Regulators use measures such as a fee on high order-to-trade ratio (OTR) to slow down high-frequency trading. Their impact on market quality is, however, mixed. We study a natural experiment in the Indian stock market where such a fee was introduced twice, with differences in motivation and implementation. Using a difference-in-difference approach, we find that the fee decreased OTR and improved market quality when it was imposed on all orders, while it had little effect when it was imposed selectively on some orders. Improvement in liquidity was driven by a reduction in adverse selection costs following lower OTR.