Funding Liquidity and Stocks’ Market Liquidity: Structural Estimation From High-Frequency Data

In accordance with trade signals that operate in the market, we design a microfounded structural model of price formation that features partially informed and noise traders. The former only have information on whether a trend in the latent price dynamic is underway. Without any trend, the partially informed agents do not trade, and prices do not update unless a noise agent activates. Assuming market efficiency, we impose zero expected net profit per trade. With dedicated parametric assumptions, we analytically derive the model’s likelihood, which allows reliable daily estimates (exclusively based on intra-day transaction prices) of the stocks’ market liquidities and funding liquidity (and their estimation errors). Theory predicates that stocks’ volatilities, stocks’ market liquidities, and funding liquiditymay interact in a non-trivial fashion. To shed light on their nature and mutual influence, we model their dynamics through an MGARCH-VAR process. The model is flexible enough to capture some of the well-known empirical features of financial data, such as fat-tailed distributions and conditional heteroskedasticity. Following an econometric methodology of standard practice in the realized volatility literature, the model is fitted on estimates (obtained fromintra-day data through the structural model estimation) of the daily proxies for stocks’ volatilities, stocks’ market liquidities, and funding liquidity. On a dataset of NYSE stocks, we find significant evidence in favor of four stylized facts: (i) stocks’ volatilities, stocks’ market liquidities, and funding liquidity co-move; (ii) co-movements are stronger when funding liquidity dries up; (iii) stocks with lower volatility are characterized by higher market liquidity, and (iv) funding liquidity restrictions have a stronger impact on stocks’ market illiquidities of high-volatility stocks.