in Albany, Buffalo, Family Income, New York, New York City, Rochester, Syracuse, Upstate New York, Utica-Rome

Census Surveys, Data Limitations, and Household Income Change in New York State

On the surface, finding the answer to how household median incomes changed over the decade from 2010 to 2020 is simple. At the national level, data from the Census Bureau’s American Community Survey is sufficiently robust to allow us to know the incomes of each group for each year and measure change during the period. The data for the most recent decade is good news. Median household inflation-adjusted income increased from $61,259 to $64,994, based on Census Bureau American Community Survey data – an increase of 6.1%. Between 2000 and 2010, real household income decreased by 2.8%.

The American Community Survey is a sample of the nation’s population. Because of natural sampling variability, the median income estimates could differ from the actual population income by an amount that is defined by a “confidence interval.” The classic example given for this concept involves a coin flip. We know that a coin flip will generate an equal number of heads and tails over time. But, as gamblers know, if we flip a coin four times, we don’t always get two heads and two tails. Sometimes three heads come up; other times, we might get lucky and get four heads, or we could get none. But the more times we flip the coin, the closer to the actual 50-50 split we are likely to get. We can calculate the sample size needed to find the true value of some characteristic within a given range or confidence interval.

The American Community Survey provides data that falls within a confidence interval of 90%, meaning that the value they present for a median income is within the specified range of the actual value in 90% of surveys. In this case, the confidence interval indicates that the increase was between $3,502 (5.7%) and $3,969 (6.5%) – a range of only 0.8%. The national-level estimate is very accurate because many people are surveyed, making the likelihood of a significant error relatively small.

But, the picture is less clear below the national level. The Census Bureau’s American Community Survey is large enough to produce precise numbers for the United States, especially when averaged over five years, as is the case here. But in some cases, for smaller units, like states, metropolitan areas, and counties, the data is not sufficiently precise to be usable for comparisons.

Household Income Change at the State Level

Source: Census Bureau – American Community Survey, 2016-2020 vs. 2006-2010. Two thousand ten data adjusted for inflation by the Consumer Price Index for Urban Wage Earners and Clerical Workers from the Bureau of Labor Statistics.

The survey sample shows that household income grew in most states. New York’s median household income increased by 8.4%, ranking 14th nationally according to the data. Only Massachusetts with 10.9% and Rhode Island with 8.5% had greater growth rates in the Northeast, Midwest, and Southeast.

The range in median income change was from a decline of slightly more than 5% to an increase of more than 30%. But, income change fell in a relatively narrow range for most states – between four and 10%. Relatively small income growth differences were associated with more significant state rank changes. Ohio, with 4% growth, ranked 37th. South Carolina, with 5.8%, ranked 26th.

There is more uncertainty in the data at the state level than in the nation. Because fewer people were surveyed in each state, the range of possible outcomes of 90% of surveys of the same population is wider.

For relatively large states, like New York, California, Texas, and Florida, the range of possible values at the 90% confidence level is relatively small. The difference between the lowest and highest inflation-adjusted household income change estimates is less than two percent. For example, for New York, the published median estimate is 8.4%, and the highest estimate found in 90% of samples was 9.6%, the lowest 7.7%.

For states with fewer residents, the potential errors in the Census Bureau’s published median incomes are more significant. For example, Wyoming’s published inflation-adjusted change in median income was estimated to be relatively low at 2.9.%. Because the survey sample in Wyoming was relatively small, we know with 90% confidence that the actual value was between a loss of 0.1% and a gain of 5.9%. Rhode Island has a published median income change estimate between 2010 and 2020 that was relatively high at 8.5%, with a confidence interval that ranged from 5.9% to 11.3%. Given the large confidence intervals around the two states’ median income estimates, both might have had an increase of 5.9%.

Because the differences in income growth between states were slight, and potential errors arising from sampling were more significant than at the national level, we can’t state with certainty the absolute or relative growth rate for states. Wyoming’s rank among states might have been between 33rd and 45th. Rhode Island’s rank might have been between 10th and 18th. New York’s actual rank could have been between 11th and 17th.

Still, for New York, the data is sufficiently strong to show that New York’s median household income growth was greater than the average for the nation and was very likely to be between 7.7% and 9.1%.

Median Household Income Change in New York State

Source: American Community Survey, 2016-2020 vs. 2006-2010. Two thousand ten data adjusted for inflation by the Consumer Price Index for Urban Wage Earners and Clerical Workers from the Bureau of Labor Statistics.

