This study investigates how cross-sectional characteristics of ETFs affect the long-term returns of ETFs. We use tracking multiples, underlying market classification, tax type, manager, underlying index, volatility, replication method, classification system, index producer, and underlying asset class as cross-sectional characteristics, all of which are characterized by nominal variables. The performance of ETFs is characterized by Dollar Cost Averaging (DCA) returns, Lump-sum deposit (LD) returns, and return differential (RD=DCA-LD). Tracking multiples are categorized into 1X Inverse, 2X Leveraged, 2X Inverse, and Normal, and underlying market classifications are categorized into Domestic, Domestic & International, and International. Both tracking multiples and underlying market categorization had a cross-sectional impact on DCA returns, LD returns, and RD. Taxation type had a significant cross-sectional effect on DCA and LD returns only for non-taxable. Dividend income tax (separately taxed real estate ETFs) had an impact only when the dependent variable was DCA returns. Dividend income tax (foreign equity only ETFs) had no effect on DCA and LD returns.