*Shadi Eskaf is a senior project director for the Environmental Finance Center at the University of North Carolina at Chapel Hill.*

*“What is the [national/state/recommended] threshold of affordable rates? Is it 2.5 percent MHI?”*

If I had a dollar for every time I get asked this question, I don’t think I’d have to worry about affording my own water and wastewater bill. Percent MHI has become a popular indicator for utilities, agencies, and organizations across the country, and even we use it in our Rates Dashboards. Although different groups have their own unique interpretation of the resulting value, the calculation is relatively standard. One of the two variables needed to calculate this indicator—the Median Household Income (MHI)—is usually obtained from the U.S. Census Bureau and taken on face value. Digging deeper into this variable, however, reveals that it is not as simple as most people consider it to be. Using the Census Bureau’s MHI as-is automatically builds in important qualifications into the percent MHI indicator that could significantly affect the interpretation of its value.

**How is percent MHI being used and calculated?**

This handy little indicator is used by utilities in rate-making, agencies in determining eligibility for funding, the Environmental Protection Agency in administering certain regulations, and advocacy groups and researchers in analyzing affordability of rates in local and regional areas. (Oh, to answer the question above, there is no standard threshold of affordable rates that all utilities are supposed to use. You set your own targets. Utilities can read more about addressing customer affordability in Stacey Isaac Berahzer’s blog post).

The formula to calculate percent MHI of the periodic residential water and wastewater rates and bills is standard for all groups:

Percent MHI = Total of Bills for One Year for a Residential Customer / Median Household Income of All Customers

The resulting number can be interpreted as “the household that is half-way in the spectrum of all households in terms of income is spending X percent of their annual income on water and wastewater.” It goes to follow that half of the customers spend less than X percent and half spend at least X percent. Of course, the number is dependent on the choice of data point for the numerator: *Do you include water and wastewater together? At what consumption point should the bills be computed for? Should a range be used?*

On the denominator side, the MHI number comes either from primary data sources (e.g.: a survey of current customers completed by the utility) or from secondary data sources (e.g.: from the U.S. Census Bureau). The rest of this blog post will focus on using the U.S. Census Bureau’s MHI estimates for calculating percent MHI.

**Where do you get MHI data from the U.S. Census Bureau?**

The Census Bureau’s American Factfinder website allows users to query all of the Bureau’s data from all of their surveys, for all topics, and for all locations in the United States. To go straight to MHI estimates, type in B19013 (which is the code for “Median Household Income in the Past 12 Months”) in the “topic or table name” box and begin to type in and then make a selection inside the “state, county or place” box (a Census Place can be roughly equated to a municipality). For regional areas such as metropolitan statistical areas, make your selection through the Geographies filter.

**How did the U.S. Census Bureau determine the Median Household Income for your utility?**

It didn’t. This is the first issue with using the Census Bureau’s MHI numbers when calculating percent MHI for utility rates: the median household income is for the entire state, county or place (hereinafter simply referred to as “communities”), not the customer base of the water or wastewater utility. A utility with a customer base that extends beyond (or is only a subset within) the county or town limits will have to approximate using one or more communities’ MHI data, making the assumption that the customer base’s household income spectrum has the same distribution as that of the communities it selected.

The U.S. Census Bureau estimates the median household income of communities by conducting surveys of samples of households in all states, counties and places, over several years. This survey is called the American Community Survey, and the results are updated **every year**. Prior to 2005, we had to wait ten years for the latest decennial survey to obtain new MHI estimates. Now, the American Community Survey updates MHI numbers for all communities annually. There are three versions of the American Community Survey (ACS): the 1-year ACS re-samples all the households every single year in large communities, the 3-year ACS re-samples households and surveys them within a 3-year period (inflating the results to the latest year), and the 5-year ACS re-samples households and surveys them with a 5-year period. Small communities will only have access to data from the 5-year ACS, but the results are still being updated and refreshed every year.

