“How does crime rate affect residential real
estate prices in New York City?”
Renata Haseth 10153489
Economics and Finance Universiteit van Amsterdam
Abstract
Crime plays a highly significant role in society due to its extensive costs. This paper focuses on one of those costs, namely the economic cost the increase of crime has on housing prices. By analyzing the changes in real estate prices due to changes in crime in New York City from 1980 to 2012 an attempt is made to measure the correlation between the two. Additionally four boroughs within New York City are analyzed to establish if the effect differs across boroughs with different attributes. For New York City we find that crime does in fact have a negative effect on housing prices. For the individual boroughs however, the coefficients for crime are all insignificant.
1. Introduction
The crime rate in New York City has been dropping since the crack epidemic in the 1980’s and early 1990’s. The year 2013 had 333 homicides, which was the lowest number since at least 1963 when reliable statistics were first kept.
Every crime has direct costs to the victim and indirect costs to society as a whole. Crime also imposes secondary impacts, of which some are harder than others to verify, in particular the wider economic effects that can spread after a crime. Intuitively, there should be a relatively strong link between crime rates and property prices, because nobody wants to invest in a house where they don’t feel safe or that will bring a lot of extra expenses due to vandalism or burglary. Recent developments have made it possible for the public to easily access crime statistics in any city. These websites are actively used by new residents or people interested in moving to a new city. Crime statistics may also influence in which neighborhoods or states people choose to live, run their businesses and raise their families; many potential new residents avoid areas with higher than average crime rates.
Media reports suggest that falling crime has driven New York City’s real estate boom. With the recent change in crime and the fact that people dislike living in crime-ridden neighborhoods, I would like to know if this has indeed had an influence on residential property prices in NYC. This brings me to my research question: “How does crime rate affect residential real estate prices in New York City?”
In order to answer the research question, I ran a regression based on the residential prices in New York City and the crime rate in New York City. I took house price, quality of nearby schooling, the violent crime rate and the non-violent crime rate in the specific neighborhood into account. I also ran separate regressions for four boroughs in New York City to determine if there are substantial differences between the boroughs. The four boroughs that were chosen for detailed study are Manhattan, Bronx, Brooklyn and Queens. For both New York City as for each of the boroughs I ran two regressions. For the first regression I used violent crime as an independent variable and for the second regression I used property crime. By doing this I could answer another
important question, namely what the difference in impact is of the two types of crime.
There are quite some studies done on the effect of crime on real estate prices, both in cities as well as in suburban areas. The studies have been done in many different ways, with different time frames and where some researchers used more variables than others etc. In this paper I tried to combine the most important aspects of different research papers and simplify this as much as possible to get the most straightforward answer. This way I can focus purely on crime and it makes it easier for the public to understand the importance of crime prevention. Of course, simplification has its
disadvantages as well due to the fact that not all aspects can be properly explained. All of the previous research papers I have come across have come to the same main conclusion, which is that crime has a negative effect on real estate prices. This conclusion is very interesting to different groups. First of all policy makers should be aware of the wide spread influence crime has on society. Not only the direct costs of crime matter, but also the indirect costs such as a decrease in property value. Second of all the police department should know how important the work they do is and these studies can increase that awareness and possibly allow the government to increase the budget for crime prevention. This will attract more investors and can help boost the city’s economy, which in turn will decrease crime again. An increase in property prices could have beneficial tax effects for the government as well due to increasing government income from property taxes. Finally this
information is interesting to future homebuyers, property investors, realtors and other people trying to decide where to invest.
The following chapter presents an overview of the literature on both crime and real estate, chapter 3 discusses the methods and data and is split up into research setting, data description and regression analysis. In the fourth chapter I show the STATA output and I discuss the results and the limitations. The fifth and last chapter states the conclusion of the paper.
2. Literature Review
It has long been known that crime has direct negative effects for the victims of the particular crime, but crime can also have large indirect negative effects on the rest of the community, or even beyond that. These secondary impacts are thought to be quite large but they are difficult to estimate accurately. With recent development in technology these secondary impacts have only increased. People have an aversion to crime and thus prefer living in low crime areas. This means that because the demand for houses in
neighborhoods with low crime is higher, the prices will be too. Nowadays it is easier for the public to distinguish between low and high crime areas. Future homebuyers can view the number of crimes, including burglaries, rape,
murder and other violent crimes online on so called “crime maps”. This recent development increases the importance of research of the effect crime has on housing prices. There has been quite some research done on the effect of crime on property prices, although there are some substantial differences in the way the research has been conducted.
