An Assessment of the Predictive Performance of Moving Average, Regression Analysis, Exponential Analysis and Natural Logarithm in Determining Future Property Values

Purpose: The purpose of this study was to measure the accuracy of performance of the prediction of moving average, regression analysis, exponential analysis, and natural logarithm in determining future value of a property.Design/Methodology: Historical property value data was obtained from an investment over a period of 16 years. The time series data was used to plot the charts and derive equations for the four models that were used for predicting the next year’s value and forecasting of values for the next ten years in MS Excel. Residuals, percentage errors and ratios were used to establish the accuracy performance of the models using actual value minus predicted value.Findings: It was revealed that the four models had capabilities to predict value in the short term, however the three series moving average had no capabilities of forecasting of values in the long run. From the findings, it was established that the most accurate model in predicting the value of property in the short term was exponential analysis with (100%) of the errors in the acceptable range of 0% to ±10%. This was followed by the regression analysis model with 69% of errors falling with the acceptable range of margin of error. The natural logarithm was third with 38% of errors falling within the acceptable margin of error. Finally, the moving average with 31% of errors within the range of margin of error. The measures of appraisal uniformity were all within the acceptable margins for COD, COC, COV, MAD, AAD, MAPE and PRD and also supported the results that were obtained based on the margin of error. This was further confirmed using coefficient of determination where R2= 98.7%, 97.0% for regression and 76.6% for natural logarithm.Research limitations/implications: These models though useful, they have little appreciation by the users in the property industry for decision making purposes as most of the assessments are based on subjective assessment of value. These could be useful in predicting and forecasting of value for investment decision making purposes.Practical implications: If adopted by practitioners these can be good models to use in predicting of values in the short term as well as forecasting of future values in the long term for decision making purposes.Originality/value of the work: It is noted that there are no studies that have compared performance of models in predicting and forecasting of values in the short and long terms. The contribution to the body of knowledge is the introduction of the four models in predicting and forecasting of property values in the short and long terms.