Crime & Home Prices – 50 Largest US Metros

Crime & Home Prices – 50 Largest US Metros


This report looks at the correlation between total crime (violent crime and property crime) and recent home prices in the 50 largest metro areas in the United States.  The study uses metropolitan statistical area (MSA) level data and zip code level data to minimize inconsistency with respect to inclusion or exclusion of suburbs of the ranked towns – a major flaw of city limit (“city”) rankings.  This study looks at data from full metro areas, metro suburbs, and metro urban areas.  For this study, the urban zone is the defined as the inner 20% of the metro area by population.  Suburbs are defined as the outer 80% of the metro area by population.

Urban experts warn that it is a common blunder in social science to assume that correlations imply cause-and-effect.  Any correlations could come from a common source, rather than cause-and-effect.  Still, we recognize that real estate sites feature crime as a major statistic they say people should consider when deciding where to live.  And popular choices on where people choose to live bid up home values.

The crime and home price data for this study comes from three sources: 2015 FBI Uniform Crime Reporting Table 6 for full metro crime data, the website for zip code level crime index data, and for zip code level average home prices and metro area average home prices.  (Note: just recently suspended inclusion of crime data on their web site for unknown reasons.)  US Census Bureau data is used to identify MSA zip code area latitudes and longitudes.  For this study, urban data is computed from the zip code zones closest to the metro area core-city City Hall and whose combined populations make up the inner 20% of the metro population.  Suburban data is computed from remaining zip codes to the edge of the metro area as defined by the Census Bureau and OMB MSA definitions.

The urban crime index was computed using the formula by combining zip code crime indexes, weighted by population, from the inner 20% of each metro area.  Suburb crime index was computed from FBI Table 6 full MSA data, and the urban crime index from above using the formula: Suburban Crime Index = (MSA Crime Index – 0.2 * Urban Crime Index) / 0.8.  This indirect approach provides the suburb crime index that, when combined proportionally with the urban index, yields the FBI Table 6 full metro index.  (For any blank entries in 2015 FBI Table 6, data from the most recent years was inserted.)


Full MSA Rankings and Scatter Plot

First consider full metro area data.  The three charts below show crime and property value data for full metro areas.  The Total Crime indexes are computed using the formulation from FBI Table 6 data.  The formula adds the crime counts of the seven FBI crime categories and normalizes to the US crime average pegged at 100.  The FBI published a crime index with this formulation until 2004.  A crime index of 200 means the area has double the US average total crime.  50 means half the US average.  The points on the scatter plot are connected with lines (left to right) just as a visual aid.

Figure 1 Full Metro Crime Rank

Figure 1.  Full Metro Areas Total Crime Index compared to US Average of 100.

 Figure 2 Full Metro Home Values

Figure 2.  Full Metro Areas Average Home Values

 Figure 3 Full Metro Scatter Plot

Figure 3.  Scatter Plot.  Full Metro Areas Average Home Values vs Full Metro Crime Index


Full Metro Data Observation

At the full metro level, there is little association between home values and crime looking at the trendline and correlation between home values and crime.  Decisions on where to live, metro to metro, are apparently not driven by overall metro crime levels.

This could be because the public does not believe crime is a major factor they should consider when seeking to move to a new metro area.  Or it could be that the public is not aware of metro area crime data, compared to widely publicized “city limit” crime rankings.  The public may be equating city crime rankings with metro rankings in their minds without realizing that there are significant ranking differences between the two.




Urban and Suburbs Rankings and Scatter Plots

The following plots compute crime and home price numbers separately for metro inner urban cores (inner 20% by population), and suburbs (outer 80% by population).

Figure 4 Urban Crime Rank

Figure 4.  Metro Urban Areas Total Crime Index compared to US Average of 100.

Figure 5 Suburban Crime Rank

Figure 5.  Metro Suburban Total Crime Index compared to US Average of 100.

