In order to maintain financial stability and meet regulatory capital requirements, banks must understand not only the default risk of individual obligors but also the degree to which defaults occur together within a credit portfolio. The tendency of obligors to default jointly, captured by default correlation, is a central determinant of unexpected losses and therefore of the economic capital a bank holds against its credit exposures. This thesis compares three approaches to estimating default correlation in credit portfolios and evaluates how their different characteristics affect the resulting estimates. Focus is placed on how robust each approach is when applied to data conditions typical of internal credit portfolios, including short historical time series, small portfolios and low default frequencies. The estimators are evaluated through a synthetic experimental framework with known parameter values and then applied to internal the SME corporate credit data from Handelsbanken. Our main findings are that the length of the historical time series and the size of the annual portfolio are the most binding constraints on estimation precision and that estimates of default correlation applied to sparse credit data carry substantial uncertainty, which makes point estimates alone potentially misleading. The results suggest that practitioners should report default correlation estimates with measures of uncertainty and interpret them with caution in settings where the available sample is short or when the number of observations is small.