In one of the earliest papers on the relationship between property crimes per capita and property values Thaler (1978) finds a negative relationship between the two. He estimated that a one-standard-deviation increase in property crimes reduces home values by about 3 percent. The found estimates could be used to compute optimal levels of criminal justice expenditures in theory. But only in theory, because the production function for the criminal justice system is not known.
In another paper “The Impact of Crime on Urban Residential Property Values” (1978), the authors Hellman and Naroff focused on the effect crime has on a central city with fixed boundaries surrounded by readily accessible suburbs. Because they chose a central city as the focus of analysis they were able to make the Polinsky-Shavell assumption that the urban area is small and open, which means that there is household mobility both within and among different urban areas. Other assumptions are that housing is a
homogenous commodity and that individuals have identical preferences. The model provides the basis for a cost-revenue analysis for city police
expenditures. They also used a minimally accepted crime level as a base for their research (Hellman & Naroff,year, p.105). In their paper they presented a methodology for calculating the property tax revenue lost to a city because of the negative effects crime has on property value. This loss of income can be incorporated in future government policy changes. Although the paper comes to some interesting conclusions, it lacks all the information needed for a completely accurate analysis. Information on police production functions and more refined information on construction of a crime index would have to be acquired. Another problem with their research is that they did not account for unreported crime or the effect of the different types of crime on house prices. Lastly and most importantly Hellman and Naroff have not controlled for characteristics such as plot size or age of the unit.
Haurin and Brasington (1996) study differs from most because of the fact that they based their study on a wide array of jurisdictions in six
metropolitan areas across the US, while most studies focus on one specific area. They wanted to test competing explanations of why house prices vary. They defined a constant-quality house as a unit where structural and land attributes, but not community attributes, are held constant. Using hedonic house price framework they did come to the same conclusions as the other researchers that aside from other community attributes, crime does affect house prices.
One study that differed in outcome is Lynch Rasmussen’s paper (2001). They were one of the first researchers to focus on different types of
crime. Weighing the seriousness of offences by the cost of crime to victims is used as an alternative to the customary measures of the number of index crimes. Their research was based in Jacksonville, Florida. They used a GIS program to develop neighborhood characteristics that are unique to each observation. To derive the weights for each type of crime two sources were used: the National Crime Victimization Survey and research done by Mark Cohen (1990). The study done by Cohen was the first to point out that the type of crime is very important. There are differences between the impact of violent and that of property crime. The outcome of Cohen’s research showed that the costs for violent crime are higher than those for non-violent crime. Lynch and Rasmussen’s regression had a very interesting outcome. It
showed that the number of violent crimes significantly reduced house values, whereas the number of property crimes had a positive and significant impact on the sales price. Thus we can conclude that the distinction between weights of crime is more important than the number of total crime. They explained the positive correlation between property crime and house prices by making the assumption that people living in good neighborhoods are much more likely to report property crime than people living in bad neighborhoods.
Schwartz, Susin and Voicu (2003) focused their research on the time between 1988 and 1998 in New York City. During this period the murder rate fell by 69 percent, the violent crime rate fell by 53 percent, and property crime dropped by 56 percent. To find out how much influence this decrease in crime had on the increase in housing prices, they decomposed trends in New York City’s property values during 1988 to 1998 into components due to crime, the city’s investment in subsidized low-income housing, the quality of public schools, and other factors. They concluded that the decrease in crime was responsible for about one third of the increase in property values. Another notable find in their research is that they found that crime fell and property values rose most sharply in poorer areas, with core Manhattan, the area with the city’s highest income and home prices, as an exception. They also found that New York City’s turnaround was strongest in both richest and poorest areas. The city’s middle-income areas experienced the smallest fall in crime rates, and they were the least affected by the upturn in property values of the late 1990s.
Titra, Petra and Greenbaum (2006), have written a more detailed research paper than most. They used hedonic regressions to make an
attempt at quantifying the intangible cost of crime. They divided crime into two sections, violent crime and property crime. On top of that they focus on what the effect of crime is on people of different wealth. Their research also differs from others because of the fact that they show that it is vital to account for the measurement error that is endemic in reported crime statistics. They address this with an instrumental variable approach. Through the use of hedonic regression they concluded, “crime is capitalized at different rates for poor, middle class and wealthy neighborhoods and that violent crime imparts the
greatest cost.” (p.299).
