might commit a crime, which includes many people, particularly those exposed to harsh conditions. What are my opportunities today, and what are they likely to be in the future? What is my neighborhood like today, and what is likely to look like in the future? How are the people I spend my time with evaluating these same risks and opportunities? How has the pandemic changed those beliefs? All of these forces are intertwined.

Policy can affect these expectations and the behaviors that result. Long-term investments in prevention for everyone in a place — child and maternal health, social and emotional learning, and school attachment — can entrench positive expectations about the future early in life. Specific investments in people and places at the highest risk of violence and victimization can have even sharper positive consequences. And, of course, the power of deterrence is real. If the perceived likelihood of detection and punishment today is high enough, fewer people will commit crimes. 

In making a prediction about 2023, the key question is: Which is more likely to change, the immediate personal incentives to commit an offense or the social ecology which is the fabric of neighborhoods? Both have to be considered, noting that personal incentives are affected by what the police are doing today while social ecology is mainly affected by prior investments in prevention. 

In this case, the shock of the pandemic has been so multi-dimensional and the stress on individuals and the social framework so profound that it is impossible to predict each mechanism. Instead, we rely on long-term trends.

Here we offer two predictions — one about the number of shootings, and one about the number of major crimes — which also include robbery, assault, motor vehicle theft, burglary and sexual assault. While the two are deeply intertwined (i.e., a non-fatal shooting is also an assault), the data suggest that two very different patterns are emerging.

Gun violence

There are two patterns within the overarching long-term trends: a period of stability from 2001 to 2011 and then a long decline through 2019 (Figure 1). Even so, the number of shootings changes slowly — between 2000 and 2019, the average change in the number of shootings year-to-year was a modest decline of 3.7%. 

But the long-term models must also account for the shock of the COVID-19 pandemic. Between 2019 and 2020 shootings increased by 97.1% — essentially doubling. That new, high level of shootings stabilized in 2021, increasing by only 2.4%. More encouraging, though, shootings declined sharply last year — 17.2% through then end of 2022.

The most recent numbers suggest a return to a time of steady improvements in shootings and gun homicide. The 2022 decline is very similar to the period from 1993 to 1999 when, at the height of the crime decline, shootings declined 19.5% each year, on average.

Complaint data on what the NYPD label the seven major felonies (murder, rape, robbery, felony assault, burglary, grand larceny and motor vehicle theft) suggest a different trend than the shooting data. Most notably, in the most recent period, shootings peaked in 2020 and declined substantially in 2021 while 2023 reported crimes were at the highest level in any period (2006-2022). 

Figure 2 tracks arrests and official police reports (complaints) on the seven majors and the relationship between the two indicators can be used to forecast short-term trends. The first point to note is that arrests are essentially stable from 2006 to 2019, which should be no surprise since the number of sworn officers is essentially fixed throughout this period and arrests per officer tend to be stable over time. The year-over-year change in the number of NYPD officers stays below 3% with the exception that — most likely because of the pandemic — the number of NYPD officers dropped 7% in 2020.

The second point to note is that changes in complaint (Part I crimes) data lag arrests. Controlling for other explanations including typical season changes in reported crimes and policing, arrests tend to be a leading indicator of increases in reported crime. That is, a rise in the number of arrests tends to precede a rise in reported crimes. This is likely not causal, but rather that police respond to rising crime by increasing arrests‚ and there is a lag of at least six months for the effects of policing to change the crime trajectory.

Thus, these short-term indicators suggest that at least the first half of 2023 will have high levels of reported crimes. Overall, we are forecasting a 10% increase in the seven major felonies for 2023.

Authors’ note: The forecast of a 10% increase in the seven major felonies in 2023 was determined in early January. Through the first week of March, the number of (seven major) felonies is remarkably unchanged from 2022 — 20,571 in 2022 and 20,561 in 2023. However, without any large-scale change in policy to point to, we are maintaining our forecast for 2023 of a 10% increase in the seven major felonies in NYC in 2023.

What could change this forecast

This kind of trend analysis is sensitive to unusual mechanisms, referred to as “shocks." Some shocks to the system appear inevitable. The Federal Reserve is rapidly increasing interest rates to reduce inflation, which (intentionally) increases the number of unemployed people. Rising unemployment, particularly within minority communities and among young adults who historically have the highest rates of unemployment, seems certain for NYC in 2023. While past scholarship finds that unemployment directly affects property crime but not violent crime, increases in unemployment create uncertainty, which is correlated with more crime and violence. Similarly, the number of guns purchased — some of which reliably become used in crimes — is subject to (upward) shocks. Thus, we have adjusted our expectations toward the top of our confidence interval.

But positive shocks are possible too. If New York City, the state of New York, and the federal government acknowledge the pressure macroeconomic effects are likely to have on crime and violence, and invest accordingly, these forces can be mitigated. The catch is, the scale of the response has to be equal to the scale of the problem to move the dial. That requires substantial investments in New York City children, young people and distressed communities right now. With the political climate leaning in the opposite direction — toward closing libraries rather than flooding poor communities with books and teachers — we did not adjust our prediction models to accommodate positive shocks.