Hurricane Maria made landfall in Puerto Rico on September 20, 2017 as a Category 4 storm and caused Puerto Rico’s entire energy grid to fail, leading to the longest blackout in American history. The storm damaged critical infrastructure and reduced households’ access to food, water, and medical care. Several studies of Maria find a significant, positive relationship between outage duration and several measures of social vulnerability. Formulating the proper policy response to address this inequality requires identifying the underlying causes, and recent studies of the Southeast United States suggest recovery procedures themselves contribute. In this paper, I apply spatial regression methods to data from power recovery crew deployments following Hurricane Maria and find, conditional on the impact of the storm itself, a significant, positive relationship between outage duration and socioeconomic vulnerability, but no statistically-significant relationship between two other forms of social vulnerability and outage duration, namely vulnerability defined by the quality of available housing and transportation and by household composition (e.g., 65+ population). This is consistent with recent studies of the American Southeast but contradicts another recent study of Maria which relies on the same data but uses methods ill-suited for the empirical setting. In addition to this primary analysis and unique to this paper, I obtain geospatial data on Puerto Rican infrastructure to explore alternative spatial weight matrices and test for potential biases caused by standard weighting practices in the literature. I find no evidence of such biases.

In the official models for projections and policy analysis (used by the Treasury, the Social Security and Medicare Trustees, and the Congressional Budget Office (CBO)), many key variables are assumed as a continuation of past trends. By contrast, in our model, these variables are simultaneously determined by supply and demand, based on logical functional forms and parameter estimates from the literature or empirical analysis. This approach better reflects real economic relationships—between health care spending, the federal budget, and investment in capital—and changing underlying conditions, especially demographics. Within the next ten years, we find the federal government budget deficit will grow significantly beyond historical experience and should be regarded as unsustainable. We project that debt-to-GDP will be 135 percent in 2032 and 268 percent in 2052, compared to CBO’s 112 percent and 177 percent, respectively. Real interest rates rise in the long run, ratcheting interest payments, deficits, and debt, and vice versa. Our projection of national health expenditures relative to GDP in 2072 is 31.4 percent, compared to 28.4 percent by the Centers for Medicare & Medicaid Services (CMS). These higher costs of health care arise from labor shortage effects in an aging economy because health care is produced in a low productivity, labor-dependent sector. Health care expenditure further deteriorates the federal budget and lowers consumer welfare.

This report expands on a standard empirical estimation of the relationship between federal deficits and debt and long-term interest rates. It follows closely a 2019 long blog post by Ernie Tedeschi, which is itself an update and extension of Francis Warnock and Veronica Cacdac Warnock (2009). Using data from September 1981 to May 2022, we find that a 1 percentage point increase in the federal debt-to-gross-domestic-product ratio is associated with an increase of nearly five basis points in the long-term interest rate. This is a larger effect than generally found in the literature and double what the Congressional Budget Office uses in its budget projections, which we attribute to our more complete specification of Federal Reserve policy.