This paper proposes a new approach to modeling financial transactions data. A model for discrete valued time series is introduced in the context of generalized linear models. Since the model specifies probabilities of return outcomes conditional on both the previous state and the historic distribution, we call the it the Autoregressive Conditional Multinomial (ACM) model. Recognizing that prices are observed only at transactions, the process is interpreted as a marked point process. The ACD model proposed in Engle and Russell (1998) allows for joint modeling of the price transition probabilities and the arrival times of the transactions. The transition probabilities are formulated to allow general types of duration dependence. Estimation and testing are based on Maximum Likelihood methods. The data are IBM transactions from the TORQ dataset. Variations of the model allow for volume and spreads to impact the conditional distribution of price changes. Impulse response studies show the long run price impact of a transaction can be very sensitive to volume but is less sensitive to the spread and transaction rate.