Episode 30: Brittany Street

 
Ep30_Street_photo.jpg

Brittany Street

Brittany Street is a post-doctoral researcher at the University of Michigan.

Date: May 26, 2020

Bonus segment on Dr. Street’s career path and life as a researcher.

A transcript of this episode is available here.

Episode Details:

In this episode, we discuss Dr. Street's work on the effects of employment shocks on criminal behavior:

"The Impact of Economic Opportunity on Criminal Behavior: Evidence from the Fracking Boom" by Brittany Street.

OTHER RESEARCH WE DISCUSS IN THIS EPISODE:

 

Transcript of this episode:

 

Jen [00:00:07] Hello and welcome to Probable Causation, a show about law, economics and crime. I'm your host, Jennifer Doleac of Texas A&M University, where I'm an Economics Professor and the Director of the Justice Tech Lab.

 

Jen [00:00:18] My guest this week is Brittany Street. Brittany is currently a postdoc with the CJARS Project at the University of Michigan, and she'll be joining the Economics department at the University of Missouri as an Assistant Professor in the fall. Brittany, welcome to the show.

 

Brittany [00:00:31] Great. Thank you for having me, Jen. I'm a huge fan of the show, and I'm excited to share my research today.

 

Jen [00:00:38] Well, thanks for being here. I'm excited you're here today. We're going to talk about your research on how local employment shocks affect crime. But to kick things off for us, it - could you tell us about your research expertise and how you became interested in this topic?

 

Brittany [00:00:52] Yeah. So I am generally interested in crime and the criminal justice system, which I consider to be pretty similar, yet distinct fields. I'm also interested in education research, which was a part of my dissertation at A&M and actually ties into the motivation behind this project that we're going to talk about today. So I wanted to understand, as you mentioned, how individuals were changing their criminal behavior in the face of these changing economic conditions. And in more recent years, we've had this concern over deteriorating economic conditions and those conditions then leading to adverse social outcomes such as substance abuse, crime, mortality. I'm going to study the opposite of this. So I'm using an economic boom from the sudden increase in hydraulic fracturing that the US experienced in the late 2000s. And we'll get into more, I'm sure. But there's this well documented increase in crime in these areas, which is actually a little bit perplexing at first when you're thinking about this typical model of crime that says researchers are familiar with where legal as legal activities have a better payoff -- and we think that individuals will start switching using their frame from illegal activities to these legal activities, suggesting the overall crime should actually decrease. So we can think of other things that might be going on. For example, individuals could be responding to changes in criminal opportunities or income inequalities, but it could also have nothing to do with how individuals change or respond with their criminal behavior, but rather with the fact that people are now moving into the area.

 

Brittany [00:02:27] And so I was interested in this idea of compositional changes that are happening in response to policies. And this is something that we often talk about when thinking about policy evaluation more broadly. So to take a step back, this is actually rather common in the education literature, where you often have a school level policy and an aggregate school level measure say something like performance as measured through test scores. The idea is that you have a policy that will hopefully improve these test scores, but if parents of high achieving students like that policy, you might see these students enrolling in those schools at higher rates. So now you see this increase in aggregate test scores after a school implemented this policy, but it becomes difficult to know if the policy improved actual students test scores or if you're just observing the higher test scores now that these higher ability students are enrolled in the school. So it becomes really a matter of are you changing individuals test scores or is it just this compositional change that is leading to this aggregate improvement in test scores? And you can maybe see why this is really important from a practitioner standpoint and for researchers.

 

Brittany [00:03:41] So this is going to be relevant going back to the context I'm specifically interested in and thinking about how changing economic conditions affect criminal behavior because this is the exact type of setting where you might worry about compositional changes that I was just describing. So specifically, we know people are willing to move for work, and when times are hard in an area, people that can will move to the stronger labor markets.

 

Brittany [00:04:10] This means that when trying to study how economic conditions are affecting criminal behavior or aggregate measures of crime, say, from like the UCR, which I'm sure many of your listeners are familiar with, has the same challenge that I was just describing. It does an excellent job showing how crime changes in aggregate. It just then becomes really difficult to say whether this change in observe local crime is due to individual changes in behavior or due to this changing composition and that was really the main focus I had in this paper.

 

Jen [00:04:43] So your paper is titled "The Impact of Economic Opportunity on Criminal Behavior Evidence from the Fracking Boom." So let's start with some basics. What exactly is fracking?

