Spring frosts are usual in many of Spain's fruit-growing areas, so it is common to insure crops against frost damage. After a frost, crop loss must be evaluated, by comparing what crop is left with the amount that would have been obtained under normal conditions. Potential crop must be evaluated quickly through the use of measurements obtainable at the beginning of the tree's growth cycle. During 1996 and 1997 and in 95 commercial plots of `Blanquilla' and `Conference' pear (Pyrus communis L.), the following measurements were obtained: trunk cross-sectional area (TCA, cm2), space allocated per tree (ST, m2), trunk cross-sectional area per hectare (TCA/ha), flower density (FD, number of flower buds/cm2 TCA), flower density per land area (FA, number of flower buds/m2 land area), cluster set (CS, number of fruit clusters/number of flower clusters, %), crop density (CD, number of fruit/cm2 TCA), fruit clusters per trunk cross-sectional area (FCT, number of fruit clusters/cm2 TCA), fruit clusters per land area (FCA, number of fruit clusters/m2 land area), fruit number per cluster (FNC), average fruit weight (FW, g), average yield per fruit cluster (CY, g), yield efficiency (YE, fruit g·cm-2 TCA), and tree yield (Y, fruit kg/tree). CS and average CY were related to the rest of the variables through the use of multiple regression models. The models that provided the best fit were CS = TCA/ha - FA and CY = -FA - FCT. These models were significant, consistent, and appropriate for both years. Predicted yield per land area was obtained by multiplying FA × CS × CY. The models' predictive ability was evaluated for 46 different plots in 2001 and 2002. Statistical analysis showed the models to be valid for the forecast of orchards' potential yield efficiency, so that they represent a useful tool for early crop prediction and evaluation of losses due to late frosts.