We study the statistical properties of escape times for stock price returns in the Wall Street market. In particular we get the escape time distribution for real data from daily transactions and for three models: (i) the Wiener process with drift and a constant market volatility, (ii) Heston and (iii) GARCH models, where the volatility is a stochastic process. We find that the first model is unable to catch all the features of the escape time distribution of real data. Moreover, the Heston model describes the probability density function for both return and escape times better than the GARCH model.