 

Brittany [00:04:53] Great question. So hydraulic fracturing involves drilling into the shale deposit. So shale deposits are these mineral resources deep beneath the ground. And so you can drill into these shale deposits and inject fluids at a high pressure, which then cracks the rock formation and release the natural resources. So fracking in general has been around for quite some time, but it's gone through several technological advances recently. So this process can be used to extract both oil and gas, and I focus on oil just because that's what's most prominent in North Dakota where I study. But in general, it became really big in the United States once these technological advances, such as seismic imaging and horizontal drilling came about. So this made these oil formations that were previously not economically viable, all of a sudden very valuable. And so in 2000, fracking extraction only made up about two percent of domestic oil production, but by twenty fifteen and made up over fifty percent of all domestic crude oil production in the United States. So during this time, domestic oil production grew faster than at any other point in history, and in doing so created this local economic shock that affected many areas throughout the United States.

 

Jen [00:06:12] And you're going to be using fracking really as a case study to understand what the impact of this type of economic shock is. So before this paper, what did we know about how economic opportunity more broadly affects crime?

 

Brittany [00:06:25] Yeah, so it's pretty well established that crime increases as economic conditions worsen and that crime decreases with better economic conditions, which is what we expect. So there are several papers that document this, and I won't go into all of them in detail, but these are - you know there are several great papers which establish this relationship using quasi random variation in economic conditions, such as better fishing waters or negative trade shocks. And so all these papers kind of point to this well documented, established relationship that we might expect. But then we see an increase in crime with increases in local economic conditions, which gets back to this perplexing relationship that I was talking about earlier. And one example of this is Freedman and Owens's paper that studies the effect of increased economic conditions through BRAC construction in San Antonio, Texas. So for those that don't know BRAC is base realignment and construction, and you can think of the construction of these bases as being a exogenous local economic shock to an area, particularly for construction workers. So they find that crime actually increases in neighborhoods that have been historically where a larger share of construction workers live, which is the group that's likely to benefit from the shock. So why is crime increasing in these areas that are most likely to benefit? And they find the individuals that had already had a felony conviction, and so are not able to be employed by these government construction jobs are actually the ones more likely to be committing crime. So this is in line with the idea that there's an increase in economic opportunities and perhaps income inequalities can result in an increase in crime from these improved economic conditions.

 

Jen [00:08:11] And then what did we know about the effects of the fracking boom in particular?

 

Brittany [00:08:15] Yeah, so we know that the fracking boom did increase wages and employment, which is documented by several studies. So, again, I won't go into all of these, but it was a local economic increase in these areas. So Gittings and Roach document that a lot of these new jobs actually went to people from outside the county. This is still consistent with increased economic opportunity for local residents in predominantly rural states such as North Dakota, as they likely benefit from the large economic stimulus in their county and then companies just need to bring in outside workers to fill in the large number of jobs that are created. This is also consistent with a forthcoming paper by Wilson, which shows a large migration to fracking counties, particularly in North Dakota, and that migrants are disproportionately male, young, unmarried and less educated. This is all to highlight the importance of accounting for these potential compositional changes.

 

Brittany [00:09:12] So moving over to crime specifically, as I have mentioned, it's well documented the aggregate crime increases in response to hydraulic fracturing activities, and a few papers demonstrate this. In particular, Smith and James's recent paper shows an aggregate increase in violent and property crime in areas with hydraulic fracturing. They investigate some potential mechanisms and document an increase in income inequality in fracking counties, likely from oil royalties and or these high paying drilling jobs, suggesting a potential driver for changes in individuals criminal propensity. They also document an increase in felony sex offenders moving into these counties, and this then suggests the other potential mechanism being compositional changes.

 

Jen [00:09:59] And so you are interested in how the fracking boom in North Dakota affected criminal behavior in the counties most affected by this economic shock, as you talk about in the paper? Fracking actually affected the state in two stages. So you differentiate between the leasing stage and the production stage. So describe the timeline for us. How did this whole process play out?

 

Brittany [00:10:21] Yes. So you start seeing the first signs of fracking in North Dakota in 2004, companies began sending what's known as landmen into counties with fracking potential and began signing oil leases with mineral owners. These leases gave the companies the right to develop the natural resources in exchange for a percentage of the oil revenues from the production paid to the owner. This activity spiked in 2004 once companies realized the potential for hydraulic fracturing in these areas was somewhat of a race to sign leases, then there's a bit of a lag until you see production start rapidly increasing in 2008 as it typically takes a year or two to set up the drilling and start actually extracting and not all start immediately once signed either, but 2008 marks a 40 percent increase in oil production in North Dakota and so that's how I kind of defined this as as a transition from the leasing period into the production period.

