By knowing the state of the economy and projections of future conditions, Bank Indonesia can determine the appropriate monetary policy response. In the past decade, nowcasting has been developed to assess the current state of the economy. The difference is that the coinciding economic index is a composite of several indicators with similar dynamics to the reference series, while nowcasting is an estimation of the size of the reference series using representative indicators.
Sørensen (2011) argues that given the large contribution of household consumption to GDP, household consumption is an important factor in assessing the state of the economy. The bridge equation is a linear regression that correlates high-frequency variables (for example, monthly retail sales) with lower-frequency variables (for example, quarterly GDP), estimating the most recent economic conditions prior to the release of the quarterly data. The aggregation process involved in the Bridge Equation, where monthly indicators are aggregated into quarterly data, can result in some of the important information being lost from the data.
Using a factor model, the unobserved state of the economy is extracted and a new coincident indicator is constructed. A detailed description of the methodology is presented in Bańbura, et al. 2008) and Foroni and Marcellino (2013), devised the approach, bridging with factors, as the (quarterly) reference series finds regression with the (monthly) common factors as a form of bridging equation. Time aggregation is applied by matching the quarterly data to that of the third month in the respective quarter, while the first and second month data are calculated using the Kalman filter.5 The next step is to project, for example, the reference series. the quarterly GDP series, based on the estimated common factors converted to quarterly data.
The first stage involved estimating representative state space parameters using principal components (from the balanced panel of monthly indicators10), thereby estimating the magnitude of common factors.
Stages
Indicator Selection
Filtering and Transformation
This approach is appropriate if similar seasonal factors are found in the reference series and the component series. Estimates can be compromised, however, if seasonal factors are different in the reference series and the component series, leading to less consistent estimates. Considering that the household consumption and investment data published by BPS-Statistics Indonesia are not seasonally adjusted, seasonal adjustments are preferred in this research.
The correlation between the reference series and the component series (indicators) depends on whether the indicators are state data or flow data, and on how the indicators are transformed before they are entered into the model (for example, stationarity requirements) as suggested by Bańbura, et. al. Data in percentages were converted into a difference compared to the previous year.
Nowcasting Exercise
Model Performance Evaluation
RESULTS AND ANALYSIS
- Indicator Selection
- Indicators of Household Consumption
- Indicators for Investment
- Nowcasting Exercise
- Household Consumption
- Investment
- Model Performance Evaluation
Of the various candidate indicators mentioned, several indicators with the highest correlation coefficient were selected. Based on these criteria, the following indicators were selected: cement sales, motor vehicle production, consumer confidence index - state of the economy, industrial production index, electricity consumption, exports YTD, imports YTD, credit (working capital, investment and total), lending rates (working capital and investment), M1 and the nominal effective exchange rate (NEER). The results of the ten best indicator combinations in the first, second and third months are presented in Table 5.
The exercise produced a composite set of the best indicators, namely motor vehicle sales, total down payment, consumer loan interest rates, M1 and the Rupiah exchange rate (NEER). Based on the RMSE, the performance of the best combination of indicators in the first month was no more robust than several other combinations of indicators, but in the second and third months that combination ranked first. The selection of the best combination showed that the five indicators mentioned above showed the closest resemblance to household consumption in terms of common factors.
However, the results of the exercise shown in Table 5 also show that the difference in performance between the top 10 indicator combinations was almost negligible, evidenced by the small difference in RMSE. In fact, the results in the third month of the second quarter were no more accurate than a month ago. The deviation then fell sharply in the second month with the inclusion of the latest auto sales data.
In addition, however, the inclusion of the latest NEER data, total deposits, consumer loan rates and M1 in the third month improved nowcasting results. Data Release Nowcasting Result [and error/. Similar to household consumption, the exercise tested different combinations of the 11 selected indicators. The results of the ten best indicator combinations in the first, second and third months are shown in table 8.
The exercise produced a series of better indicator components, namely cement sales, motor vehicle production, electricity consumption, credit surplus and M1. However, the results of the exercise presented in Table 4.8 also show that the difference in performance between the 10 best indicator combinations was almost negligible, evidenced by the small difference in RMSE. When investing now, the indicators of cement sales, motor vehicle production and electricity consumption are almost always selected as components of the best combination.
In the third month, however, current times forecast results improved in the first and second quarters, but actually worsened in the third and fourth quarters. As presented in Table 11, the comparison of the accuracy of the model for forecasting household consumption showed that the prediction error of the dynamic factor model was smaller than the prediction error of the bridging equation and.
CONCLUSION 5.1. Conclusion
Recommendations
Therefore, the nowcasting of household consumption using the DFM model turned out to be the most robust. On the other hand, the comparison of model accuracy for nowcasting investments showed that the prediction error of the Dynamic Factor Model was smaller than the prediction error of the Bridge Equation and ARIMA, as shown in Table 4.12. The prediction error of the DFM model was considered significant, but smaller than that of the benchmark models.
Short-Term Eurozone GDP Growth Forecasts. ECB Working Paper Series, No. Borage; Diebold, Francis X.; and Scotti, Chiara. A look into the factor model black box: release lags and the role of hard and soft data in GDP forecasting.