**Issues and assumptions when using the U.S. Census Bureau’s MHI estimates to calculate percent MHI for residential rates**

1. **The MHI estimate is for the state, county or place, and not the customer base of the water or wastewater utility itself.** This is explained in more detail above.

2. **MHI estimates are different across the three versions of the ACS (1-year, 3-year and 5-year), even for the same year.** Communities with fewer than 20,000 people will only have access to 5-year datasets. However, communities with more than 20,000 people will have access to the 5-year and 3-year datasets (and also the 1-year dataset for communities with at least 65,000 people). MHI is determined in each of these datasets, and the estimates are different even for the same year, as you can see from the example below for four communities of varying sizes.

Census Place or County | Census 2010 Population Estimate | 2011 MHI from the 5-year ACS | 2011 MHI from the 3-year ACS | 2011 MHI from the 1-year ACS |

Irving city, Texas | 216,290 | $ 48,005 | $ 47,866 | $ 47,266 |

Oshkosh city, Wisconsin | 66,083 | $ 43,203 | $ 40,187 | $ 39,387 |

Lumpkin County, GA | 29,966 | $ 43,704 | $ 43,306 | NO DATA |

Rose Hill town, North Carolina | 1,626 | $ 33,125 | NO DATA | NO DATA |

The U.S. Census Bureau suggests using the 5-year ACS data (for all communities) for the best precision and most reliability of the estimates, although some of the households sampled may have responded to the survey four or five years prior to the publication of their data.

3. **The latest MHI estimates are usually more than a year old.** By the time the U.S. Census Bureau collects, compiles, analyzes, reviews and then publishes its data, a year will have passed between the end of the survey administration and posting the data online. Hence, the latest MHI data will always be at least one year old. For example, the 2011 5-year ACS data were published in December 2012. Between December 2012 and November 2013, we can only use the 2011 ACS data with MHIs estimated for and reported in 2011 dollars. Comparing 2013 water and wastewater rates to 2011 MHI data provides an imprecise measure of affordability, since current MHI will be different from the MHI estimate of two years ago. Following the Census Bureau’s methodology, however, you can adjust the MHI estimate to the current year using inflation, although median household income levels have not kept up with inflation in many years.

In addition, the 5-year ACS is conducted over a five year period (2007 through 2011 for the latest dataset), meaning that the 2011 MHI estimate was determined partially from households that completed the survey in 2007, 2008 and 2009. The income levels of households in communities can change drastically within a five-year period, not accounting for the year it takes to compile and publish the data.

4. **MHI estimates for a community can significantly change from one year to the next.**

The Census Bureau’s relatively new practice of updating the MHI estimates for all communities annually, rather than decennially, is welcomed by those computing percent MHI indicators. However, the MHI estimate for a community can jump significantly from one year to the next, especially since the estimate is determined from a statistical sample of households and not the entire community. This effect is usually more pronounced in smaller communities since the sample size is small and the MHI estimate is sensitive to changes in income levels of only a few households. In the example on this graph, Rose Hill’s MHI estimate dropped 37% between 2010 and 2011.

If the MHI estimate changes significantly from one year to the next, the percent MHI indicator will reveal very different assessments of the affordability of rates. A utility that doesn’t change its rates may be surprised that their affordability measure reads much lower (or higher) than the previous year. Importantly, policy and funding decisions that are dependent on the percent MHI indicator can yield varying results for the utilities they are assessing from one year to the next. If long-term trends and consistency are important, it may be better to only update the MHI estimate every few years instead of using the latest available MHI estimate every year.

5. **The Median Household Income is not a single number (anymore).** This is an important point that escapes many people using the Census Bureau’s MHI data to compute percent MHI indicators for their varying purposes. To determine a single value for the percent MHI indicator, as is often required for programs and assessments, the MHI number (the denominator) must be a single number. In prior years, the Census Bureau primarily listed a single median household income estimate for every community, at least on their factfinder website. Now, using the ACS sampling methodology, the U.S. Census Bureau computes a

*range*that the MHI for the community falls within, and displays both a single point estimate (often, this is used for calculating percent MHI) as well as the margin of error for that point estimate. Due to the sampling technique, the actual MHI for the community may be any number within that margin of error, and may not necessarily be the same as the point estimate number itself. However, most percent MHI indicators are calculated using the point estimate alone, ignoring the margin of error.