Figure 6 Urban Home Values

Figure 6. Metro Urban Area Average Home Values

Figure 7 Suburban Home Values

Figure 7.  Metro Suburban Average Home Values

Figure 8 Urban Crime - Urban Home Values Scatter Plot

Figure 8.  Scatter Plot. Urban Crime Index vs Average Urban Home Values

Figure 9 Urban Crime - Suburban Home Values Scatter Plot

Figure 9.  Scatter Plot. Urban Crime Index vs Average Suburban Home Values

Figure 10 Suburban Crime - Suburban Home Values Scatter Plot

Figure 10.  Scatter Plot. Suburban Crime Index vs Average Suburban Home Values

Figure 11 Suburban Crime - Urban Home Values Scatter Plot

Figure 11.  Scatter Plot. Suburban Crime Index vs Average Urban Home Values


Urban / Suburban Data Observation.

Here are three observations from the crime and home value data from data split into urban / suburban components for each of the 50 largest metros in the United States.

  1. While there is little correlation between crime in the suburbs and property values in either the suburbs or the urban areas, the same is not true for urban crime. As we might expect, metros with high crime in urban areas has some correlation with low home values in urban areas.

But what is most pronounced in the data is the correlation between urban crime and suburban home values.  Metros with high urban crime indexes correlate even more with low suburban property values, in the top 50 metros, than high urban crime rates correlate with low urban property values.

While not claiming one causes the other, one could reasonably project that a metro with high urban crime will likely have diminished suburban home values, in spite of the safety of the suburbs in that metro.

  1. Another interesting observation is that many metros with poor crime rankings on the urban crime list, such as Detroit and St. Louis, rank well on the suburban crime rankings. This should not be too surprising since suburban crime would have to be very low in those cities to get their full metro crime averages down to the middle of the pack positions they hold in the full metro crime rankings created from FBI Table 6.

One explanation for the big difference for urban vs suburban rankings for metros Detroit and St. Louis may be geography.  Both cities were founded against major mobility barriers – Canada/Detroit River for Detroit, and the Mississippi River/Illinois for St. Louis.  Their core city halls are almost on peninsulas causing extreme metro sprawl away from downtown.  Downtown core for both metros is far from the current metro population centers, and their downtowns are no longer major crossroads for traffic spanning the metro area, leading to urban core decline.  Their downtowns form part of the outer edge of the city boundary elbow, rather than the inner edge of the elbow, such as Chicago or Boston.

  1. “City” crime rankings (using core city limits only) often line up with the urban ranking shown here for many older Eastern cities, since those cites usually encompass only a fraction of their metro areas with the oldest housing in the metro core. When boundaries are defined consistently using Census and OMB defined MSA rules based on population, instead of politically determined city limits, more Western towns fare poorly in crime rankings.


Crime Ranking: St. Louis vs. Kansas City

Crime Ranking: St. Louis vs. Kansas City

Midwest Mayors complain that “city” crime rankings are not accurate because cities are defined by city limits, which are determined by politics, not consistent rules of population statistics.  Rust belt cities typically have city limits locked in place encircling a small old inner portion of the metropolitan area and containing few if any low-crime suburbs.  Newer Western cities, by contrast, cast their city limits far out into farm fields where they encircle the majority of their metro low crime suburbs, which dilutes their average crime rates.


But researchers put both types of cities into the same “city” ranking, and then declare the cities with most crime per resident as the most dangerous.  Since the general public associate cities with entire regions, “city limit” rankings can unfairly paint an entire region as crime riddled while masking growing crime issues in so-called “hot” younger cities.  These rankings tell you almost nothing about personal danger, since any city can change its ranking just by moving a boundary without actually lowering crime.

By contrast, crime rankings based on Metropolitan Statistical Area (MSA) boundaries, instead of city limits, make use of city definitions set consistently, metro to metro, by the Federal government using statistical rules based on population.  For some reason, these more valid MSA rankings are ignored by the media. In 2012, Forbes Magazine switched from MSA crime ranking to a city limits crime ranking, with scant rationale, saying “We used cities instead of larger metropolitan statistical areas, which gave the disadvantage to older cities with tighter boundaries.”

