About Publication

Contact Us

C2ER Home


Click to Compare

Purchase COLI
Purchase COLI Back Issues

Licensing Data - COLI Calculator

 


County Cost of Living Index

COLI Adjusted Median
Household Income

 


COLI Release Highlights

COLI in the Media

Review the COLI Methodology

Tips for Interpreting Data

Sample Data (PDF&Excel)

Listing of COLI Geographies

Frequently Asked Questions

Talk with a COLI Expert

Sign up for Free Press Releases

 


Can My Community Participate?

COLI Pricing Help Wanted

 

 

COLI Manual

Helpful Hints in Pricing for COLI

Pricing Survey Form

Register for Data Collection

 

 

2014 Q3 Index: October 27, 2014
2014 Annual Report: January 30, 2015
2015 Q1 Index: May 29, 2015
2015 Q2 Index: August 29, 2015
2015 Q3 Index: October 30, 2015

 

 

2015-Q1: January 8 - 10, 2015
2015-Q2: April 9 - 11, 2015
2015-Q3: July 9 - 11, 2015

 

Site Map Search Contact Us COLI Participants Data Users About Us


County Cost of Living Index

Since 1968, the Council for Community and Economic Research (C2ER) has collected and published cost of living index data at the local level. However, the voluntary nature of the index means that not every area is covered. Increasingly, customers are seeking to better assess the pricing differences for their local area. In response to this demand for more localized data, C2ER has developed a county-level cost of living index for the United States based on an econometric model that identifies key determinants of an area's cost of living.


Estimated County Cost of Living Index Frequently Asked Questions

1.  What are the variables used to estimate the county level index?
2.  What are the economic logics using these variables?
3.  What are the county-equivalent units in this report?
4.  What are the data sources?
5.  What is the regression model that best measures total cost of Living?
6.  How reliable is this data?
7.  Do you have any other related products?


1.  What are the variables used to estimate the county level index?

The key framework for price determination in economics is simple supply and demand. Some factors cause demand to be higher, tending to raise prices; others cause supply to be lower, tending to lower prices. In addition, there are also non-market forces that can cause costs to be higher in one area than in another, creating regional price variations.

Kurre (2000, 2003) examined these issues in the effort to estimate cost of living differentials for all counties in Pennsylvania. As part of that project, Kurre reviewed the literature and tested a number of variables using C2ER's cost of living index data. Enlightened by the results of that study, and with further research and data availability, we identified the following variables as the most important determinants of cost of living in an area:

COLIi = f (POPi, DENSITYi, INCOMEi, GROWTH RATEi, GCOSTi, REGIONi)

where:

  • COLIi: overall cost of living index in area i;
  • POPi: population of area i;
  • DENSITYi: people per square mile of land in area i;
  • INCOMEi: either aggregate personal income or personal income per capita, of residents in area i;
  • GROWTH RATEi: rate of growth of population, income per capita, or aggregate income in area i. One, two, and five year growth rates were tested;
  • GCOSTi: government cost per unit of service (per full time equivalent employee) for all local governments in area i;
  • REGIONi: C2ER defined regions in which area i is located.

The model was developed after testing the significance of any number of variables. Specifically, the model reflects the theory that the overall cost of living in a community is a function of the community's population, population density, income, growth rate, utility rates, efficiency of the government sector, and region of location. The following section provides an overview of why these variables were selected for testing.

2.  What are the economic logics using these variables?

Population
In theory, if there are more people competing for the supply of goods or services, the price will be driven higher. Thus, if two areas are identical in every way except that one has a higher population, we would expect the cost of living to be higher there due to demand factors.

On the other hand, a larger demand may mean that firms producing for local consumption can attain the scale of operations necessary to make use of large-scale production processes which are important in some industries. A larger scale may allow greater specialization in the production process with resulting lower costs per unit. This concept, economies of scale, is very important in some production processes, but certainly not in all. In fact, a larger scale may lead to higher costs in some industries, not lower. This means that an area's distinctive mix of services and industries, and the relative sizes of firms within those entities, can affect its cost of living.

Given the offsetting nature of these two factors, it is not immediately clear which would predominate. Previous research finds that a larger population tends to be associated with higher costs of living. This was our working hypothesis.

Density
Population Density is defined for this model as the number of residents per square mile. Population density is expected to have an effect on cost of living that is distinct from the effect of sheer numbers of people. If two cities each have a million inhabitants, but one has them concentrated into a land area that is only one-fourth of the other's, we may expect that city to have greater congestion and resulting transportation problems, higher land costs, and more environmental issues. As a result, the cost of living for residents in the area with the higher density is expected to be higher.

