Flood damage assessment is a necessary tool in the planning of flood-prone areas. There are several factors affecting the flood damages. It is not easy to detect these effective factors by classical methods. In this study, correlation coefficient and cross wavelet analysis are used to look for a possible connection between flood losses and large-scale climate indices. Some strong connections suggest that sea surface temperature anomalies influence the general characteristic of flood damage distribution across the United States. Time-series analyses of flood damage data reveal that there is an upward trend in the flood losses. This apparent trend can be related to increase in population. Also, an autoregressive model and a regression model are proposed on the predictability of flood damages. It is shown that model errors stay within the acceptable error limits.