International Workshop on Consumption and Full Accounting of

International Workshop on Consumption and Full Accounting of the Household Sector 26 - 28 March 2012 Benchmarking of Korean Quarterly GDP Yungtaek Jung Director National Accounts Division

Contents I. Introduction II. Benchmarking Method III. Korean System of National Accounts IV. Benchmarking Results V. Conclusion

I. Introduction § There exists inconsistencies between ANA and QNA due to different source data and compilation methods. - ANA is based on Input-Output account or annual survey, whereas quarterly indicators or survey results are used in QNA. § Benchmarking is the adjustment process which combines QNA with ANA in consistent time series. § The main objective of benchmarking in NA is to preserve short-term movement of QNA and benchmark ANA for longterm trend.

II. Benchmarking Method § Benchmarking techniques can be categorized into two approaches. - Numerical approach: the family of Denton, Bassie. - Statistical modeling approach: Chow-Lin, GLS model, etc. § The proportional Denton is most widely used. - It is easier to implement. - It preserves the short-term movement of original series. ⇒ It avoids the step problem. § IMF recommends the proportional Denton in QNA manual.

II. Benchmarking Method § Benchmarking process consists of distribution and extrapolation. - Distribution: alignment of quarterly data within annual benchmark - Extrapolation: adjusting data for recent quarters whose annual benchmark is not available. § In this study, we apply proportional Denton to the Korean value added for electricity, gas, and water supply. Also, we compare the benchmarked results with pro-rata distribution.

II. Benchmarking Method : pro-rating § Pro-rating distributes the annual total according to the proportions of quarterly values within a year. : benchmarked value for quarter q of year β; : quarterly indicator or estimates before benchmarking; : annual benchmark for year β; : annual BI(Benchmark to Indicator) ratio. § It is simple but it causes a break between years.

II. Benchmarking Method : proportional Denton § Proportional Denton minimizes differences between neighboring quarterly BI ratios. Subject to § It preserves the original quarterly growth rates well. So it avoids the step problem which is occurred in prorating. § However, due to the smoothing process, the revision span becomes longer compared to pro-rata distribution.

II. Benchmarking Method : extrapolation § § For recent quarters whose annual benchmark is not available, quarterly values can be extrapolated using the last quarterly BI ratio. This extrapolation method is equivalent to using the growth rates of quarterly indicator.

II. Benchmarking Method : extrapolation § § The enhanced version of proportional Denton incorporates the forecasted value of annual BI ratio into benchmarking process. Constraint 1: 2: Where is the quarterly BI ratio; is the forecasted value of annual BI ratio. This method is efficient when annual BI ratios follow systematic pattern.

III. Korean System of National Accounts Estimates Release Schedule Contents Quarterly advance around 25 days after the quarter constant prices only. GDP, GDI Quarterly preliminary around 70 days after the quarter current & constant prices. GDP, GNI, gross savings Annual preliminary around 3 months after the year NDI and its appropriation, product & income accounts Annual final around 15 months after the year In addition to the contents of annual preliminary, capital accounts

IV. Benchmarking results : pro-rating § § The annual BI ratio of the value added for electricity, water, and gas is 1. 01 for year 2008. Pro-rata distribution causes a break between the 4 th quarter of 2007 and the 1 st quarter of 2008.

IV. Benchmarking results : proportional Denton § § The proportional Denton reduces size of the break. Since it smoothens quarterly BI ratios between two adjacent years. Instead, the quarterly values for year 2007 are revised even though annual estimate is not revised. In pro-rating: 30. 6%

IV. Benchmarking results : step problem § § The step problem can distort seasonality of original series. The seasonal factor of 1 st and 4 th quarter have been changed when distributed by pro-rata.

IV. Benchmarking results : extrapolation § § § We compare two types of extrapolation methods, i. e. , the proportional Denton and the enhanced version. For the enhanced version, we need to forecast the annual BI ratio. However, we do not have enough information on the behavior of annual BI ratios. So, we use the number(=1. 01) in 2008 for the forecast in 2009. The real annual BI ratio in 2009 is 1. 002 which is quite different from our forecast. We compare two types of extrapolated results with the distributed values in 2009.

IV. Benchmarking results : extrapolation § In the enhanced version, the size of revision is larger than proportional Denton. This is because we incorporate imprecise forecast.

V. Conclusion § § § In this study, we apply the proportional Denton to the Korean value added for electricity, gas, and water supply. As a result, the proportional Denton alleviates the step problem which can be found in pro-rata distribution. Therefore, it preserves original quarterly growth rates very well. From the extrapolation results, we can see that it is crucial to forecast annual BI ratio precisely in order to apply the enhanced version of proportional Denton. ank a n he § compilation of Korean GDP from 2014. Since then, we will also change the revision policy of original series.