For economic researchers and policymakers, mapping and analyzing the level of similarity between states and regions can be a useful exercise. By identifying the regions with close similarities, decision makers can help narrow their search for successful policies to use in their own communities. This article examines a common methodology for comparing statistical similarities, evaluates how similar Idaho’s economy is to that of the other 49 states and analyzes the overall usefulness of the results.
Using a variation of a common statistical method called a nearest neighbor analysis allows researchers to take a set of observations (in this case, the 50 states) and rank them in order of similarity. For this analysis, the method analyzes the economic composition of each state, based on the employment in each state across 285 different industries. Then the mix of industries in each state is compared according to the percentage of that state’s employment concentrated in that industry.
For example, in Idaho, 33 industry categories had no employees at all in 2018, including pipeline transportation and coal mining. At the other end of the spectrum, Idaho’s largest industry – restaurants – employed more than 52,000 people to make up more than 8 percent of Idaho’s total employment.
In this analysis, the nearest neighbor algorithm analyzes the mixture of industries in each state and compares it to the mixture in Idaho in order to rank the other 49 states in order of similarity to Idaho. This concept of similarity can be easily represented graphically – when comparing jobs distribution – two states that are highly similar will graphically show a strong linear relationship, while states with low similarity would appear to be distributed randomly.
For Idaho, similarity analysis reveals the most similar state economy belongs to close neighbor Oregon. This is not surprising as the similar geographic profile leads both states to have similar profiles in natural resources, notably logging and wood products. The other states with high similarity scores compared to Idaho are also near neighbors, including Montana and Colorado.
At the other end of the spectrum, Idaho is the least similar to Hawaii. This, too, is unsurprising. Hawaii has virtually no employment in key resource sectors, like logging and cattle ranching, which are important to Idaho, while the state’s heavy reliance on tourism leads to employment concentration in hosptiality sectors, like hotels, resorts, and restaurants, that are many times greater than in Idaho. The second-least similar state, Nevada, is similarly tourism-dependent.
When this nearest neighbor analysis is conducted, two important observations immediately stand out. The first is that states are far more alike than they are different. In the case of Idaho, even the states least similar in their employment mix show a high degree of statistical correlation. This directly relates to the second key observation, which is the differences between states are most apparent among smaller industries.
In almost all U.S. states, the largest industries are exactly the same. For example, in 42 of 50 states, the three largest industries are identical: restaurants, elementary and secondary schools, and hospitals. It is worth emphasizing, however, that the industries that make states unique tend to be relatively small.
In Idaho, the most oversized industry – meaning, not the largest in absolute terms, but the largest relative to other states – was cattle ranching and farming; yet in 2018 this industry only represented 1.5 percent of Idaho’s employees. In fact, in all but four states the most oversized industry still employed less than 3 percent of the state’s workers. The lone exceptions were automobile manufacturing in Michigan (3.6 percent of the state’s employment), seafood processing in Alaska (also 3.6 percent), scientific research in New Mexico (3.7 percent) and credit intermediation in Delaware (4.5 percent).
The implication of this analysis is that, while similarity analysis may be of academic interest, it is likely to be of limited value to policymakers, because the outlier industries in each state tend to comprise a relatively small share of total employment. Thus, for the most part, similarity analysis is unlikely to reveal profound insights for the simple reason that most states are more similar than they are different. For policymakers seeking to find useful examples and models in other states, a quantitative approach is unlikely to be useful.
Sam.Wolkenhauer@labor.idaho.gov, regional economist
Idaho Department of Labor
(208) 457-8789 ext 4451