Michael Hicks
In 2022, the US passed the “Creating Helpful Incentives to Produce Semiconductors for America Fund,” or the CHIPS Act for short. This is a very large bill, $280 billion, to be spent over several years promoting the growth of the semiconductor industry in the US.
I’m not usually a fan of this sort of activist industrial policy. Free markets and free trade typically move the production of goods and services where they are best suited. However, there are a few things about this policy that warrant a second look.
In the wake of the pandemic, it is clear that the US is overreliant on distant supply chains for semiconductors. Several key American industries, most obviously automobiles, continue to suffer from supply chain interruptions of semiconductors. The typical American automobile uses more than 1,400 semiconductors out of 30,000 different parts.
The availability of semiconductors is critical to the national economy, but that alone wouldn’t justify spending $280 billion to support the industry. Surely, automakers and others are asking the world’s chip makers to come to the United States to forestall just such future risk. But there is more.
Semiconductors are also a key part of national defense. Handheld javelin missiles, best known for expediting Russian tankers into the afterlife, use well over 200 semiconductors each. These products are used in every handheld radio on the battlefield, in the devices that ensure secure communications and in every military vehicle. This is perhaps the only American industry in more than a century that legitimately warrants government support on national security grounds.
The good news is that some of the industry was bound to return to the United States under any circumstance. That means the direct costs of the bill are lower and the effect will be more immediate than would be the case otherwise. It is also worth noting that the CHIPS Act was introduced by Republican Indiana Sen. Todd Young, who has been working on this effort since the start of the pandemic.
Of course, there is interest in the location of these new factories, and the potential economic impact to cities where they locate. That interest motivated myself and colleague Dagney Faulk to study the issue. We were specifically interested in trying to assess where the new plants would be located.
Our study (https://projects.cberdata.org/191/semiconductor-plant) examined every county the United States for which there was available data, a little more than 3,000 counties. The statistical model we used was designed to analyze where the currently 50 or so semiconductor plants were located. We judged these variables to be the most predictive set of regional characteristics for a new plant. Our model used data on population, employment, current industry mix, age, wages, immigration, commuting, agglomerations (city-specific productivity), quality of life, a whole slew of human capital measures, and the presence and strength of research universities.
We found that the most important regional characteristics were the share of adults with a bachelor’s degree or higher, the share of adults in graduate school, population growth and agglomerations. This was not surprising. Fast-growing places with highly educated workers near a research university with large, unexplained productivity (agglomerations) are precisely the factors we’d expect in a high-wage, highly sophisticated industry.
We then isolated the top 1.5 percent of counties, and ranked them as the most likely to see the opening of a new semiconductor plant. We called these Tier 1, Tier 2 and Tier 3 cities, along with all those that already had a plant. And, to test our model, we then compared our findings against a list of 20 already proposed factories that were not yet in our database. Our model predicted 18 of the 20 cities on that list. Even more happily, since we completed the study, another proposed plant opened in one of our “predicted” metro areas, in Indiana. That’s a pretty happy outcome for an economic model of this sort.
Among the happier findings is the likelihood that the Midwest will see a disproportionate share of new plants. As of now, there is only one semiconductor plant in the Great Lakes states. Our model predicts a dozen more counties are among the most likely locations for plants.
Of course, the model isn’t perfect. Our model missed two proposed plant openings. One of these was in Tippecanoe County, which just missed the cutoff criterion for our three tiers. The second one was in a Michigan micropolitan area adjacent to two metropolitan areas that made the list. These errors point to a more useful interpretation of our model.
It is difficult to directly link the CHIPS Act to the new factories here in Indiana, or the other two announced so far in the Midwest. While it will be years before anything is produced at these plants, the U.S. is now entering a period of more predictable supply of semiconductors. This will be important to every manufacturing and logistics industry, and anyone who relies upon any modern technology. Most importantly, the coming decade ensures the US will have access to the semiconductors we need for national defense.
Michael Hicks is the director of the Center for Business and Economic Research and an associate professor of economics in the Miller College of Business at Ball State University. Send comments to [email protected].
