This anonymous essay was submitted to Open Philanthropy's Cause Exploration Prizes contest and published with the author's permission.
If you're seeing this in summer 2022, we'll be posting many submissions in a short period. If you want to stop seeing them so often, apply a filter for the appropriate tag!
I’d like to recommend further support and study for greening blighted vacant lots in urban environments and turning them into public pocket parks. Greening in this case means “removal of trash and debris, grading the land, planting new grass, planting a small number of trees, installing low wooden perimeter fences, and maintaining the newly treated lots.” A research team has been looking into this intervention for the last decade or so, in impoverished US communities like Youngstown, Ohio. The largest implementation that they’ve studied has been in Philadelphia, covering about 4,400 vacant lots. The research team has done a “meta-analysis”, which primarily turned up their own research, and their best estimates are that this relatively inexpensive intervention (at least by US standards — $1600 + $180/yr) reduces nearby gun violence by about 6%. Other research has found significant effects of nearby small green spaces on property values.
A little about me, first, to ground my analysis: while I’ve read a lot of popular work about EA, I’m not an active participant in forums or discussions, so I apologize in advance for not being 100% up on the lingo. I’m going to crunch some numbers in this piece, and I don’t have any formal background in statistics, so fair warning about that, too. Finally, I also live near Philadelphia, where a lot of this research was done, so I may be biased because the research was done locally.
Back to the topic at hand: although there has been a recent uptick of articles, until recently this does not seem to have been a particularly highly-regarded research topic. These articles did not, until recently, have many citations, and I’ve run into several solo-authored papers from graduate students. However, the direction of result of the papers I’ve read point the same way, towards greening vacant lots, or the establishment of “pocket parks” or community gardens, lowering crime rates and creating an economic boost. So, given that the topic seems to be historically somewhat understudied and yet promising, I think there’s room for more analysis.
I think that a close examination of the existing research, followed by direct investment in a poor community designed to create a more conclusive research result, could yield a strong impact. The same communities that are most festooned with vacant lots are also the least able to fund discretionary projects and the most in need of assistance in general, so there’s a case for direct funding to them. In addition, my larger hope would be that establishing that this sort of intervention reduced crime and raised property values to the point of paying for itself several times over would lead to independent adoption of this policy by wealthier municipalities.
Crunching the numbers
Let me start with a quick sanity check. The claim is that greening all vacant lots will provide a 6% reduction in crime. How does that compare to hiring police, which has also been shown to reduce the crime rate? I find an estimate that a 10% increase in police results in a 3% decrease in crime, so I’m going to assume that increasing a police budget by 20% would result in a 6% decrease in crime. For the case of Philadelphia, with its $788 million police budget, that would cost $157.6 million annually. In contrast, maxing out the vacant lot greening program, according to the assumptions in the paper above, would cost $1600 per vacant lot * ~50,000 vacant lots = $80,000,000 initially and then $180 per vacant lot * 50,000 = $9,000,000 a year annually in maintenance; over a 30-year span that’s 7.2% of what it would cost for the police expansion. So at least on an initial pass it’s interesting just from the benefits of crime reduction alone.
However, part of the problem with this research is that the numbers are all over the place. Using one lead researcher’s calculations, the group studying Philadelphia argued in 2013 that the cost effectiveness from reduced gun violence was 968x over 46 months; with a different lead author’s calculations, much more detailed, they argued in their meta-analysis (which, again, mainly re-analyzed their own published work) that the cost effectiveness was .2x , so there’s nearly a factor of 5000 difference in the two calculations. While the calculations in the first paper are somewhat opaque, the most recent paper breaks down their calculations very granularly in an appendix. I’ve discovered what I feel are serious methodological problems with the most recent paper — their model seems to vastly underestimate the number of murders an average city experiences — although I have yet to receive a response from the research team. Based on my reanalysis of that data using actual murder rates, I believe that with their current assumptions there is on the order of 2x cost effectiveness over 30 years from the prevention of gun violence.
Now, this is only the effect for gun violence. They use gun violence because it’s the easiest to get information for and all of the studies they look at have studied it, but I’m going to make an assumption that if greening vacant lots lowers gun violence, it lowers all crime. To be conservative I’ll cut the effect size by 50%. The cost to society of murder also dwarfs the cost of other crimes, so the net change by assuming that all crimes are reduced is about a 20% value add, to make it a total of 2.4x value. (I’m using the crime cost numbers from this study and crime incidence rates as reported here.)