 

Jen [00:11:19] And so what mechanism should we have in mind for how these two stages might affect local crime?

 

Brittany [00:11:25] Yeah, so let's start with the leasing period. Really, the only thing that's changing is that companies begin signing leases, which is signaling the start of economic expansion with jobs starting to trickle in and production gradually starting to tick up. I think of households just responding to these new job opportunities, whether expected or realized during this period. And there are a few mechanisms through which the job opportunities could operate, including a type of incapacitation effect from being employed, eased financial constraints, the need to have a reasonably clean background for employment opportunities, the cost of being caught is now higher as individuals having better jobs that they may risk losing. Or it could just be that there's this better economic outlook in an area that previously did not have a lot of economic activity and the economics of the despair are beginning to be reversed.

 

Brittany [00:12:20] The production period starts to be a little more complicated, which is why I separate it into two periods. The production period is more labor intensive, so the jobs are beginning to pour in and so do a lot of people and with the production begins the oil royalty payments to some households. So now we can expect changes in crime to operate through several different channels. First, we still expect households to be responding to these increased economic opportunities and for them to operate through the various channels that I just described. Then, even though we are isolating these local residents, we still expect that the influx of new people to affect how the local residents, their decision and how they commit crime, and this could be through increased social interaction, friction with newcomers or the negative peer influence. And then finally, you have this these large royalty payments, which could affect both the leaseholders and the non leaseholders criminal behavior.

 

Brittany [00:13:19] So those with leases could change the criminal behavior to the extent that the money eases, financial constraints is spent consuming complements to crimes such as drugs or alcohol or creates within household conflict. There is also a potential for the non-leaseholders crime to change due to changing criminal opportunities and being on the lower end of the increasing income inequality. They may also have similar changes in crime related to increased household income through higher wages. Yet, households that do not sign leases is the group that we actually expect to be the most responsive to the new economic opportunities since they don't receive the non labor income transfer. So since there are different activities within these different counties and thus potentially different mechanisms between the leasing and production periods, I consider each of these periods separately as we go through the paper.

 

Jen [00:14:14] And just to clarify on the lease front, how much money are we talking about when when a household signs one of these leases and gives the company the right to drill oil from underneath their land, how much money could they expect to get from that?

 

Brittany [00:14:27] Yeah, so that's a really excellent question. So typically, leases are signed to allow 12 to maybe 18 percent even of total oil revenues, and so basically I was able to collect several different data sets in order to try to back this number out. So how this works in North Dakota is that you have spacing units - and so this is so that, you know, individuals have different land and it's kind of hard to say, well, the oil beneath this property belongs to this person, but the oil beneath this property belongs to someone else, even though they're adjacent. Right. So these oil fields kind of spread different people's property. So in order to try to estimate this number, I was able to collect different data sets covering what spacing units cover different households in North Dakota, as well as the total production for that given spacing unit or the wells contained in the spacing unit. You can then dollarize this just using the price of a barrel of oil and then subtract off some taxes and other fees and at the end of the day it looks like the the median leaseholder receives roughly two thousand dollars per month. So this is a sizable amount considering the median monthly income is $5,000 in North Dakota in 2015.

 

Jen [00:15:48] Yeah, that's a lot of money. So, okay, so we've got people who - you know, production - the leasing period, start their sort of the promise of lots of jobs coming in. There's the promise of all this money.

 

Jen [00:15:59] You've got people who are local who stand to gain from both of those things, some more than others and then but then in the end, as you talked about before, you have you're going to have a bunch of people moving in during the production period because there are a lot more jobs to fill than than locals can fill themselves. So lots going on. But but I think it's interesting to start thinking about how all of this might wind up playing out in practice.

 

Jen [00:16:28] So as you started thinking about this project and wanted to measure the effect of economic opportunity on local residents criminal behavior, what were the main challenges that you had to overcome in order to do that? Were the hurdles mostly in getting the right data or was it in finding a good natural experiment or was it both?