In 2008 Linden and Rockoff stated “understanding the relationship between property values and local crime risk is useful for measuring the willingness of individuals to pay to reduce their exposure to crime risk. This, in turn, can help determine the appropriate level of public expenditures that reduce crime, such as police services.” (p.1103) They attempted this in their paper “Estimates of the Impact of Crime Risk on Property Values from
Megan’s Laws”. They made improvements on previous papers on the subject by exploiting both the timing of move-in and the exact locations of sex
offenders, we can improve on past estimates of the impact of crime risk on property values. They were the first to exploit both inter temporal and cross-sectional variance in the presence of a sex offender. They found that
individuals have a strong distaste for living in close proximity to a sex offender. A single offender depresses the value of a property substantially, although they were not able to test the two assumptions underlying the victimization cost estimate.
Lochner and Moretti (2001) wrote a paper on the relationship between education and crime; “The Effect of Education on Crime: Evidence from Prison Inmates, Arrests, and Self-Reports.” They used data from the U.S. Census and changes in state compulsory attendance laws. The study calculates social savings from crime reduction associated with high school graduation. The externality is about 14-26 percent of the private return,
suggesting that a significant part of the social return to completing high school comes in the form of externalities from crime reduction. (p.3).
The affect of schooling on house prices has been studied quite a lot. It has been found that the quality of schooling does influence house prices, but there are a few different ways to measure this. For instance, school quality can be approximated by test results or school expenditure. There are two recent papers that each come up with a different answer to the question which measure is the better one.
Although the research has been done quite differently in each paper and some researchers went a step further than others, they all have found a negative impact on housing prices because of an increase in crime. Another good point is that research has been done in different types of cities, from industrialized cities to more suburban areas. This topic has interested researchers for years and as mentioned earlier it has become increasingly important. Although quite some research has been done through the years it’s important to keep reevaluating to keep government officials aware of the importance of crime prevention, not only for the victims directly involved but also for the effects it can have on the housing market and thus the economy as a whole. This review allows stating a hypothesis that a decrease in crime does increase residential property value.
3. Methods and data
In order to answer my research question, I carry out a regression analysis based on the residential prices in New York City and the crime rate in New York City. I use regression analysis to understand how the typical value of house prices changes when any one of the independent variables is varied, while the other independent variable is held fixed. For the New York City regressions I use simple linear regressions and for the individual boroughs I use multiple linear regressions.
3.1. Research setting
New York was chosen because of the clear distinction between
neighborhoods and because of the easily accessible data. New York City is an appropriate city to measure the affect of crime rate on real estate because since 1990 violent crime has drastically decreased. Recently crime has come down to 1950’s levels because of the policies of mayor Bloomberg. The year 2013 had 333 homicides, which was the lowest number since 1963. In New York the 1980’s and early 1990’s crime was high because of the American crack epidemic. Because of the well-known changes in the crime rate a comparison of the different levels will be quite straightforward.
Within New York I will chose a few areas across the city to focus on. The first is in Brooklyn; specifically Bedford-Stuyvesant because this is one of the most drastically changed areas in Brooklyn. Houses fetch prices no one ever thought possible in the days of the crack epidemic of the 1980s.
As a second area I chose to focus on the Bronx, specifically Melrose neighborhood in South Bronx. I picked South Bronx because of the easily accessible data and the proximity to Manhattan, which gives the area a more urban vibe than more northern neighborhoods.
The third area I will analyze is Midtown Manhattan. I picked this area because it has the highest per capita crime rate in New York.
Finally the fourth neighborhood I will focus on is Glendale in Queens. Queens is the largest and second most populous borough in New York.
3.2. Data description
Housing
Although the initial plan was to look at the time frame xxx, after contacting the New York City Department of Finance I found out that unfortunately, prior to 2005 the housing sales data is confidential under the New York City law. Therefore the regressions for the four boroughs have been shortened to now include the years 2005 to 2013. I found the data for the boroughs on
Zillow.com. Figure 1 to 4 show the median sale prices for houses in each borough. I used these graphs to compute the median sale per year for each borough.
Figure 1
Crime
I have collected crime data for New York City as a whole and for the four different boroughs. The New York City crime data has been split up into violent and non-violent crime data for 1980-2012. The data was collected
Figure 3
through the website of the U.S department of Justice. Violent crime includes murder, no negligent manslaughter, forcible rape, robbery and aggravated assault. Non-violent crime includes burglary, larceny theft and motor vehicle theft. For the boroughs I found data on New York City Police Department’s website where records are held in which reported crime and offense data can be found based upon the New York State Penal Law and other New York State laws. The data here only goes back as far as 2000. The data is also separated into two categories, the seven major felony offenses (violent crime) and the seven major non-felony offenses (non violent crime).