There’s also evidence that proximity to pocket parks and community gardens raises property values in the nearby area. One study finds anywhere from a 3% to 10% rise in property values for properties near a community garden compared to other similar nearby properties, and calculates the additive effect on property value as $2 million per community garden, compared with an outlay of $270,000 to create and maintain the garden over 20 years (7.4x). The study further argues that a city would receive a 3x return on investment in that time frame from investing in community gardens from increased tax revenue.
Recent studies by other research groups analyzing the Philadelphia data have also found a significant effect — a 4-6% boost in property values, but they don’t go into as much detail quantifying the effect. An earlier consultant’s report done on behalf of the City of Philadelphia calculated the effect of a blighted vacant lot at -6.5% to nearby property, suggesting that replacing it with something with a neutral effect would create a result in the range of the earlier estimates.
I will take the NYC community garden estimates as a baseline — which is an 7.4x economic boost to property values and a 3x boost to municipal tax income. While a small flat square of grass is not as appealing as a community garden, and a 5% boost in property value in NYC is worth a lot more than a 5% boost in property value anywhere else, the cost of the greening intervention over 20 years is estimated at $7200, which is about 2.6% of the estimated $220k cost of creating and maintaining the community garden. To quantify other differences, I see estimates that New York City is about 2-3x as expensive as Philadelphia to live in; I’m going to use the high end prediction there and assume Philly property is 1/3 as valuable as NY property. New York is also twice as dense as Philadelphia, so there should be approximately twice as many units near each vacant lot. I’m also arbitrarily going to assume that a greened vacant lot is only 50% as effective in raising property values as a vacant lot turned into a community garden. So 1/3 as valuable property * 1/2 as dense * 1/2 as effective = 1/12 as much effect on property value, for 2.6% the cost, which comes out to 3.2x as cost effective as a NYC community garden, for an estimated value add of 24x to property value and 10x for municipal tax revenue.
There’s also direct value to nearby inhabitants on having a vacant lot turned into a usable quasi-public space. If you’re able to walk to a nearby green space with your dog or your significant other where nothing like that existed before, that’s worthwhile. I’m going to arbitrarily put a number of $10 per resident on the same block per year, and about 70 residents/block, for $700/yr. However, some blocks have multiple vacant lots, and I’m not super confident about this calculation, so I’ll drop it by 50%, for $350/yr, compared to $255/yr for the cost of maintaining the lot, and given uncertainty this is basically another 1x value add.
I also believe, though I’ve been unable to find studies on the topic, that a remediation program like this makes it more likely that the vacant lot in question will be developed. (I’ll do some math here, but I have very low confidence in these numbers.) I’m going to make an assumption, which is that replacing a pocket park with a new building will not have any effect on crime and local property values. The value from creating a pocket park is not that you’re creating a very valuable amenity, it’s that you’re getting rid of a vacant lot; the community gardens study finds that buildings near a vacant lot are valued at 10% or so less than similar buildings that are not near a vacant lot, and other studies find a similar negative effect. I would posit that one non-vacant lot use is pretty similar to another, which makes the key element the value of a new housing unit. In terms of tax value, it looks like the value of land in Philadelphia is typically 1/3 the value of the building on top of it. So, a new building would generate 3x the taxes. Average house cost in Philly is $250000, so let’s say a new house is $400,000, and property taxes are ~1%, making the value to the municipality of redeveloping from property taxes $3k/yr. A study for Philadelphia also finds wage, real estate transfer, and sales tax benefits equal to about 2/3 of the property tax revenue, so that pushes it up to 5k/ year. Value to society is harder for me to estimate; real estate developers seem to shoot for a 15% ROI, so let’s say the value to society is around there, $60k, for a new $400k development. I will estimate that the chance of a lot being redeveloped goes up 1% per year in absolute terms if that lot is greened vs. it being vacant. That is worth .01 * (60000) +.01 * (150000) = $600 (2.7x) to society and $1500 (4.8x) to the municipality over a 30-year horizon.
So far, I find 2.4x cost effectiveness for crime reduction, 24x for nearby property value increase, 1x for added utility for neighborhood residents, and 2.7x for a very speculative increased possibility of development, for my best guess of a ~30x cost-effectiveness for societal benefit and ~15x in revenue to the municipality. Even taking into account the fact that many of the beneficiaries of this would be poorer US residents making less than $50,000 a year — I realize that Open Philanthropy is looking at things it believes have at least a 1000x multiplier, so this is small potatoes. Here’s why I think this still is worthy of consideration:
First, the research data is all over the place on the value of this intervention. I think it’s been moderately well established over a number of papers that crime rate goes down and property values nearby go up with vacant lot interventions, but I honestly don’t trust the value calculations in any of these papers— there’s a factor of 5000 difference from the same research group looking at the same data with a different modeler! So I think a careful analysis (from someone else, I am not a trained statistician) is in order to do a proper job of accounting for costs and benefits; my 90% confidence interval on the true value of this intervention is extremely wide — I might say 5x to 100x; I’m confident that it’s positive, but I’m really really unsure about the magnitude.