 

Brittany [00:16:45] Both, definitely. So I think there are - the two main challenges are exactly what you just described. So first, you know, economic conditions are not random, and this is a pretty standard problem for empirical studies. So why a company such as Amazon chooses a particular location for their new headquarters rather than another? Or why a farm closes has to do with a lot of factors, and these factors could be directly related to crime or crime prevention or that could be indirectly related to crime. If they're related to say something like the education, public infrastructure, our poverty levels in the area, so the first challenge has to do with the credible variation in economic conditions and the need for this viable natural experiment. But even if you have the coin flip dictating when economic conditions improve or worsen, you're still going to be up against some pretty big data challenge and so the data challenge here is actually two fold. And first, crime data needs to be at the individual level rather than commonly used county level measures. Fortunately, administrative data is becoming better available in more recent years, but often it needs to be collected county by county or, if you're lucky, state by state, and it might not go back 20 years. The second data challenge is that even if you have individual level data, you don't actually know who was already living there prior to the boom, starting without some other data set and it's knowing who was where and when that is especially tricky when thinking about figuring out when trying to figure this out for over a decade ago. And so this is really the real barrier,.

 

Jen [00:18:25] Right, because in this case, what you're really interested in is separating this compositional effect from these other these other mechanisms that might be going on. And so you were able to overcome that data hurdle. So tell us about the data you were able to use in this project and how you found it.

 

Brittany [00:18:42] Yeah, so the data for the project comes in three main parts. I first needed to identify which households were already in each county more than a decade ago. This was probably the hardest part for me. The initial plan was to use property records, but many counties don't have electronic records or keep paper records dating that far back, especially when you need it from all the counties for statewide coverage. Fortunately, I came across the Great Plains Directory Service. This is a company that printed rural residential directories for several states, including North Dakota. Basically, every few years they collected the property tax records from each county for people living outside the city limits and then verified who was living at the property and recorded their name, address and phone number. I ended up finding these in the basement of a state library and hired a temp worker to take pictures of each individual page on a cell phone and then zip them over to me. These were OCR to create a registry of over thirty thousand households. This was a big win for the project in the sense that I can use this as a set of households that I knew lived in each county prior to the large immigration that's associated with fracking, because the fracking boom does have this unique feature of oil royalties in addition to the economic expansion and as we mentioned, it's a rather substantial amount. I wanted to be able to identify which households receive these additional non labor related income shocks.

 

Brittany [00:20:12] I collected this information from Drilling Info which is a private company that collects almost all leases signed throughout the United States. This database is becoming more common for researchers studying mineral leases and importantly for me, it includes who the mineral owner is, their address and the record date. Ultimately, I want to track changes in criminal behavior of these local residents, so to measure this, I collected administrative individual level records covering all 53 counties from the North Dakota judicial branch, spanning from two thousand until twenty seventeen. Importantly, these records also contain name, offense, level information, file, date and address information. These records were provided under an MUA, which is common for receiving administrative records for research purposes that are not publicly available and are similar to many court records that some of your listeners might have used as well. The critical feature then behind all these data sets is that they contain name and address, and this is what allows me to link across the data sets. So I define households in the Great Plains directory by their last name, the street number of their address, the city, and the zip code. And then I use a Levinstein index to match households to whether or not they signed a lease and when they signed the lease. And then I link this to all the cases filed. So I know when the household members committed a crime as well.

 

Jen [00:21:41] And so we should think of those court records as being charges or convictions?

 

Brittany [00:21:46] Yes, you should think of these court records as being charges.

 

Jen [00:21:49] Okay, great. So then how do you use all that data to measure the causal effect of the fracking boom on crime? What's your empirical strategy here?

 

Brittany [00:21:59] So I use a generalized difference-in-differences strategy which uses household in your fixed effects rather than a pre post or treatment control indicators. Intuitively, I will compare households in fracking counties to those in non fracking counties before and during hydraulic fracturing activities.

 

Brittany [00:22:17] This is why it's important for me to have data going back to 2000 before the fracking activity started. The household fix effects are going to control for these static differences between households. So if some households are just more crime prone in general, that is fine. Then you have your fixed effects which take out due to your shocks, which will affect all households both in the treatment and control counties such as the Great Recession or other statewide policies. What I'm interested in then, is the relative change in criminal behavior between households in fracking counties compared to households in non fracking counties. So I'm estimating the treatment effect as the difference in criminal behavior between these households during each period - the leasing and production period relative to their baseline difference between these households before any of the fracking activity started.

