The scintillation index is a common measure of the effects of atmospheric turbulence. Using a photon-counting sensor, the integration time for each sample needs to be short enough to ensure that the intensity is constant during this time. Simultaneously, hardware limitations, including detector dead-time, restrict the count rate so that the number of counts in a single time segment is extremely low. The dead-time also introduces nonlinear effects. The variance calculation in the scintillation index formula is then dominated by quantization error, and the scintillation index is severely overestimated. We investigate two methods of correcting the scintillation index based on data from a time-correlated single-photon counting laser radar system. The first approach is based on the covariance calculation of the data and can be used for very low count rates and high temporal resolution. This method may also be useful in other cases where the variance of noisy, time-resolved data needs to be calculated. The second method is based on fitting the theoretical probability density function for the intensity fluctuation caused by propagation through turbulence to the experimental data. This method can take dead-time effects into account and be used for higher count rates.