MSAs usually have an inner business cores at their centers, older smaller homes and multi-family homes further out, and suburbs beyond that.  I contend that statisticians could do a much better job of comparing major cities by going down to the zip code level and identifying zip codes in the inner 10%, 20%, 30% etc. of their MSAs for crime statistics.  Then one could compare the inner 10% core of the Pittsburgh metro area with the inner 10% core of the Houston metro area if one was planning to live near downtown.  Or compare the 50% population ring of two metros for folks comparing suburbs.  But that takes some work, and most crime rankers just paste FBI tables into a spreadsheet, combine crime categories into a single score for each city, and then sort on that score.  This is something almost anyone could do in an afternoon.

I decided to take my own advice and see how hard it would be to go onto the internet and address just two cities using the percent of population rings approach to compare crime rates.  I chose to compare St. Louis and Kansas City.  St. Louis is the last old Eastern City as you go West, and KC could be seen as the first Western style city.  In the free 2014 CQ Press Cities Crime Ranking, St. Louis ranked at #5 worst for crime while Kansas City ranked better at #61.  But in the 2014 CQ Press Metro Ranking, the orders were reversed with Kansas City ranking worse at #52, while St. Louis ranked safer at #95.  So I was anxious to see how the plots would show a transition as the data extended further from City Hall.


Here are the steps I used to plot St. Louis and Kansas City crime.  The same steps and data sources (links at the end of the piece) work for all metro areas.

Method for Average Crime index for percent population rings from City Hall.  10%, 20%, etc.

  1. Find zip codes for each Metropolitan Statistical Area (MSA)
  2. Find LAT LONG of City Hall of the primary city of each MSA and each MSA Zip Code area
  3. Compute distance of each zip code from city hall in miles with LAN LONG to mile conversion.
  4. Get Crime Index for each zip code
  5. Get population for each zip code
  6. Determine distance rings containing 10% of the population, 20%, etc.
  7. Identify specific zip codes within each ring
  8. Compute total crime index for each % ring using zip code crime index weighted by population.
  9. Plot crime index for the 10% population ring, 20% ring, etc. as a histogram.

I was able to find free databases online for each of the steps in this approach, but it was a bit tedious copying crime indexes by zip code from one of various neighborhood data realty sites and pasting the indexes into my spreadsheet one at a time.  Professional researchers could probably purchase the entire crime-by-zip-code database in XL format to make that part a lot easier.

I computed the distances from City Hall for 10% of the metro populations, 20%, 30%, etc. at these distances:

STLKCpercents_milesTable 1.  Distance from City Hall where 10%, 20%, etc. of the metro population live.

I realized that since St. Louis and Kansas City metro areas are fairly similar, it would be interesting and pretty easy to go through step 4 above and just plot zip code crime indexes for each zip code as a function of distance from city hall.  The data include zip code areas in Illinois for St. Louis, and Kansas for Kansas City as well as Missouri zip codes.


Here is the scatter plot of zip code crime indexes vs. distance from City Hall for St. Louis and Kansas City.  US average crime index is 100.

STL v KC Zip Crime by Distance.JPG

Figure 1.  Zip Code Crime Index by miles from City Hall for St. Louis and Kansas City Metros.

The website posting the crime index for each zip code said the index is a combination of rape, murder, assault, robbery, burglary, larceny and vehicle theft normalized against a US average score of 100.  So 200 means twice the average US crime.  The counts of each crime category are unweighted according to the web site, so murder counts the same as robbery.

Continuing with the remaining steps to get a histogram of crime indexes by percent of metro area rings, I combine crime indexes within each percent ring.  For this, I weighted the indexes by zip code population, so a zip code with just 3 people would contribute proportionally less than one with 1,000 people within a percent ring.  I collected data up through the 60% ring.  For the full metro numbers, I computed the St Louis and Kansas City crime indexes directly from the FBI data tables.

Here is the crime index histogram for each percent ring of population out to 60% from City Hall.

STL v KC Pop Rings Crime

Figure 2.  Crime Index by 10% rings of population out from City Hall.

And here is what the data looks at the 10% core, the entire inner 50% of the metro population and the full metro.