 

Brittany [00:23:08] So the assumption behind this approach is going to be that households in fracking counties would have had similar changes in criminal behavior as households and non fracking counties had the fracking boom not taken place. The nice feature of this shock that affected areas based on where the mineral deposit was beneath the ground and the massive increase in crude oil production was not anticipated, even by the Energy Information Administration, which creates these, you know, yearly updated forecasts. This overcome some of the common challenges with difference in difference designs related to endogenous policy adoption. It's also it's still possible, though, that my treatment control counties don't track for other reasons, as with any difference in differences design. And so I'll employ a series of checks to make sure I'm not picking up something other than the response to the boom, such as my treatment county systematically responding to year to year shocks differently than my control counties.

 

Jen [00:24:07] Yeah, but this really - the fact that this is sort of a natural phenomenon really does go a long way to helping on the natural experiment front. So the example you gave earlier was Amazon recently was trying to figure out where to open its next headquarters and where they might carefully choose one county over the other counties, maybe because there are lots of highly educated people there who could be workers or they get good tax breaks or they're good roads or whatever it is. And here it's really hard to argue that the oil companies chose certain counties in North Dakota to drill in for any reasons like that. It just the oil was either there or it wasn't. And when no one see that coming, then that should avoid any problems, that there would be underlying differences between the counties that we that we might worry about.

 

Brittany [00:24:53] Yeah, that's exactly right. And I think when using this type of methodology, you want to try to convince yourself that, you know, this is kind of the ideal setting in which you can use difference-in-differences because you do expect these areas to be tracking because there isn't going to be some other thing unobserved to the researcher that dictated these decisions. Right. So it was completely driven due to this natural resource, which creates a really nice and natural experiment.

 

Jen [00:25:22] Yeah, really, it's the epitome of a natural experiment. So what are the main main outcome measures that you're going to be focused on here?

 

Brittany [00:25:31] So I care about how criminal behavior changed. So my main outcome is whether or not any case is filed for a household in a given year. I also split this into four of my most prevalent crime categories - so drug related crime, theft, driving, or this other category, which includes crimes that are not as common in my sample, such as violent crime, criminal conspiracy and several others. So those are my main outcomes, although you can also use the number of cases or even charges filed, but I relegate that to kind of appendix material to make sure that how the the main conclusions of the study are sensitive or not sensitive to this. But really, at the end of the day, that doesn't matter. And I can just focus and tell the story with whether or not any case was filed as the primary outcome measure for how criminal behavior is changing.

 

Jen [00:26:23] Okay, so let's talk about what you found. What was the effect of fracking on the criminal behavior of people who had lived in the affected areas before the boom started?

 

Brittany [00:26:32] Yeah, so we have discussed that the fracking boom happens in two periods during the leasing period. I find that households living in fracking counties have a 28 percentage point reduction in the likelihood of having a case filed relative to household and non fracking counties. This represents a 14% relative drop in cases filed. This continues into the production period with a .35 percentage point reduction or roughly a 17% drop in having any case filed. You might be concerned that fracking counties are just different from my control counties. After all, they might be on different sides of the state, so you may wonder if these effects are going to be picking up something else systematically different about these counties or how households in them might be responding to these year to year shocks. But overall, these results that I just mentioned are robust to including interactions between pretreatment controls such as population or per capita income for the overall county and these year to year effects. So this allows for counties with different levels of observable characteristics to counties that are just more rural or lower income to respond differentially to year to year shocks. And since the results are robust to the specification it supports, that these detected effects are a change in criminal behavior in response to the changing economic conditions rather than something else.

 

Jen [00:27:56] Right. When you look at the graph, you've kind of - you've got these beautiful flat pre-trends, which we love to see in these kinds of graphs, which basically shows you that the estimated difference between the treatment and control counties is zero during the pre-period and that gives us some confidence that all your controls or are soaking up whatever differences there might be between these places. But then the the the leasing period starts and you start seeing crime drop. So it's a pretty striking to see it in the graph.

 

Brittany [00:28:22] Absolutely.

 

Jen [00:28:24] Were those effects driven by particular types of crime?

 

Brittany [00:28:28] I do find it decreases in all crime types, but the decrease is particularly pronounced for drug crimes during the leasing period, and this holds up to correcting for multiple hypothesis testing. So the idea that we're ruling out that these effects are happening by chance, because now I'm testing for several different groups or crime types. So why drugs? This could be for a few different reasons, which I can't quite disentangle, but also just a few. It could be that residents have a better outlook on their future economic situations as companies begin starting up in their county, which might lower drug use in line with the economics of despair, hypothesis or hope in this case. Or it could be that workers are preparing and applying to these jobs where many oil companies, drug test employees and do background checks. It could also be the incapacitation effect as individuals begin acquiring jobs early in the leasing period and have less time or financial need for drug related crime.