Census Bureau estimates show that median income grew in most counties. Growth was greatest in Kings County (Brooklyn) – 25%, New York County (Manhattan) – 18%, Herkimer County – 18%, Rensselaer County 14%, and Lewis County – 12%. Census Bureau estimates showed declines in 10 counties, among them Orleans County – 6%, Schuyler County – 4%, Ulster County – 3%, Madison County – 2%, and Ontario County – 2%.

Although the confidence intervals in estimates of median household income change for highly populated states are relatively small, the confidence intervals for estimates for counties in New York State are larger in most cases. In 57 of 62 counties, the range of estimated actual values was 10% or greater at the 90% level of certainty.

In several of the state’s largest counties, the range within which actual income change might have fallen was between four and five percent. In some cases, that is nearly as large or larger than the published median income change estimate. For example, in Monroe County, the Bureau data estimates the change in median household income from 2010 to 2020 was 3.4%. But, the actual value could have been as small as 1% or as large as 5.8%. For Erie County, the Bureau estimated growth to be 7.3%, even though the change could have been as small as 4.7% to as much as 9.8%, based on the 90% confidence threshold. Given the uncertainty associated with the relatively small samples used at the county level, Erie County’s median income growth might have been substantially higher than Monroe’s, or it could have been about one percent less.

Still, the data about median income changes are significant enough to draw some conclusions. Many counties had median income growth during the period. Some had robust growth. For example, the increase in Kings County (Brooklyn) was so great – 23% to 28% – that it was undoubtedly the most in the state. Growth in Nassau, Herkimer, Saratoga, Rensselaer, New York(Manhattan), and Queens County was also significant.

In less populated counties, the confidence intervals are substantial. For example, in Delaware county, although the change shown using the Bureau’s median household income estimates for each year is a decrease of -0.6%, the actual value could have been a decrease of as much as -12.6% to an increase of 9.8%. In the state’s least populous county, Hamilton, median household income could have declined by as much as -24.7% or increased by 29%.

Change in Metropolitan Areas in New York

Because metropolitan areas consist of groups of counties (areas with a large population center and surrounding areas that are economically and socially connected), they have larger populations than counties in most cases. For that reason, the confidence intervals around Census Bureau median income estimates are often smaller.

Source: American Community Survey, 2016-2020 vs. 2006-2010. Two thousand ten data adjusted for inflation by the Consumer Price Index for Urban Wage Earners and Clerical Workers from the Bureau of Labor Statistics.

The data shows two things – the estimated amount of median income change between 2010 and 2020 in New York metropolitan areas, represented by the vertical distance between blue and orange dots, and the confidence intervals for each year – represented by the bars attached to the dots. In the case of the New York metropolitan area, the bars are too short to be visible, because the estimates are very precise. The dots for the metro area show that income grew substantially – from about $75,000 to $82,000.

Most metropolitan areas in the state had median income growth during the period. In most cases, the confidence intervals for 2010 and 2020 do not overlap, indicating a high level of certainty that the change was real. Only the Poughkeepsie, Kingston, Elmira, and Binghamton metros had overlapping confidence intervals, meaning that there might have been no real median income growth during the period. The gaps between the dots for New York City, Albany-Schenectady-Troy, Buffalo- Niagara Falls, and Utica-Rome were large, indicating that those metros had the most significant income growth – 10% or more.

Conclusions

2010 to 2020 was a period of robust household income growth for New York State. New York’s growth was substantially greater than the nation’s – 7.7% to 9.1% compared with 6.1%. Within the state, most of the state’s metropolitan areas showed increases, with a number having income growth greater than the nation. Given the range of sampling variation, we can confidently say that the New York metropolitan area, Albany-Schenectady-Troy, Syracuse, Ithaca, Buffalo-Niagara Falls, and Utica-Rome had greater growth.

Unlike the Decennial Census, which attempts to contact every American, the American Community Survey is an ongoing annual survey that reaches 3.2 million households. Although the survey is extensive, confidence intervals for data below the state level can be too large to provide highly precise information. In larger, more populated places, we can get reasonable estimates of the change in median incomes. For less populated locations, the estimates are too uncertain to permit conclusions with high levels of certainty.

Assessing income change over time is particularly challenging because two data points are used. Each has an associated range around a sample number associated with the actual population value. The two confidence intervals increase the uncertainty related to a particular change estimate.

For the nation, the states, most metropolitan areas, and larger counties, the ACS data provides useful information about income change over the past decade. For less populous locations, the relatively small household samples lead to large confidence intervals around the Bureau’s estimates that make conclusions about income change uncertain.

Write a Comment

Comment

This site uses Akismet to reduce spam. Learn how your comment data is processed.