Income per Capita
Income is expected to affect cost of living in much the same way as population. If two cities have the same population but one has a higher per capita income, the richer city would experience greater demand for most goods, resulting in upward pressure on prices. Of course, the "economies of scale" effect could have the same impact here, as well. However, it is not immediately clear that all prices would be affected equally. Higher incomes may result in greater demand for luxury items rather than necessities so that the price of shoes and canned peas might not be affected as much as the price of champagne and facials.

The impact of these two variables (population and income per capita) can be measured as an aggregate (population multiplied by income of residents in the area). This study looks at the effect in both ways, factoring population and income per capita separately and in the aggregate form. One or the other of these approaches may yield better results, especially if population size and one of the income measures are so highly correlated that it is hard to distinguish between them statistically.

Growth
Aside from size, either in terms of population or disposable income, the rate of change from that size may have an important effect on the area's cost of living. As the growth rate increases, demand also increases, leading to a likely rise in the price of goods. This higher price means a higher profit for suppliers of the product involved in the short run. We would expect this to give current producers an incentive to supply more of the good or service now, and also induce new producers to enter the market in the long run. The increase in the number of producers can lead to an increase in supply which will eventually bring prices back down to previous levels. However, this won't happen overnight for most products. It will take time for the supply response to occur, meaning that living costs may remain higher in an area during a growth spurt than they would after a reasonable period of adjustment. Alternatively, living costs may be higher for areas with recent or more severe increases in population and income than in areas without them.

The time period necessary for adjustment depends on the type of product and the degree of difficulty with barriers-to-entry for new firms. The supply of milk or shampoo in an area can be increased quickly by shipping in more from other areas. On the other hand, an area's housing stock takes a significantly longer time to supply. Since our key goal is to identify a formula allowing us to estimate costs of living for areas, it makes sense to experiment with various time frames and see which seems most closely related. In this study we tested one-, two-, and five-year growth rates for population, income per capita, and aggregate income.

Although this discussion is cast in terms of growth, it applies equally well inversely relative to declines in population or income. Forces that cause demand to decline should exert a negative influence on prices.

Government
Size, type, and quality of services provided by local governments vary dramatically in this country, having both direct and indirect effects on the cost of living in a locality. A government that provides excellent education and efficient garbage collection saves residents the costs of providing similar services out of their own budgets. Similarly, effective police protection and local street maintenance helps keep costs low for local producers, resulting in lower prices for locally produced goods and services.

Measuring local government efficiency involves two components: the amount and quality of services provided, and the cost to local residents in terms of taxes and other charges by local government. Actually measuring these, especially the services provided, is difficult, but the idea is conceptually simple: governments that provide better service for a dollar of cost (or more and better services for the same cost) are more efficient, and contribute to a lower cost of living.

Regional Dummy Variables
Aside from all the factors discussed above, there are others that affect the cost of living in different areas. Although we expected to identify the most important determinants, there are others that we haven't captured, or that affect some areas but not others. One way to try to account for some of these effects is to introduce an individual dummy variable for each region to make our estimating process a little more accurate. Below are the regions and assigned states:

RegionState
1OR
2AZ, CO, ID, MT, NM, NV, UT
3IA, KS, MN, MO, ND, NE, SD
4AR, LA, OK
5AL, KY, MS, TN
6DC, DE, GA, MD, SC
7IN, MI, WI, WV, WY
8NJ, NY, PA
9CT, MA, ME, NH, RI, VT
10AK, CA, HI
11FL
12IL
13NC
14OH
15TX
16VA
17WA

*We used Census region definition and introduced states as a new region where there were at least 10 participating areas in the third quarter of 2013 Cost of Living Index publication.


3.  What are the county-equivalent units in this report?

Since the key goal of this project is to index county-level living costs in the U.S., the county is the basic geographical unit in this report. However, the concept of "county" is not as simple as it seems. For example, in Louisiana they call their counties "parishes", but these are functionally equivalent to counties and therefore we treated them as counties.

While Connecticut and Rhode Island have dissolved their county governments, data are still available for those county areas and we also included them in this report. The same is true for nine counties of Massachusetts.

The District of Columbia is a unique unit that is not really a county. In fact it is often treated as a state when studies do state-level analyses. However, DC has no political subdivisions and therefore we treated it as equivalent to a county for this study.

While counties are the primary administrative/governmental unit below the state level, some states have independent political units, most often cities that are NOT counties or parts of counties. These units exist in four states: Maryland (Baltimore city), Missouri (St. Louis city), Nevada (Carson City), and Virginia (multiple independent cities). Although these are not technically counties, for our purposes we also treated them as equivalent to counties.

