What’s Not There: The Odd-Lot Bias in TAQ Data

We investigate the systematic bias that arises from the exclusion of trades for less than 100 shares from TAQ data. In our sample, we find that the median number of missing trades per stock is 19%, but for some stocks missing trades are as high as 66% of total transactions. Missing trades are more pervasive for stocks with higher prices, lower liquidity, higher levels of information asymmetry and when volatility is low. We show that odd lot trades contribute 30 % of price discovery and trades of 100 shares contribute another 50%, consistent with informed traders splitting orders into odd-lots and smaller trade sizes. The truncation of odd-lot trades leads to a significant bias for empirical measures such as order imbalance, challenges the literature using trade size to proxy individual trades, and biases measures of individual sentiment. Because odd-lot trades are more likely to arise from high frequency traders, we argue their exclusion from TAQ and the consolidated tape raises important regulatory issues.