Compared to the point estimate, the margin of error can be very large, especially for the smaller communities (again because of the smaller sample size and thus greater sampling error). Since the MHI value is actually somewhere between the lower and upper bounds determined by the margin of error, the percent MHI indicator should produce a *range *of values instead of a single, deterministic estimate. The range of percent MHIs, therefore, can also be large, based on the size of the margin of error. This chart displays the point estimate and margin of error of the 2011 5-year ACS estimate of MHI for four communities, and the percent MHI of the utility’s residential water and wastewater bills for 5,000 gallons/month computed at the point estimate and the lower and upper bounds of possible MHIs. As you can see, the percent MHI value can be very different for a utility, based on which MHI estimate you use for the community. In Rose Hill, NC, the utility’s $63.30/month (= $ 759.60/year) charge for 5,000 gallons/month of water and wastewater in 2012 amounts to 2.3% of Rose Hill’s 2011 MHI using the point estimate, but could really be anything between 1.6% MHI to 3.8% MHI, depending on what the actual MHI of the community is (which is somewhere between $20,074 and $46,176 – a very wide range not unlike those determined for many other small communities across the country). If a program or policy depends on the percent MHI value being above or below a single-point threshold, Rose Hill would meet or not meet the threshold, based on which MHI value is used. Larger communities, like Irving, TX and Oshkosh, WI, have smaller margins of error for their MHI estimates, and thus their percent MHI estimates fall within a much narrower range.

6. **MHI does not provide a complete picture of income distribution of households.** The median household income, by definition, indicates the income level that half of the households in the community make more or less than. However, the distribution of incomes could be skewed in strange ways. Two communities could have a MHI of $40,000. In one community, nearly everyone is making between $38,000 and $42,000. In the other community, perhaps 40% of households are below the poverty line (which varies by household size but is close to $20,000), and another 40% have very high income levels. If the rates charged in the two communities are identical, then the percent MHI values will also be identical. Yet, the affordability of the rates for the first community will be different than the affordability of the same rates in the second community, where 40% of households are below the poverty line and are paying much higher portions of their income for water and wastewater service.

**The U.S. Census Bureau’s MHI estimates can still be reliably used to compute percent MHI indicators for residential rates, with caution**

Have I completely discouraged you from using the U.S. Census Bureau’s MHI estimates to compute percent MHI of residential rates? I hope not, and it’s not my intent. Despite the challenges I’ve listed, there are still good arguments for using the MHI estimates to compute percent MHI indicators. Unless the utility or local government conducts its own survey to determine customers’ income levels, the Census Bureau’s MHI data are the most reliable, up-to-date and easily accessible data on the local community’s income levels. Since the Census Bureau now updates its surveys and data annually, and includes MHI estimates for all states, counties and places, agencies and researchers looking to analyze percent MHI for multiple utilities at the same time can use the Census Bureau’s website for quick data collection and consistency in MHI estimates (at least for one-year intervals). Despite the lag time between surveying and publishing the data, MHI estimates from the U.S. Census Bureau are still relatively recent, within a few years, and are not based on a survey conducted a decade ago and adjusted using an adjustment factor. Although MHI does not provide a complete picture of income distributions of the customer base, it is a relatively good approximation of the income below which nearly (but not exactly) half of the residential customers earn. Percent MHI itself is a familiar indicator, useful for what it measures and provides multiple interpretations based on context, and is easy to understand after one gets his or her head wrapped around it the first time. The U.S. Census Bureau’s MHI estimates can still be used to compute percent MHI for utilities, as long as the challenges are well understood and handled thoughtfully, and the assumptions are clearly stated.

## 0 Comments

## 7 Pingbacks