This paper investigates differences in returns between the ETF price, Net Asset Value, and Benchmark Asset Baskets for five popular futures-backed ETFs. We decompose tracking difference to examine the relative size of tracking differences attributable to managers versus the arbitrage process. Tracking differences attributable to managers is found to be significantly smaller than that attributable to the arbitrage process. We then test for average Tracking Differences using the Mincer-Zarnowitz Equation. We find evidence of bias in returns for multiple ETFs and demonstrate the usefulness of the decomposition. Furthermore, we investigate the dynamics of Tracking Error using a GARCH methodology. We find support that the volatility of the ETF effects Tracking Error but find no evidence that rolling futures contracts influences Tracking Error. / Master of Science / This research focuses on futures-backed commodity ETFs. ETFs are exchange-traded instruments and are a convenient way for investors to gain commodity exposure without having to have access to a margin account, deal with futures contract expiration, or the large size of futures contracts. We investigate the ability of these instruments to achieve their investment goals: namely to perfectly replicate the exposure of a benchmark of futures contracts. We find that differences in the returns of the benchmark and ETF exist on average and that the bulk of these differences are attributable to the Creation and Redemption process rather than the ETF manager. Finally, we find that market volatility effects the volatility of these differences, but roll dates have no effect.