 

Jen [00:29:26] Did you find different effects on people who were leaseholders versus those who weren't?

 

Brittany [00:29:31] Yes, I do. And this is an interesting result because you could expect several different things. So as we're talking about mechanisms earlier and how households receive non labor income shocks, do these oil royalty payments while other households don't? Again, you could expect this to change both leaseholders, criminal behavior and non leaseholders criminal behavior, but for different reasons. So I consistently find that larger decreases for non-leaseholders or those that are not receiving this large non-labor income transfer through the oil royalties. These estimates are statistically different during the leasing period and remain larger for non- leaseholders into the production period. This suggests that it's the increase in job opportunities that is driving the reductions in crime rather than just an increase in income per say, and that the effect of job opportunities seems to be stronger than the effect of increased criminal opportunities. And perhaps here that is because this type of economic shock is an all ships rise type of shock.

 

Jen [00:30:31] And then how did these effects change over the fracking period? So as you moved from the leasing period to the production period, you can look at how these crime patterns are evolving over time and what do we learn from how those patterns change?

 

Brittany [00:30:46] Yeah, I find a decrease in crime for households in fracking counties relative to household and non- fracking counties in both the leasing and production period. However, I find the effect seems to be offset for drug related crimes in the production period. This suggests that some other things that are changing during the production period offset the effect of the job opportunities. It could be a combination of illegal drug markets moving in during the economic expansion and the production period, perhaps to cater to this influx of people or the negative peer influence from newcomers or this new disposable income being used on drugs. Empirically, I saw some descriptive evidence of a sharp increase in establishments with liquor license in fracking counties, which somewhat supports this notion of more availability and demand for things like drugs and alcohol. And thinking more about these offsetting behaviors we can also think about variation in fracking intensity within fracking counties. So I find a decrease in both minor and more major fracking counties during the leasing period. However, as the production ramps up, I find a sustained decrease in the minor fracking counties, which don't have as many other offsetting effects, such as the larger population influxes just to name one. And in contrast, the effects diminish in the major fracking counties during the period, which likely has a lot of other things also happening.

 

Jen [00:32:11] So as a good way to summarize this, that basically everyone is kind of on better behavior when the leasing period starts and then once a bunch of young men start moving in to take these jobs, the behavior starts getting worse for everyone involved. As you know, bars are opening and there is perhaps more going on in their small towns. You see drug offenses go up and so on. Is that is that intuitively the story here?

 

Brittany [00:32:38] Yes, in some cases. So I do see still an overall decrease, but you start to see kind of these specific types of behaviors exactly where you might expect them to start to be offset. And so, yes, that's exactly right.

 

Jen [00:32:53] And then you talked earlier about how other papers have found that crime rates went up overall during this period, you can also look at this. So what did you find in your data? Was the overall effect of fracking on crime rates? And how did those changes in crime rates then compare to the changes in criminal behavior that you're finding among the incumbent residents that had been there before all this started?

 

Brittany [00:33:16]  Right, so consistent with the literature, I also document an increase in aggregate crime in fracking counties relative to the non- fracking counties during this production period, I'm able to add somewhat to this body of literature by using the administrative court records, which also include other types of crime which are not readily, readily observed in other data sets such as drug and driving related crimes, and also seen increased during this period for those types of crime. Specifically, I measure a relative overall increase of roughly 40 cases per 1,000 people in fracking counties during the peak of the fracking boom in 2012. This starts to ramp up during the production period and begins to drop coinciding with the fall in oil prices starting in 2014 and 2015. This is in stark contrast to the relative decrease in criminal behavior for households in those same fracking counties, which is consistent with people decreasing their criminal behavior as they're exposed to better economic conditions and suggesting that the increase is driven by other factors.

 

Jen [00:34:22] And other factors, in particular, the changing composition of the local population, so basically crime rates are like, you know, when you look at the graph, the crime rates are just like skyrocketing places. But at the same time, you've just shown us that that crime is falling dramatically among people who are already living there. And so that really highlights how important it is that you have this influx of people who, because they tend to be young men, tend to be among the you know, the more crime prone population. That's that's the group that tends to commit most of the crime and get into most of the trouble. So you've got this policy that brings a lot more of those people into your local area and crime rates go up by quite a bit.

