MODELLING THE SINGAPORE ECONOMY : A STUDY ON THE INCORPORATION OF SURVEY DATA

Abstract

Government and private bodies have been constructing macroeconomic models to analyse and forecast the economy. Some of the models are assembled econometrically and are quantitative in structure. They are called "quantitative" models. In contrast, some include survey data. These are known as "mixed" model. Empirical evidence have suggested that the latter type of models are gaining in popularity. The objective of the study is to explore the incorporation of survey data into the modelling of the Singapore economy. More significantly, the bottom-line is to test whether their inclusion has any significant impact on the analysis and forecast of the economy. To achieve this aim, two models of the Singapore economy are constructed -- one consisting only of quantitative data and the other consisting of quantitative and survey data. Three sets of survey data are used -- business expectations, output forecasts and investment plans. Six criteria are used to evaluate the performance of the models. These are 1. Individual Performance of Model Equations 2. Simulation Fit of Equations 3. Turning Points in Historical Data 4. Correlation and Regression Results 5. Forecasting Errors 6. Overall Sensitivity of Models On the whole, the mixed model 1. gives higher adjusted R2 and F values 2. provides lower simulation errors 3. predicts the number of historical turning points more accurately 4. offers better correlation and regression results 5. contributes fewer forecasting errors 6. yields more robust estimates. In addition, counterfactual analysis and forecasting exercises are carried out. Two counterfactual events are undertaken. The first experiment involves the appreciation of US$ against the S$ by 10% since 1986. The other experiment deals with the increase of government expenditure by another 10% since 1986. Interestingly, it is found that 1. both events, which are favourable to Singapore economy, resulted in higher predicted GDP growth rates as compared to growth rates without the experiments. 2. the new predicted growth rate of the quantitative model is lower than the actual growth rate while those of the three versions of the mixed model are higher. Thus, it can be deduced that the mixed model captures the favourable impact of experiments better. As for the prediction of Singapore GDP growth rates, the quantitative model forecasts 7% and 5.5% for 1993 and 1994 respectively. The three versions of the mixed model, on the other hand, forecast 8.6-8.9% and 7-7 .2% respectively. Judging by the better performance of the mixed model and the official estimates of 9% and 6-8% for 1993 and 1994, it can be concluded that Singapore will grow by about 8.8% in 1993 and 7.1% in 1994. The study on the incorporation of survey data in econometric modelling is a relatively new field in Singapore and the rest of the world. With the need to complement and supplement quantitative data, it is foreseeable that there will be great potential in this area in the near future.