This paper evaluates four popular algorithmic models (SLIM, COCOMO, Function Points, and ESTIMACS) used to estimate software costs. Data from 15 large completed business data-processing projects were collected to test the accuracy of these models' ex post effort estimation. The results show that Albrecht's Function Points model was validated by the independent data, while models not developed in business data-processing environments required significant calibration. The models failed to sufficiently reflect underlying productivity factors, indicating the need for further research. The study addresses three key research questions: the generalizability of the models to different environments, the relative accuracy of SLOC-based versus non-SLOC models, and the comparison between proprietary and nonproprietary models. The findings suggest that models developed in different environments need calibration, non-SLOC models perform better in business data-processing environments, and proprietary models may not always outperform nonproprietary ones. The paper concludes with implications for practitioners and directions for future research, emphasizing the importance of collecting historical data for calibration and the need to improve models to better capture productivity factors.This paper evaluates four popular algorithmic models (SLIM, COCOMO, Function Points, and ESTIMACS) used to estimate software costs. Data from 15 large completed business data-processing projects were collected to test the accuracy of these models' ex post effort estimation. The results show that Albrecht's Function Points model was validated by the independent data, while models not developed in business data-processing environments required significant calibration. The models failed to sufficiently reflect underlying productivity factors, indicating the need for further research. The study addresses three key research questions: the generalizability of the models to different environments, the relative accuracy of SLOC-based versus non-SLOC models, and the comparison between proprietary and nonproprietary models. The findings suggest that models developed in different environments need calibration, non-SLOC models perform better in business data-processing environments, and proprietary models may not always outperform nonproprietary ones. The paper concludes with implications for practitioners and directions for future research, emphasizing the importance of collecting historical data for calibration and the need to improve models to better capture productivity factors.