Farris and Padberg, Krenz, and Boxley have discussed different ways of projecting the distribution of firm sizes. Farris and Padberg utilize Markov chain analysis where transition probabilities were derived from observed changes in firm size for Florida citrus packing firms over a five-year period. Krenz, recognizing the difficulty of securing data with which to estimate transition probabilities, developed rules which could be used to derive transition probabilities from census data. Krenz applied this technique to the size distributions of North Dakota farms. Finally, Boxley applied a completely different technique to estimating farm size distributions in the United States. Boxley's technique involved fitting negative exponential functions to firm size distributions observed from census data. According to Boxley, these functions were sufficiently stable to permit estimation of future firm size distributions.