However, despite the current mushiness of the data, I feel like this is a tractable problem to study and analyze. A 5% reduction in crime and a 5-10% real estate property value boost, which is what is claimed, are significant and noticeable effects over a multiyear horizon. It should be possible to get a definitive answer as to whether these effects are real and about their general magnitude.
Second, assuming there is a real, strong effect here, I think more publicity, research, and implementation has a high potential for inspiring outside groups to implement this policy on their own, which would leverage the investment. If this really is ~14x as cost effective as police at reducing crime, convincing municipalities to invest in parks instead of police at the margin is quite valuable. As I’ve said before, it’s an attractive idea for municipal leaders: invest in your poorest areas, create a tangible benefit for community members, and reduce crime, all at the same time. Implementation is not complicated, and it’s something that can easily be handled by municipalities using their existing public works crews, or by neighborhood groups or local charities using volunteer labor. Furthermore, depending on the laws in place, the expense of cleaning up nuisance buildings can sometimes be passed along to the building’s owner. So I could see this scaling up fairly rapidly outside of the initial investment if it works out.
Finally, and somewhat cynically: I think that given that the EA movement is concerned right now about its public image, there is a positive potential public relations effect for a significant investment of this kind. I’ll reiterate that the result would be creating hundreds or thousands of pocket parks in impoverished areas. From the perspective of the general public, compared to, for instance, grants to an AI research foundation, this is a more understandable and more viscerally attractive intervention, as well as a camera-friendly one. The initial academic papers have attracted a fair amount of press attention.
Why this all might be wrong
Well, first of all, I’m contradicting the most recent paper by the research group most involved in studying this very topic, which found a much smaller multiplier effect for crime prevention. As I’ve said, I believe I’ve found flaws in their current model, but they’re the ones with PhDs.
Second of all, this seems like a really positive, feel-good sort of investment for municipalities or other outside funders to make. And yet, this is not a common project for them to invest in. So why aren’t they already doing it? Maybe it doesn’t work? Philadelphia still has a lot of crime, despite remediating a significant number of its vacant lots, At a macro level, the remediation hasn’t had a detectable impact.
I’m also putting a lot of weight here on property value effects. Those could be wrong; the research on specifically community gardens and pocket parks / vacant lot remediation is fairly limited. There’s better info on larger parks but it may not apply as directly to something very small. The property boost effect may also be confined to a subset of only the most blighted vacant lots out there, which would limit the scope of a possible intervention — this study indicates that “stable vacant lots”, defined as lots whose owners pay property taxes and presumably maintain them, do not reduce property values.
How to move forward to find out if it’s right
I would envision, in addition to a close reanalysis of the existing data, working in partnership with (at least one!) impoverished city on a large-scale intervention and analysis of the results. As I’ve said, for the entire city of Philadelphia, based on the data here, total costs would be on the order of $75 million + $9 million a year. Extrapolating from that number, for a smaller, poorer city of 200 to 300 thousand people, which might be a better partner, the costs might be more like $15 million initially and $2 million a year in maintenance.
As a sidebar — I admit that I’m mostly thinking in terms of the US here, because that’s where I live and that’s where all the studies I’m reading are from. However, this intervention does not require a significant amount of state capacity to implement. I don’t see any reason why it couldn’t be carried out in poorer countries. The key factors in terms of effectiveness are crimes (particularly murders) prevented, (which I think would be broadly similar in poorer countries as in the US, although less so in areas with a low murder rate), cost of intervention (which is a function of the cost of labor), and value of property (which is also lower in general in poorer countries, but this is an urban intervention, and property costs in cities, especially in large cities, are fairly high.) For a quick back of envelope comparison— in Lagos and Abuja, Nigeria, the average salary seems to be about 1/5 of the US, but the cost of renting an apartment is comparable to US cities (!!!), suggesting that the cost effectiveness of this intervention would be ~5x as high as in the US. If I properly understand your system for weighting the effect of giving more money to low income individuals — which I have just noticed near the due date of this application, so I’m not going to be able to do it justice— incomes being ~1/5 US normal would result in another 5x adjustment to the effect, which gets the overall number near 1000x. However, I’m not at all confident here that my internet research is giving me accurate information, so this is even more speculative than my math elsewhere.