 

Brittany [00:35:02] Right.

 

Jen [00:35:04] Was the number of police officers per capita changing at this time? It seems possible that these changes in crime and criminal behavior could be affected by changes in police capacity during this period. Was that happening at all?

 

Brittany [00:35:15] So I do not find evidence of that, but in general, this is a concern and a common issue for crime research. So really, we just want to know what's driving the changes in crime, whatever it is, whether it's the police or the economic conditions. So I thought a lot about, OK, you know, it's important to get this right and and how best can we attempt to measure this? So first, I look at institutional evidence and I call the police chief in Williston, North Dakota, which for those that know fracking, they know that this is the town that has really become the poster child for the fracking boom and the chief had worked in Williston for close to 20 years. So, I asked him, you know, what changes he had observed in his role over that time and specifically related to the hydraulic fracturing activities. And he said everything was pretty normal until about 2010 when it was more like a light switch was just flipped and all of a sudden companies were bringing people in by the truckload and this suggests that there really wasn't a change in police resources until later into the production period.

 

Brittany [00:36:18] I also look for empirical evidence relating to changes in the number of police officers in total and the number of police officers per capita. I use the main difference in differences specification, but this time at the county level to show how counties change with respect to their police force during these periods. And doing this, I find no change in the number of police officers during the leasing period, but the number of police officers does start to increase later in the production period and so it could have some role towards the end of my sample. However, then you care about whether the police force was covering more or less people. So is this an effective change in police resources? And again, you see no change in the population during the leasing period, but increases in the production period coinciding with when companies were bringing workers in. So ultimately, I find that there's no change in the police force during the leasing period, either in total or per capita.

 

Brittany [00:37:13] And during the production period, while there's an increase in the number of police officers, it turns out there's no change in police per capita. This is also in line with work by James and Smith, in which they rule out changes in police resources as a driving mechanism behind the aggregate changes in crime that they observe in fracking counties.

 

Jen [00:37:32] And then you run a few other checks to confirm that your estimates aren't driven by other potential confounding factors, such as people moving out of the area or changes in household composition to talk about what you do there and how you're able to convince yourself that such changes aren't driving your results.

 

Brittany [00:37:50] Yeah, so this is an important point. I observe houses in each county in the early the 2000s. There's likely some moving in attrition, which on its own is not a major concern, but you worry that there might be differential attrition and treatment or control counties, and that's more concerning. This can actually happen in a few ways. So I'll test for it in many ways as well. However, increased moving in non- fracking counties relative to fracking counties is less of a concern because it would bias me against my result. So I'm going to focus my attention on whether or not people are more likely to move in within the fracking counties. First, you can have households differentially move out of the county or state to check for this I'm measuring outmigration using county migration files from the IRS. These files record what county a person files their taxes more specifically, I count the number of exemptions which better represent the number of people on the tax return. And so this is commonly used as as a yearly estimate of population. So a person that filing county in one year and counting being the next would be an outmigration of A, an immigrant of B, and this way I can measure differential outmigration from fracking counties. I find no evidence of differential outmigration until 2012, when those who had moved in at the start of the production period likely begin to turn out.

 

Brittany [00:39:14] This is also consistent with Wilson's analysis that out-migrants from fracking counties are often just the transient workers finally moving Next there could be a concern of differential moving within a county. I should note that there is a housing shortage during the fracking boom and people in the rural residential directories are likely a fairly stable households, which may make moving within county less likely. Nevertheless, to check for signs of movement within county, I collected county level property sales and prices. I don't see major changes in either until the production period, which is reassuring. I also use the American Community Survey to measure respondents that report moving within county, which unfortunately only covers the whole state starting in 2009. Descriptively, it looks like people are moving at similar rates across non-fracking and fracking counties, but there is an increase within county movement in the major fracking counties, again starting after 2012.

 

Brittany [00:40:14] Then finally, since this analysis is at the household level, you might worry that people are moving within the household. So, for example, particularly if young men were more likely to leave the household in fracking counties relative to non- fracking counties, then that could explain the drop in crime that I observe. Again, I'm not picking up differential outmigration from the county and if anything, young men would likely leave the non fracking counties in order to move to the fracking counties. But I still want to check for this, the specific story, and so I limit my sample only to crime committed by household members that are already 25 at the start of my sample and so they're more likely to be a stable group. I then tracked changes in household criminal behavior and find the same effects even for this more residentially stable population. So at the end of the day, after several of these different checks to all trying to get at the same thing, all signs point to picking up changes in criminal behavior rather than differential outmigration or moving.

