It has been well documented that using the Black-Scholes model to price options with different strikes generates the so-called volatility smile. Many previous papers have attributed the smile to the normality assumption in the Black-Scholes model. Hence, they generalize the Black-Scholes model to incorporate a richer distribution. In contrast to previous studies, our model allows for not only a richer distribution, but also the relaxation of another crucial assumption in the Black-Scholes – continuous trading. We show, using S&P 500 call options, how relaxation of continuous trading explains a non-trivial change in the volatility. When an empirical distribution is considered, the smile is almost completely removed. Market prices of options that differ only in their strike prices are inconsistent with a single volatility for the underlying asset, and this well-known feature is called the volatility smile or smirk. The volatility smile is often attributed to either the non-normality of stock returns or to the impossibility of performing costless arbitrage (rebalancing) in continuous time. In this paper, we consider option pricing models that incorporate no rebalancing and/or a nonparametrically estimated density. We attempt to empirically identify which of these two factors contributes more to the smile bias. Using S&P 500 index options, we find that the model with no rebalancing but with normality has a somewhat diminished smile, and the model with a nonparametric density, but with continuous trading, has a slight smile. Only the model with both a nonparametric density and no rebalancing has no perceptible smile bias.