Essays on Macroeconomics and Economic Modeling

Abstract

This dissertation contains three chapters on topics in macroeconomics and dynamic modeling in fishery economics. In Chapter 1, I study the effect of immigration on the natural rate of interest and household welfare in an economy subject to secular stagnation. I build a three-generation overlapping generations model with heterogeneous skills among native and immigrant workers, incorporating higher immigrant fertility and skill complementarity among high-skilled workers. A no-capital version of the model yields closed-form expressions for the natural rate, aggregate supply, and aggregate demand; an extended version with physical capital provides the quantitative analysis. Immigration unambiguously raises the natural rate through population growth: a decomposition of the decline in U.S. real interest rates between 1970 and 2015 shows that the immigration expansion raised the natural rate by 0.22 percentage points relative to a counterfactual with negligible immigration. The welfare implications are regime-dependent. At full employment, immigration lowers native welfare through wage competition, with low-skilled workers bearing the largest losses. Under secular stagnation, the ranking reverses: immigration reduces the severity of the demand shortfall, and the resulting employment gains make it welfare-improving for all household types, including natives. A policy experiment calibrated to a doubling of the H-1B visa cap shows that 85,000 additional high-skilled workers can reduce the initial output gap following a Great Recession--scale deleveraging shock from 7.3% to 1%. In Chapter 2, co-authored with Yu-Chin Chen and Pushpak Sarkar, we use machine learning techniques to re-examine the long-standing difficulty in predicting currency returns with macroeconomic indicators by focusing on three possible causes: the general lack of information in the macro predictors, mis-specifications in the forecasting equations, and inherent instabilities in the relationship between the exchange rate and its macro determinants. Using a large international dataset that captures current macroeconomic conditions as well as forward-looking market expectations and perceived uncertainties, we forecast monthly returns from 1995 onward of four major currencies (AUS, CAD, GBP, and JPY) against the USD. In in-sample regressions, we see that while market expectations embedded in derivatives markets may help predict subsequent exchange rate returns, there is little evidence that they contain predictive content above and beyond what is in the macro indicators themselves. Moreover, both types of predictors perform better in non-linear specifications than under the linear specifications which often deliver adjusted R^2 around zero. We take these findings as indicative that the exchange rate is not disconnected from indicators of the macroeconomy--be their current values or expectations, though their functional relation may be more nuanced than simple linear specifications can capture. Moving the analyses to pseudo out-of-sample (OOS) forecasts, we find that a multilayer perceptron neural network can generate improvements over the long-standing Random Walk benchmark, some of which are statistically significant under the Clark-West test. More prominently, we see that the majority of the ML methods considered do not outperform a RW forecast given our small sample context. In fact, unlike results for other asset returns, ML does not appear to help resolve the FX forecasting puzzle. Nevertheless, our ML explorations unveil significant empirical instabilities, especially around the GFC period. These findings support the views that pseudo-OOS exchange rate forecasting in finite samples can be overwhelmed by inherent statistical issues such as parameter and model instabilities, and that the exchange rate dynamics are inherently difficult to distinguish from a RW process statistically (Engle and West, 2005). They also indicate that predicting exchange rates are a different endeavor from predicting bond yields. Chapter 3, co-authored with Christopher M. Anderson, applies a similar dynamic modeling framework to fishery management, where institutional design shapes economic outcomes even when biological targets are held fixed. The maximum economic yield (MEY) is increasingly adopted as a fishery management objective, yet standard bioeconomic models treat the cost of harvesting as independent of institutional design. We develop a dynamic model in which biological dynamics are common across management regimes while the law of motion for fishing effort depends on the institutional environment. Under open access, profit-driven entry dissipates rents along the extensive margin. Under limited entry, competition shifts to the intensive margin through capacity investment. Under a total allowable catch (TAC), fishers compete for quota shares by expanding capacity, inflating costs even when the biological target is met. Under individual fishing quotas (IFQ), secure harvest shares eliminate the strategic motive for overcapitalization, and fishers instead minimize costs. Phase diagrams, numerical simulations, and comparisons to the static bioeconomic model show that regimes achieving identical biological outcomes can generate very different economic rents. In our calibration, steady state profit under IFQ is roughly double that under TAC at the same stock and harvest level. The entire gap is attributable to endogenous differences in the equilibrium cost structure. We introduce the concept of an implementable steady state set: the frontier of achievable stock-profit combinations specific to each regime. The sole owner MEY may lie outside the feasible set of regimes that leave competitive margins open, making it an unreliable target for decentralized fisheries. Analytically, IFQ can replicate the sole owner optimum when fleet size and harvest cap are jointly chosen, but this level of coordination is rarely available in practice. These results imply that instrument choice matters as much as target-setting for economic outcomes in fisheries.