Market microstructure study on seven US stock exchanges: panel vs. VAR methodology

Abstract

This article questions the validity of regression models when high correlations exist between independent variables and presents the application of VAR as an alternative technique through the comparison of two groups of selected stocks that represent components of Dow Jones and S&P 500 indices, respectively. The results indicate that panel regressions face serious specification problems, while the impulse response function underlines that the shock to the volume innovation has a mostly positive impact on the volatility in both S&P and Dow Jones sample, but the tendency cannot be easily accounted for. The positive impact of volatility shocks on the inter market depth is rather unexpected, but it may be associated with an increase in volume that does not enormously enhance the spread up to the point where it will be too costly for market-makers to trade, and accordingly, quickly narrows the spread to absorb new liquidity influx in the market. In the Granger causality tests Dow Jones stocks with comparatively larger average volume depth values and price levels provide slightly stronger relations between analyzed variables compared to the stocks included in the S&P sample.