This dissertation consists of two chapters spanning household and behavioral finance. Chapter 1 studies how lower trading costs affect retail investors and the market. Brokerage commissions have declined substantially over recent decades worldwide. This paper studies how lower trading costs affect retail investors and the market through increased speculation. We leverage a 2017 reform in Taiwan that reduced the transaction tax specifically for day trading. Using detailed account-level transaction data, we show that even though lower trading costs mechanically benefit investors, their responses offset these gains and reduce portfolio returns. First, day trading volume increases significantly but this increase is driven disproportionately by less sophisticated investors who tend to lose money on each trade. Second, and surprisingly, day traders' gross returns per dollar traded worsen, eroding the mechanical tax-cut benefit. Consistent with the view that transaction costs serve a disciplinary role, we trace this performance deterioration to less attentive decision-making. Together, these responses result in losses concentrated among investors with smaller holdings, whereas large investors benefit. Despite individual-level losses, market quality improves: intraday liquidity increases and volatility decreases. Overall, our findings highlight a policy-relevant trade-off: increased retail speculation from lower trading costs can benefit markets while harming individual investors. Chapter 2 studies the source of predictable currency returns by interest rate differentials using survey data on exchange rate and interest rate expectations. With a present value decomposition of exchange rates, I highlight that the predictable innovation in subjective currency risk premia plays a crucial role in explaining the ex-post predictability of exchange rate forecast errors, which results in predictable currency returns widely-documented in the literature. This is a novel channel of predictability not emphasized in exchange rate models featuring rational expectations or assuming investors make predictable errors for interest rates. As an illustration, I propose a reduced-form model with time-varying subjective perceptions of risk that generates predictable innovation in subjective risk premia. I further present empirical evidence that is consistent with a key property of the model---a positive relationship between subjective perceptions of risk and subjective return expectations.