Schooling
Based on my findings in the literature review I included schooling as one of the independent variables. For New York City as a whole I decided to use school expenditures. In 2001 Bradbury, Mayer and Case found that increases in school expenditure did increase property values. Federal and state
policymakers agree that education expenditures are an important factor in determining school quality. They frequently argue that more resources should be provided to schools to improve opportunities for students. Many Americans seem to share this view.
From 1980 through 2012 the state expenditure on education as a fraction of GDP has almost remained constant, with just small changes from 4% to 5%. This is sufficient to assume that the change in education quality in New York City has not been substantial enough to have a noticeable effect on house prices.
For the individual boroughs in New York I decided to use ELA results to approximate school quality. For each borough in New York I found the results to the ELA test. The ELA test stands for the English, Language and Arts test. It is taken by grades three to eight in New York. It is a timed test that contains several different types of questions. Students answer multiple choice
questions based on short passages they read, and write responses to open-ended questions based on stories, articles, or poems they listen to or read. Educators examine the results for each school to identify broad instructional areas that need improvement. The results are made public and the data is available on the website of the department of education for New York. I used this data to estimate the quality of the schooling system in each borough and the changes in the quality over time. Unfortunately the data only goes back as far as 2006. Historical data on schooling is very difficult to find.
3.3. Regression analyses
In order to answer the research question, two regression analysis were run. The first set was run on data for New York City as a whole. I ran one
regressions using violent crime and one using property crime as an
independent variable. The second set was run on four individual boroughs in New York City. For each borough I ran two regression, one using the
independent variables violent crime and ELA-results and the other using the independent variables property crime and ELA-results.
Regression A : New York City
The first regression set was run to see if the decrease in crime in New York City is related to the increase in residential real estate prices in New York City. I used data on residential real estate value, more particularly the Freddie Mac house price index for New York. With this data two regressions were run, one using the crime data for violent crime and the second using crime data for the non-violent crimes. These two are run separately to add another important question to my research, which is what the difference in impact is of the two types of crime. It is generally believed the quality of schooling in an area influences house prices in that area. I also take this into consideration. However, for New York City as a whole I find this is not a greatly relevant variable in the analysis. Research has found that expenditures on education are a good indication of the quality of schooling. My statistical analysis has found that the amount spent on schooling has not changed much in the past 30 years in New York. This has lead me to make the assumption that school quality in New York City has stayed relatively constant and that the city has not attracted many new home buyers based solely on an increase in school quality.
Regression B: The boroughs
The second regression set is based on more specific areas. The aim of these is to find out if the correlation between crime and real estate prices differs between boroughs and where this is the strongest. To answer this question I run eight different regressions. For each borough I run two regressions, one using the violent crime data and one using the non-violent crime data, as I did with the New York City regression. As the crime rate goes down I expect to see the prices of residential properties increase.
4. Results
This chapter reports the results of the regression analysis mentioned above. First I will discuss the first set of regressions, which are the ones done for New York City. Then I will discuss the regressions done for the individual boroughs.
Regression A : NYC
Figure 5 and 6 present the results of the regressions run for New York City.
Figure 5 & 6
In figure 5 the independent variable is violent crime. In figure 6 property crime is the independent variable. We first analyze figure 5.
First of all we check the F-statistic for the overall significance of the regression model. Specifically, we check the null hypothesis that all of the regression coefficients are equal to zero. From the STATA output we read that p > F = 0.0000, so we can conclude with extremely high confidence, above 99.99%, that the overall model is indeed significant.
