About Me

Government subsidies bring firms attention, not information. After a first award, analyst coverage rises while the content of analyst output does not change; insiders trade more, tilt toward selling, and their post-award sales predict returns less well than ordinary insider sales.

Related work shows that executives buy their own stock when analysts are pessimistic but do not trade on the firm's public conduct record, and that 44.7 percent of the variance in monthly consensus changes comes from analysts entering and leaving the roster rather than from revised views.

Hidden duration losses become priced crash risk only when a salient shock forces investors to revalue long-duration assets. In the 2022 UK gilt crisis, banks with higher pre-shock gilt sensitivity were more likely to crash, while a placebo year shows nothing.

A census of 11,622 U.S. banks from 2000 to 2025 shows that depositor discipline is small in normal times and about 10 times larger in the 2008 to 2010 crisis. A further paper studies how markets reprice unrecognized held-to-maturity losses in U.S. bank equity.

Under one fixed, cost-aware protocol on real market data, deep reinforcement learning does not robustly beat naive 1/N diversification or a well-implemented mean-variance optimizer, and a controlled simulation shows why.

Other work compares how large language models and managers allocate capital against q-theory, and measures what earnings calls carry that the 10-K leaves out.

Earlier work studies nonfarm income and migration in Bangladesh (Development Policy Review, 2023), Islamic banking and finance, and trade, fuel efficiency and emissions with the GTAP model. I have spoken at the U.S. Department of State's Foreign Service Institute on remittances and on the economy, infrastructure and energy of Bangladesh.