This paper introduces a generalized order flow imbalance (GOFI) construction method that accounts for non-minimum quotation units in real transactions, improving upon traditional Order Flow Imbalance (OFI) and Stationarized Order Flow Imbalance (log-OFI). The key innovation is to consider price changes beyond the minimum tick size by aggregating order flow across multiple price levels, which better captures the true market imbalance. Using high-frequency order book snapshot data from CSI 500 constituent stocks, the authors empirically compare the explanatory power of these indicators for short-term price changes at 30-second, 1-minute, and 5-minute intervals.

The results demonstrate that both GOFI and its stationarized version (log-GOFI) significantly outperform traditional OFI and log-OFI. Specifically, log-GOFI achieves average out-of-sample R-squared values of 83.57%, 85.37%, and 86.01% for the three time scales, compared to only 32.89%, 38.13%, and 42.57% for OFI. Moreover, log-GOFI exhibits stronger stability in its explanatory power across different time scales. These findings suggest that incorporating multi-level order book information into order flow imbalance measures can substantially enhance the ability to explain and predict short-term price movements in stock markets.