Response to Michael Levi’s Council on Foreign Relations Blog Post–Part 2

Another criticism made by Levi is that the estimated employment impacts in our study are invalid because more capital-intensive industries will be more heavily affected than labor-intensive industries by climate policy.

---

As we make clear in our study, our figures for potential job losses are only order-of-magnitude estimates designed to give a general idea of the size of the employment effects we can expect from a policy that reduces GDP by the amounts predicted by EPA in various years. We don’t model the entire American Power Act bill. Instead, we show about how many jobs can reasonably be expected to disappear if GDP falls by a given amount, holding all else constant. In our study, we assume overall GDP reductions will be felt by industries in proportion to the fossil-fuel carbon intensity of their products. Levi is right that if industries are affected in different proportions than what we assumed, the pattern of employment losses — and potentially the overall total job losses — will differ from our estimates. But it’s easy to see that they won’t differ by much. In fact, it turns out our estimates are robust across a wide variety of assumptions about the distribution of GDP impacts among industries. To see why, suppose Levi is correct that capital-intensive industries will be most heavily affected by Kerry-Lieberman. Rather than dividing the overall GDP impacts among industries by carbon intensity as we’ve done, we can instead divide it by an estimate of capital intensity by industry. A back-of-the-envelope way of doing this is to use data on the relative share of labor income as a percentage of value added for industries. In capital-intensive industries, labor income will be small relative to total value added. We can then weight these “capital intensity factors” by total industry output to arrive at a reasonable proxy for capital intensity by industry. These figures can then be used to distribute overall GDP impacts to industries, consistent with Levi’s argument above. Re-running our model using this method, we find the employment effects of Kerry-Lieberman would be significantly larger than our estimates—not smaller as Levi assumes. Here are the figures for total job losses in various years under the assumption that capital-intensive firms are more heavily affected: 2015 : -653,783 2020 : -895,924 2030 : -3,511,055 2040 : -4,915,477 2050 : -6,440,970 Overall, these figures are broadly comparable to our original estimates. However, they are higher by roughly 25 percent. It is simply not the case that our study has over-stated employment effects from the bill as Levi claims. To the contrary, if Levi’s argument above is correct, our estimates may in fact may err on the conservative side. This should not come as a surprise—our estimates of job losses should be considered order-of-magnitude estimates, which are unlikely to vary dramatically to changes in the assumption of how overall GDP declines are distributed among industries.