The U.S. Bureau of Economic Analysis (BEA) is the primary source for both aggregate and per capita income data in the U.S. Unfortunately the BEA aggregates some of the smaller independent geographical units of Virginia with their surrounding counties. The BEA's variables are very important for our analysis, so it was necessary for us to use the same geographical aggregations that the BEA does. Typically this involved aggregating one or two independent cities with their surrounding county. There are 53 Virginia geographical units involved and these were aggregated into 24 units, for a loss of 29 "county" units. No data were lost or excluded in this process and the sums of the income and population variables for Virginia were the same before and after the aggregation. Simply put, this process aggregated smaller jurisdictions with their neighboring counties and simply reduced the number of geographical units that could be included in the analysis.


4.  What are the data sources?

Most of the independent variables were available for 2009. However, data for "Density" and "Government Cost per Unit of Service" variables were only available for 2007. For dependent variables, we used the 2013 Annual Cost of Living Index (COLI) data since this was the most recent annualized dataset available at the time this study began.

We also used the FIPS (Federal Information Processing Standards) codes to align the variables across counties, and to ensure that all data were for counties rather than other geographical units with similar names.


5.  What is the regression model that best measures total cost of living?

By utilizing ordinary least squares regression analysis, we tested various combinations of the independent variables to identify the best model for use with the data for the 291 areas around the country for which Cost of Living Index data exist. To allow for nonlinear relationships, we also tested squared versions of appropriate independent variables in the model. Criteria for inclusion of a variable included statistical significance (typically at the 10% level or better), intercorrelation with other variables, impact on the adjusted coefficient of determination (R2), and economic logic.

The model that best measures total cost of living is:

COSTOFLIVING = -96.53528 + 0.00000129*Population + 0.0020101*Density + 19.09507*LogofIncomePerCapita + 0.1001465*FiveYearIncomePerCapitaGrowth - 8.403868*Region2 - 13.46978*Region3 - 15.64299*Region4 - 15.39385*Region5 - 9.585289*Region6 - 12.7959*Region7 - 5.191123*Region8 + 7.772021*Region9 + 9.379046*Region10 - 12.11232*Region11 - 12.04831*Region12 - 10.62401*Region13 - 15.36975*Region14 - 16.4688*Region15 - 10.51892*Region16 - 6.27974*Region17

Once the model is calibrated using data for 291 areas, data for the remaining counties were applied to the same model to get their cost of living index estimates. For each county, we plug in the values for those driver variables into the regression equation and get a cost of living estimate for each county.


6.  How reliable is this data?

One way to answer this question is to compare the model's estimates with those areas for which we have actual data, collected on-site — in other words, the 291 areas for which we have actual cost of living data.

The estimated cost of living indexes for these 291 areas have the same average index value, 97.8. The range of the estimates is a bit smaller, from 81 to 173, rather than the range for the actual COLI values which is 82.4 to 178.5. The two sets of numbers have a strong positive correlation; the correlation coefficient is 0.89. This means that those places with a higher COLI value also tend to have higher estimated values. The relationship is not perfect, but it is quite strong.

County cost of living index numbers are estimates, and like any other estimates, they include some error. Error can enter through a number of doors. First of all, the estimates are based on other data and errors in any of those data will, of course, be reflected in the estimates. The regression process tested a large number of explanatory variables, all with their own data. All of these data come from reliable sources, but there are many thousands of individual data points involved and it would not be surprising if a few contained errors of some type.

It is also important to recognize that the regression models estimated here do not account for all of the factors that affect cost of living; no single model could. It is encouraging to note, however, that this model for the overall cost of living accounted for 78% of the variation in cost of among the 291 places in the sample. Despite these cautions, the cost of living estimates presented in this report should serve as useful guides, roughly speaking, to differences in the cost of living across counties. Because of the potential error, however, it is best not to assign great important to small differences in cost of living indexes between places.


7.  Do you have any other related products?

C2ER created two new datasets: Cost of living adjusted median household income and state level cost of living index data.

Median Household Income (MHI) data provided by the U.S. Census Bureau is crucial to understanding the economic conditions of counties across the country. However, to accurately compare counties with higher and lower than average costs of living, these data must be used in conjunction with a cost of living adjustor.

C2ER has also created a state level index based on the estimated county level numbers by using the population weighted index numbers for each county within a state.

Learn more about COLI adjusted Median Household Income data
Purchase The State Level Cost of Living Index