Determining the co$t of vacancies in Baltimore.

[ILLUSTRATION OMITTED]

As the number of home foreclosures rises nationwide, cities face mounting costs related to vacant and abandoned properties. Aside from subsidizing these costs through general funds, fees and fines, cities are suing mortgage financing companies and banks to recover costs, charging that irresponsible lending practices have led to increased foreclosures and vacancies. However, with any of these measures, it is necessary to demonstrate that there are specific costs associated with vacancies.

STATISTICAL ANALYSIS

In 2008, the city of Baltimore, Maryland, used econometric analysis to rigorously quantify the cost of police and fire services associated with vacant properties. Many studies have addressed the cost of vacant properties in a general sense, but Baltimore was the first city to undertake a detailed study of this nature. Working with a consultant, the city determined that the cost per block of police and fire services showed an annual increase of $1,472 for each vacant and unsafe property on that block (vacant and unsafe is a technical designation used by the City of Baltimore Housing and Community Development Department).

The cost of the extra safety effort was determined based on 911 call data, a list of properties designated as vacant and unsafe for at least one month, and police and fire department budgets, all for fiscal 2008. The call data and vacant property information was used to construct an econometric model that measured the statistical relationship between the number of properties designated as vacant and unsafe on a block and the amount of time police and fire personnel spent on that block. In the analysis, "officer minutes" and "fire fighter minutes" were the base measures used, indicating one minute of time for one officer called to a specific block. For police officers, the officer minutes were calculated from the time an officer was dispatched until he or she left the scene, multiplied by the number of officers who responded. For fire fighters, the number of staff responding was not broken out in the data, so the Fire Department provided an estimated number of fire fighters who responded, based on the type of call. The econometric model used zip codes as a proxy control for geographic differences in public safety costs (i.e., areas with more or less crime). The analysis determined a positive relationship between vacant and unsafe properties and the officer and fire fighter minutes spent on a given block.

To calculate the cost of one officer or fire fighter minute, the portion of police and fire department budgets dedicated to responding to 911 calls--known as the response budgets--was divided by the total police and fire response minutes. These response budgets included most, but not all, of the departments' activities. They excluded, for example, police special investigative teams and the fire prevention inspectors.

This cost per officer or fire fighter minute was multiplied by the marginal increase in public safety response minutes associated with each additional vacant and unsafe designated property on a given block, as calculated by the model. The results of this analysis are shown in Exhibit 1. The analysis indicates a correlation between vacant and unsafe properties and increased police and fire cost, and it measures the size of that increase. But the analysis does not address whether the vacant properties caused the increase in cost or whether increased criminal activity might have caused an increase in vacancies. Further analysis would be needed to answer the causality question.

The statistical results for police and fire fighter minutes yielded by the econometric model are shown in Exhibit 2. The coefficients are the incremental number of minutes spent for each vacant and unsafe property by block for police officers or fire fighters, respectively The coefficients are positive, as one might predict (i.e., more time is spent servicing blocks with vacant and unsafe properties than those without), and they are statistically significant. That is, there is a greater than 99 percent chance that these coefficients are not zero. The [R.sup.2] value indicates the amount of variation (in officer minutes) this equation explains. While vacant properties and zip codes can explain 8.4 percent of the variation of officer time spent on a specific block, 91.6 percent is explained by other variables such as poverty, density, transience, etc.

EVERY ADDRESS IN BALTIMORE

Conducting the analysis required four city departments to work together. The Department of Housing and Community Development (HCD) established which properties were vacant and unsafe. The police and fire departments provided specific call data, and the Department of Finance provided budget information. HCD also provided the real property file to determine the universe of addresses and blocks within the city. Each of the four departments assigned staff to a steering committee that guided the analysis.

Over the 12 months of data, there were approximately 19,000 properties designated as vacant and unsafe for at least one month, out of 236,569 properties in the real property file. Of the addresses in the real property file, those that had descriptive names rather than address identifiers were removed, leaving in 229,020 addresses as the universe of properties in Baltimore.

Unfortunately, while the police and fire data were input in the same format, the real property file used slightly different naming conventions. The following methodology was developed to rectify these differences and match real property addresses to call data. Four pieces of the street address were used to create a unique street address identifier: street name, house number, street type, and street direction (where applicable). The resulting address code is a unique 11-digit number. The first piece of the address code is the name of the street where the property is located. The consultant created a unique four-digit identifier for each of the approximately 2,500 street names in the real property file. The numbers were assigned to streets in alphabetical order, starting at 1000. The next digit of the address code was for the street direction. A unique number was assigned to each cardinal direction (i.e., North, South, East, and West), and for streets without a direction. For every street type (e.g., street, road, lane) in the real property file, a two-digit value was assigned. Street names that did not have a street type (e.g., Broadway) were assigned a value as well. The final piece of the address code was the house number. No house numbers were longer than four digits, so each one was converted to a four-digit number (e.g., 300 would become 0300). In the real property file, some of the addresses contained a range of house numbers; for the purposes of this analysis, the first house number in the range was used.

As an example, the address code for the Baltimore Police Department headquarters, located at 601 E. Fayette St., is 18491350601. The first portion of the code is for the street name, Fayette, which was assigned the number 1849. The next number, 1, is the numeral assigned to the direction of East. The value of 35 corresponds to the street type, in this case "street." The final portion is the four digit number for the house number, 0601.

Using this methodology, an 11-digit number was created for each of the records in the real property file. In addition to the longer address code created for the specific address, a block-level address was created as well. The block-level code is a nine-digit number created by removing the last two digits in the house number. The same methodology was used to create the address code and block-level code for each record in the police and fire call data. There were 12,824 blocks identified in the City of Baltimore real property file.

The police provided 1,103,091 call records, of which approximately 143,000 were designated as administrative (i.e., lunch, transportation, etc.) and thus removed from analysis. Originally, the analysis tried to match specific addresses to specific calls, but the Police Department does not always use the exact address and at times uses what it calls an "implied" address. That is the first two digits of the address number with either an odd or even designation to indicate the specific side of the block. To overcome this hurdle, the city's budget director suggested conducting the analysis on a block level rather than a specific address level. Due to data irregularities, however, not all records could be matched to a block. Once all the quality checking was complete, there were 870,201 police records, or about 90 percent, that could be matched to a specific block in the real property file.

A similar process was conducted on the fire data (although there were no administrative codes in the fire data). The original file contained 158,728 records, and once the quality of all the data was checked, 122,275 records remained, or approximately 80 percent.

[ILLUSTRATION OMITTED]

CONCLUSIONS

While a great deal of literature generally estimates or hypothesizes the cost of police services associated with vacant properties, the Baltimore study represents the first time direct public-safety costs have been rigorously measured with statistical validation. Using this analysis, the city is armed with financial information that will allow it to pursue additional action in fighting problems caused by vacant properties. City staff are currently determining what type of program to use--an additional fee, fine (citation), or both to help defray the costs of vacant and unsafe properties. Whichever path it pursues, the city now has defensible data it can use to hold owners who allow their property to become vacant and unsafe financially accountable.