We continue the analysis with the slope coefficients, violent crime. As seen in the table violent crime has a very small effect on residential value (-0.0007221). So for every violent crime the value of the nearby house decreases with 0.07%. The hypothesis that crime has a negative effect on housing prices has not been rejected by this research. From the p-value and the t-value we see that the result is highly significant, the p< 0.000 so our
_cons 177.7548 11.53653 15.41 0.000 154.0411 201.4684 propertycri~l -.0002295 .0000343 -6.69 0.000 -.0003001 -.000159 housingindex Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 54006.2215 27 2000.23043 Root MSE = 27.638 Adj R-squared = 0.6181 Residual 19860.8369 26 763.878344 R-squared = 0.6322 Model 34145.3846 1 34145.3846 Prob > F = 0.0000 F( 1, 26) = 44.70 Source SS df MS Number of obs = 28 . regress housingindex propertycrimetotal
_cons 203.3172 14.08634 14.43 0.000 174.3623 232.2721 violentcrim~l -.0007221 .0001008 -7.16 0.000 -.0009293 -.0005149 housingindex Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 54006.2215 27 2000.23043 Root MSE = 26.432 Adj R-squared = 0.6507 Residual 18165.0962 26 698.657545 R-squared = 0.6636 Model 35841.1254 1 35841.1254 Prob > F = 0.0000 F( 1, 26) = 51.30 Source SS df MS Number of obs = 28 . regress housingindex violentcrimetotal
coefficient is significant at the 99.99+% level. This shows that people are willing to pay more for a house in a safer neighborhood. This is a very
important point for the police as well as the government because this could be used as an extra incentive to create safer living environments. Investors, project developers and future homebuyers should also pay attention to this finding because planned increased law enforcement will raise future house prices.
The constant is 203.31 and is the value of the equation if the slope is zero. It is also highly significant but has little meaning in our model because the slope will hardly ever be zero.
The R-squared is 0.6636, which means that the model explains the variance 66% better than if we were to use the mean value of the house price as a predictor of the dependent variable. This is a pretty high value
considering the small value of the coefficient. This could mean that even though the coefficient is small and thus the change in residential real estate value is small with an increase in violent crime, the change is relatively big considering New York real estate prices are quiet stable. This statement is supported by Abel and Deitz (2010) who found that upstate New York has largely been insulated from this volatility.
We further did the same for non-violent crimes. First we check and see that the overall model is significant as well, with a p>F = 0.0000. We then focus on the independent variable non-violent crime noted here as property crimes. As we can see in the STATA output (Figure 6) the value of the
coefficient is -0.0002295. This means that with every property crime the value of a nearby house decreases with 0.023%. This is less than a decrease in the value of house prices when violent crime increases. This makes sense
because property crime is much easier to prevent than violent crime. Installing an alarm or closed garage could help the problem substantially. It is much more difficult for citizens to prevent violent crime. In addition the social cost of violent crime is larger than that of non-violent crime so the (financial) cost of a house with less violent crime is higher than the cost of a house with less non-violent crime. Finally people tend to be much more afraid of non-violent than of non-violent crime and thus are more willing to pay for a decreased chance of violent crime than for a decreased chance of property crime. The coefficient for property crime is also significant at a 99.99+% level. The r-squared is about equally large as for violent crime, 63%. I would explain this high value of r-squared in a similar way as I did for violent crime, which is that even though the value of the coefficient for property crime is quite small, it is
relatively bigger because residential real estate prices in New York have been quite stable.
Regression B: The boroughs
The second regression set is based on more specific areas. I am interested in finding out whether the correlation between crime and real estate prices differs between boroughs and where this is the strongest. To answer this question I run eight different regressions. For each borough I run two
regressions, one using the violent crime data and one using the non-violent crime data, as I did with the New York City regression. As the crime rate goes down I expect to see the prices of residential properties increase.
Stata output : Violent crime
Figure 7 presents the STATA output for Brooklyn. Figure 8 presents the output for Manhattan, figure 9 presents the output for Bronx and figure 10 presents the output for Queens.