 

Jen [00:41:18] Okay, so that's your paper and your paper has been around for a couple of years now. Is there any other more recent research that's come out and been done since you first read this paper that contributes to our understanding of how local economic shocks affect criminal behavior?

 

Brittany [00:41:33] Yeah, so there's this cool new paper by Aaron and Owens looking at recidivism in fracking counties, which is very consistent with this paper. So, again, even though crime is increasing in these areas, they find the ex-offenders who have a high recidivism rate in general are actually less likely to commit new crime when moving back to fracking counties. This is also related to a body of work by several others in this area in the past few years, finding it this consistent inverse relationship between economic conditions and recidivism. This literature seems to be converging on the importance of economic conditions in specific sectors as being more important when it comes to crime and this is closely related with what we've been talking about, but focuses more specifically on ex-offenders.

 

Jen [00:42:21] That is something that I've been thinking a lot about with the in the current pandemic era and during which we're recording as we've got lots of people coming out of jail and prison and as you said, there's a growing amount of evidence about how the how weak the local labor market is, can help determine the recidivism rate for those folks. And so that that leads us into my next question, which is what your take on the policy implications are both of your findings and the other work in this area.

 

Brittany [00:42:53] Yes. So first, I think it adds to the policy discussion on combating rising crime in fracking areas, there's been a lot of news coverage related to an increase in crime and I think having a better understanding of what is driving this is helpful for policy discussions. Second, it sheds light on policy discussions related to the decline in rural areas and economic stimulus. These results seem to reverse the pattern associated with economics of despair as I showed drops in crim, but the overall crime could increase depending on the type of shock and the potential for large compositional changes with the changing economic conditions. Also, even though I don't find evidence in line with increased criminal opportunities being a driving mechanism in my context, related work by Freanyman and Owens does illustrate the important role of criminal opportunities, particularly if only part of the economy benefits.

 

Jen [00:43:47] And so what's the research frontier here? What are the next big questions in this area that you and others will be thinking about going forward?

 

Brittany [00:43:55] I think the frontier of research related to labor markets and crime is really continuing to focus on the individual, this this is typically difficult because it's very data intensive, but our data keeps getting better and better. And I think, you know, as you know, in the coming years, we'll see more research that's able to pick apart the mechanisms even better.

 

Jen [00:44:15] Yeah, I think I would add to that that, you know, we we have this growing amount of evidence of how important economic opportunity is, but we can't engineer economic shocks, the booms whenever we feel like it. Right. And so figuring out how to translate the benefits that we're seeing when economic opportunity happens to increase into a policy intervention that we can actually control, that is something we we don't know how to do yet and will be interesting to see how what kinds of things people come up with and whether they work.

 

Brittany [00:44:52] Yep. And then something else that I think would be of interest is also looking and thinking ahead to the next paper, looking at how these income effects actually matter for criminal behavior separate from these economic conditions.

 

Brittany [00:45:07] And so that's kind of the other half of this fracking boom, which I don't think has been developed yet. And hopefully if I could increase my sample size, could look at more directly and so this is a slightly different literature, but I think given that we do have this natural experiment that has happening, it does create a possibility to contribute to this literature on how income shocks affect criminal behaviors.

 

Jen [00:45:33] Lots of room for new work in all of these areas. Yes. My guest today has been Brittany Street from the University of Michigan. Brittany, thanks so much for doing this.

 

Brittany [00:45:41] Thank you so much for having me Jen. It was the blast.

 

Jen [00:45:49] You can find links to all the research we discussed today on our website, probablecausation.com. You can also subscribe to the show there or wherever you get your podcasts to make sure you don't miss a single episode. Big thanks to Emergent Ventures for supporting the show. And thanks also to our Patreon subscribers. This show is listener supported. So if you enjoy the podcast, then please consider contributing via Patreon. You can find a link on our website. Our sound engineer is Caroline Hockenbury with production assistance from Elizabeth Pancotti. Our music is by Werner and our logo is designed by Carrie Throckmorton. Thanks for listening and I'll talk to you in two weeks.

Jennifer DoleacEmployment, Resources, Difference-in-Differences