STL v KC Crime 10 50 100

Figure 3.  Crime Index for 10% core, the entire inner 50% of the metro population, and the full metro.

Since St. Louis and Kansas City are similar in size, the histogram information roughly lines up with the distance scatter plot.  If I was comparing St. Louis to a much smaller or larger metro, the percentage histogram would be more useful.

Here are maps of St. Louis and Kansas City with the Crime Index shown as a number from 1 to 6, where 6 represents crime 6 times the national average.  The links below the maps go to short videos of each map in a circling motion to see around the data pillars.

St Louis Crime Index by Zip

Figure 4.  St. Louis Crime Index by Zip Code.  US average is 1.


Kansas City Crime Index by Zip

Figure 5.  Kansas City Crime Index by Zip Code.  US average is 1.



I was surprised at how different the plots turned out between the two cities.  As expected, core areas of both metros have higher crime rates, and suburbs have lower crime rates.  Since the full St. Louis metro crime rate is lower than the full Kansas City metro crime rate, I was guessing that the two metros were similar enough in configuration that St. Louis would come out slightly safer at every percent of population and distance out from City Hall.  Instead I learned that the inner 20% of St. Louis zip codes had around a 20% higher crime index than their Kansas City counterparts.  I was even more surprised to see how much safer St. Louis inner suburbs are than their Kansas City counterparts. The Kansas City crime index was around 40% higher than St. Louis for the 30% through 60% population rings.  The higher suburban crime in Kansas City more than makes up for the higher inner core crime in St Louis to account for the overall higher crime rate in the entire Kansas City metro area.

The crimes per person may be higher in St. Louis inner core because the number of people living there (the denominator) has plummeted over the last 70 years until recently, while the number of people working, driving through, and doing business during the day is still pretty high.  But crime indexes always divide only by the resident count, not the visitor count.  I suspect these patterns may be typical for older rust belt cities where the middle class has moved to larger modern homes in the suburbs long ago, whereas Western cities still have many newer homes close to the central core.  Some cities like St. Louis and Detroit have an additional factor pulling residents westward – a central business district built almost out on a peninsula up against a major barrier – the Mississippi River for St. Louis and the Canadian border for Detroit.


If all the City and Metro crime rankings were replaced with charts like these, planners could make better decisions about the status of crime in major cities.  This approach completely eliminates the city limits as a factor driving a false ranking.  Planners can better see how their metro area stacks up against other metros at similar distance rings when assigning resources to fight crime.  The next step would be to go down to zip code level directly to address specific crime problems within the metro areas.  If publishers must have crime rankings to sell magazines, the full MSA boundary ranking, or the 50% inward stats are more representative of relative crime rates.


Links to data sources:

Zip Codes that make up each MSA:

LAT LONG of each metro – City Hall

Zip Code Area LAT LONGS

Convert difference between two LAT LONGs to statute miles

=ACOS(COS(RADIANS(90-A2)) *COS(RADIANS(90-A3)) +SIN(RADIANS(90-A2)) *SIN(RADIANS(90-A3)) *COS(RADIANS(B2-B3))) *3958.756

Crime rating and population size for each zip code

Population distance spread from City Hall

Zip code map images

FBI Table 6 to get Full Metro Stats to compute full metro counts 2013

FBI Table 1 to get Full US Stats to scale computer full metro crime Indexes 2013

Total Crime Risk Index Used by description:

Total Crime Risk – A score that represents the combined risks of rape, murder, assault, robbery, burglary, larceny and vehicle theft compared to the national average of 100. A score of 200 indicates twice the national average total crime risk, while 50 indicates half the national risk. The different types of crime are given equal weight in this score, so murder, for example, does not count more than vehicle theft. Scores are based on demographic and geographic analyses of crime over seven years.

CQ Press 2014 (2013 data) rankings of safest cities and safest metro areas.

Forbes Most Dangerous Cities

2011 When Forbes used the MSA Ranking

2012 When Forbes switched to the City Limits Ranking

Gary Kreie is a recently retired missile software engineer/manager.