Figure 7 _cons 2043.098 1288.366 1.59 0.188 -1533.98 5620.177 elaresults -1.942548 1.805252 -1.08 0.342 -6.954732 3.069636 majorcrimeb~n -.0146493 .009525 -1.54 0.199 -.041095 .0117965 brooklynhouse Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 2644.35951 6 440.726586 Root MSE = 20.224 Adj R-squared = 0.0719 Residual 1636.10937 4 409.027344 R-squared = 0.3813 Model 1008.25014 2 504.12507 Prob > F = 0.3828 F( 2, 4) = 1.23 Source SS df MS Number of obs = 7 . regress brooklynhouse majorcrimebrooklyn elaresults
Figure 8 Figure 9 Figure 10 . _cons -3729.237 4954.983 -0.75 0.494 -17486.47 10028 elaresults 6.51252 7.333764 0.89 0.425 -13.84927 26.87431 majorcrimem~n .0286072 .0245789 1.16 0.309 -.0396347 .096849 houseprice Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 60836.5597 6 10139.4266 Root MSE = 104.13 Adj R-squared = -0.0694 Residual 43372.857 4 10843.2143 R-squared = 0.2871 Model 17463.7027 2 8731.85135 Prob > F = 0.5083 F( 2, 4) = 0.81 Source SS df MS Number of obs = 7 . regress houseprice majorcrimemanhattan elaresults
_cons 1431.228 568.0484 2.52 0.065 -145.927 3008.384 elaresults -1.602151 .7761879 -2.06 0.108 -3.757194 .5528925 crimebronx -.0024795 .0037205 -0.67 0.542 -.0128093 .0078504 houseprice~x Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 657.886592 6 109.647765 Root MSE = 7.5393 Adj R-squared = 0.4816 Residual 227.366987 4 56.8417466 R-squared = 0.6544 Model 430.519605 2 215.259803 Prob > F = 0.1194 F( 2, 4) = 3.79 Source SS df MS Number of obs = 7 . regress housepricebronx crimebronx elaresults
_cons 6056.809 2608.102 2.32 0.081 -1184.444 13298.06 elaresultsq~s -8.390652 3.686525 -2.28 0.085 -18.62609 1.844783 majorcrimeq~s -.0017782 .00851 -0.21 0.845 -.0254058 .0218494 housepriceq~s Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 6340.94475 6 1056.82412 Root MSE = 11.637 Adj R-squared = 0.8719 Residual 541.700843 4 135.425211 R-squared = 0.9146 Model 5799.24391 2 2899.62195 Prob > F = 0.0073 F( 2, 4) = 21.41 Source SS df MS Number of obs = 7 . regress housepricequeens majorcrimequeens elaresultsqueens
Above is the output for the regressions run on violent crime for each borough. When checking the F-test we find that the regression for Brooklyn, Manhattan and Queens all have a very high “probability > F value”. The f-statistic tests the null hypothesis that all of the regression coefficients are equal to zero. In the first three models the probability that all the independent variables are just random with respect to the dependent variable is pretty high. In the last
regression this probability is low, 0.73% so this model is overall more significant.
The coefficients for violent crime are larger for the boroughs than for New York City as a whole. What is surprising is that for Manhattan the coefficient is positive. While for the other boroughs the coefficients are
negative, Manhattan is an exception. For both an increase in violent crime as well as an increase in ELA results the house prices in Manhattan increase. This can be explained by the fact that Manhattan real estate value is
increasingly rising, no matter what. A paper by Glaeser, Gyourko and Saks (2003) states that Manhattan house prices have been rising since the 1990’s. This is the result of land use restrictions. Because regulation is constraining the supply of housing, the prices will keep increasing.
Another peculiar observation from the results is the fact that the
coefficients for the ELA results for Brooklyn, Queens and Bronx are negative. When we look at the data we see that the ELA results for each borough
increased by very little. So over the time period 2006-2013 the ELA results did not change much and the changes were similar in each borough. This could mean that people did now move based on the ELA results because they all changed very similarly, so an increase in the ELA results did not lead to an increase in housing prices. But did it lead to a decrease? It probably did not either since the coefficients are all insignificant, except for Bronx, which is significant at a 10% level.
The R-squared for Queens is very high, above 90%, which indicates that the data fit the statistical model very well.
Stata output : non-violent/propert crime Figure 11 Figure 12 Figure 13 _cons -666.9435 2418.904 -0.28 0.796 -7382.898 6049.011 nonmajorcri~n .0136596 .0193885 0.70 0.520 -.0401716 .0674907 elaresults 1.515133 3.376668 0.45 0.677 -7.86 10.89027 houseprice Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 2644.35951 6 440.726586 Root MSE = 24.063 Adj R-squared = -0.3139 Residual 2316.19786 4 579.049466 R-squared = 0.1241 Model 328.161649 2 164.080825 Prob > F = 0.7672 F( 2, 4) = 0.28 Source SS df MS Number of obs = 7 . regress houseprice elaresults nonmajorcrimebrooklyn
_cons -12118.04 5409.275 -2.24 0.089 -27136.6 2900.512 elaresults 18.73374 7.87009 2.38 0.076 -3.117133 40.58461 nonmajorcri~n .0791725 .0317062 2.50 0.067 -.0088581 .1672031 housepricem~n Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 60836.5597 6 10139.4266 Root MSE = 75.317 Adj R-squared = 0.4405 Residual 22690.6466 4 5672.66166 R-squared = 0.6270 Model 38145.9131 2 19072.9565 Prob > F = 0.1391 F( 2, 4) = 3.36 Source SS df MS Number of obs = 7 . regress housepricemanhattan nonmajorcrimemanhattan elaresults
_cons 897.4813 500.4807 1.79 0.147 -492.076 2287.039 elascore -.8944986 .7139944 -1.25 0.279 -2.876865 1.087868 nonmajorcrime .0015922 .0031325 0.51 0.638 -.007105 .0102894 housepriceb~x Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 657.886592 6 109.647765 Root MSE = 7.702 Adj R-squared = 0.4590 Residual 237.286119 4 59.3215298 R-squared = 0.6393 Model 420.600472 2 210.300236 Prob > F = 0.1301 F( 2, 4) = 3.55 Source SS df MS Number of obs = 7 . regress housepricebronx nonmajorcrime elascore
Figure 14
The figure 11 to 14 show the regression output for each of the four boroughs for property crime, stated in the output as non-major crime.
The only significant regression is the one for Queens, although the coefficient for property crime is not significant. One finding I want to focus on is the fact that the coefficients for property crime in Manhattan, Brooklyn and Bronx are positive. This means that property crime is positively correlated to house value. Other researchers have found this before as well. It has been explained by Lynch and Rasmussen (2001) who made the assumption that people living in good neighborhoods are much more likely to report property crime than people living in bad neighborhoods. Another possible explanation is that people who buy more expensive houses don’t mind property crime in the neighborhood that much because they can afford better security systems.
The effect of property crime on house prices is smaller than the effect of violent crime on house prices. This shows that people fear violent crime much more than property crime, which makes sense. The R-squared for the regressions are all also quite high, except for the Brooklyn regression, where the R-squared is 12,4%.
Limitations
There are several limitations with the research. First of all we need to be aware of the fact that there are very few variables used in this research. This is partly due to the shortage of data available, but also because it is nearly impossible to account for all the factors that could affect house prices. If we take the first regression, violent crime for New York as an example, we see that the R-squared value is about 66%. So in this case 34% of the total affect on housing prices is unaccounted for. Government regulation, interest rates, demographics, speculation and many more factors affect residential real estate value. A second limitation is that the research is only done for reported crime. There is also a lot of crime that goes unreported, especially vandalism
_cons 5606.244 1612.716 3.48 0.025 1128.625 10083.86 elaresultsq~s -7.753609 2.226813 -3.48 0.025 -13.93623 -1.570984 nonmajorcri~s -.0009219 .0186641 -0.05 0.963 -.0527417 .050898 housepriceq~s Coef. Std. Err. t P>|t| [95% Conf. Interval] Total 6340.94475 6 1056.82412 Root MSE = 11.697 Adj R-squared = 0.8705 Residual 547.279933 4 136.819983 R-squared = 0.9137 Model 5793.66482 2 2896.83241 Prob > F = 0.0074 F( 2, 4) = 21.17 Source SS df MS Number of obs = 7 . regress housepricequeens nonmajorcrimequeens elaresultsqueens
and crimes of that sort. This can also have an effect on the way people view a certain neighborhood.
Most of the borough regressions are insignificant. For the boroughs I used two variables, ELA-results and the crime rate. The New York
regressions that used just one variable are significant. This could have something to do with multicollinearity. Multicollinearity occurs when multiple predictor variables in a model are highly correlated. This increases the standard errors of the coefficients. Increased standard errors in turn mean that coefficients for some independent variables may be found not to be significantly different from zero. So in terms of my research this means that ELA-results and crime are correlated. This does make sense since people with better education are much less likely to commit crime. Lochner and Moretti (2001) estimated the effect of education on participation in crime and incarceration and found that a significant part of the social return to
completing high school comes in the form of externalities from crime reduction.
5. Conclusions
In summary, the crime rate in New York has been dropping since the 1980’s and early 1990’s. Crime has the obvious effects for the direct victims but it also has less obvious secondary effects. Intuitively, there should be a
relatively strong link between crime rates and property price, because nobody wants to invest in a house where they don’t feel safe or that will bring a lot of extra expenses due to vandalism or burglary. This leads to the research question: “How does crime rate affect residential real estate prices in New York City?”
I ran ten regressions to answer this question, two on New York City data and eight using data for four boroughs in New York City, Brooklyn, Manhattan, Bronx and Queens.
The regressions run for New York City gave us the expected result, the effect crime has on house prices is indeed negative. All of the literature I reviewed came to this result as well. The result from the regressions done for the boroughs varied a bit more. None of the coefficients for violent crime were significant, although the ones for Brooklyn, Bronx and Queens were negative. The positive coefficient for violent crime in Manhattan was explained by the high demand and low supply for Manhattan housing.
When we ran the regression using the independent variable property crime we observed big differences between property crime in New York City and property crime in the individual boroughs. First I have to state that since the coefficients for property crime for the boroughs were all insignificant, the variable property crime has no effect. That said, the coefficient for property crime in New York is negative, which means that as property crime increases house prices decrease and vice versa. On the contrary the coefficient for property crime in three of the four boroughs is positive. So for those boroughs an increase in property crime is related to an increase in housing prices. As
mentioned in the literature review, other researchers have observed this as well. The positive correlation is explained by the two facts. People living in good neighborhoods are much more likely to report property crime than people living in bad neighborhoods. On top of that people who buy more expensive houses don’t mind property crime in the neighborhood that much because they can afford better security systems.
With this research we have proven that, for New York as a whole at least, a decrease in crime increases home values. Policy makers should take this into consideration. This research encourages the budget for the NYPD to be increased, because according to the output this will increase wealth. Through the increase in housing prices, property tax income increases for the government and investment is stimulated. Future homebuyers should also be aware of this fact since buying a house in a neighborhood.
The coefficients for crime in the regressions for the boroughs were all insignificant. So unfortunately there was no statistically significant linear
dependence of crime on house value detected in the individual boroughs. The topic is very interesting for future research. My research has quite a few limitation and getting rid of these could improve results and teach us more about the effect crime has on real estate value. It is clear that quite some additional research has to be done to completely understand the relationship between house prices and crime. Future researches should focus on
combining what is already known from previous research and incorporating new data and more elaborate data to obtain new knowledge on the subject.
Literature
Abel, J. R., & Deitz, R. (2010). Bypassing the bust: the stability of upstate New York's housing markets during the recession. Current Issues in Economics and Finance, (Mar).
Bradbury, K. L., Mayer, C. J., & Case, K. E. (2001). Property tax limits, local fiscal behavior, and property values: Evidence from Massachusetts under Proposition 212. Journal of Public Economics, 80(2), 287-311.
Cohen, Mark. “A Note on the Costs of Crime to Victims.” (1990) Urban Studies volume 27
Glaeser, E. L., Gyourko, J., & Saks, R. (2003). Why is Manhattan so
expensive? Regulation and the rise in house prices (No. w10124). National Bureau of Economic Research.
Grosskopf, S., Hayes, K. J., Taylor, L. L., & Weber, W. L. (1999). Anticipating the consequences of school reform: a new use of DEA. Management
Science,45(4), 608-620.
Haurin, D. R., & Brasington, D. (1996). School quality and real house prices: Inter-and intrametropolitan effects. Journal of Housing Economics, 5(4), 351-368.
Hawley, A. H. (1956). Estimating aggregate residential real estate values from population income data. Land Economics
Hellman, D. A., & Naroff, J. L. (1979). The impact of crime on urban residential property values. Urban Studies, 16(1), 105-112.
Kelling, G. L., & Bratton, W. J. (1998). Declining crime rates: Insiders' views of the New York City story. Journal of Criminal Law and Criminology, 1217-1232. Leonard, V. A. (1952). Effect of Crime on Real Estate Values. The journal of criminal law, criminology, and police science
Linden, L., & Rockoff, J. E. (2008). Estimates of the impact of crime risk on property values from Megan's laws. The American Economic Review, 1103-1127.
Lochner, L., & Moretti, E. (2001). The effect of education on crime: Evidence from prison inmates, arrests, and self-reports (No. w8605). National Bureau of Economic Research.
Lynch, A. K., & Rasmussen, D. W. (2001). Measuring the impact of crime on house prices. Applied Economics, 33(15), 1981-1989.
Schwartz, A. E., Susin, S., & Voicu, I. (2003). Has falling crime driven New York City's real estate boom?. Journal of Housing Research, 14(1), 101-136. Thaler, R. (1978). A note on the value of crime control: evidence from the property market. Journal of Urban Economics, 5(1), 137-145.
Tita, G. E., Petras, T. L., & Greenbaum, R. T. (2006). Crime and residential choice: a neighborhood level analysis of the impact of crime on housing prices.Journal of Quantitative Criminology, 22(4), 299-317.
