Abstract: This article collects a ries quarterly data of China’s GDP from 1992 to 2010, and we u the method of factor decomposition to collect the long-term increasing trend and asonality, then u ARMA model to fit the residuals, do analysis to get the final model and u it to generate a short-term GDP-forecast of china.
Key words: factor decomposition; ARMA model; GDP forecast；
1. Introduction
1.1Background
From 1978, since the reform and opening up, china’s economy is developing rapidly and steadily. After joining the WTO, the developing speed has reached a new level. GDP (Gross Domestic Product) , which is the basis of national economic production of statistical indicators, can be ud to reflect a country’s economy. It is the core of Statistical indicators in the national economy. GDP combines respons of the most basic aspects of macroeconomic, can not only measure the overall national output and income scale, but als
o can explore the economic fluctuations and cycles. Hence, it is of great importance to fit and analyze GDP accurately for exploring a country’s macroeconomics trend. The aim of this article is to generate a GDP forecast model and u it to predict the future GDP of china.
1.2 Method
A lot of methods have been ud to analysis economy phenomenon, time ries analysis is one of the most efficient methods. A time ries is a collection of obrvations of well-defined data items obtained through repeated measurements over time. Time-ries methods u economic theory mainly as a guide to variable lection, and rely on past patterns in the data to predict the future. An obrved time ries can be decompod into three components: the trend (long term direction), the asonal (systematic, calendar related movements) and the irregular (unsystematic, short term fluctuations). When the factors occur, we can u the method of decomposition, from which can we collect uful information of the data, we defined it as factor decomposition here.
The trend component typically reprents the longer term developments of the time ries of interest and is often specified as a smooth function of time.
The recurring but persistently changing patterns within the years are captured by the asonal component. It is quite common in economic time ries, when it occurs, we should u asonal adjustment method. Seasonal adjustment is the process of estimating and then removing from a time ries influences that are systematic and calendar related. Obrved data needs to be asonally adjusted as asonal effects can conceal both the true underlying movement in the ries, as well as certain non-asonal characteristics which may be of interest to analysts.
The irregular component reprents the irregular fluctuations which are affected by causal factors. It usually defined as residual. Considering the Insufficiency of the deterministic decomposition, we should test the residuals, if there is no autocorrelation among the residual, it means that the information of the time ries is totally recovered by the deterministic decomposition.
In the ca of the existence of autocorrelation, ARMA model can be ud to fit the residuals. ARMA is a one of tho most common time ries model which was ud to make preci estimation according to short term data. Its main idea can be concluded as a combination of veral time-related components which can be ud to predict the future data. The time ries components from the ARMA model is a t of random variables which related to time itlf, which shows uncertainty when obrved individually combined with each other shows some kinds of regularity and can be expresd by corresponding statistical model. The ARMA model consists of two parts, an autoregressive (AR) part and a moving average (MA) part. The model is usually then referred to as the ARMA(p,q) model where p is the order of the autoregressive part and q is the order of the moving average part.
2. Data Analysis
2.1 Datat
The data we collected contains historical GDP from 1992-2010, the reason we choo
this time duration rather than the 1978-2011 which most other prediction article would like to u is that during the first 10-15 years the economic growth rate is relatively slow compared with the later year’s (1990-now) growth. So we would like to wipe out the interference of the early data. Another reason we u recent years data (1992-2011) is that it is hard for us to look for the quarterly GDP data before 1992 due to the imperfection of the statistical system of China in the end of 20th century.
Table 1 quarter GDP data from 1992-2010 (Unit: 1000 million CNY)
time | GDP | time | GDP | time | GDP | Time | GDP |
1992.1 | 4974 | 1997.1 | 16257 | 2002.01 | 25376 | 2007.01 | 54755.9 |
1992.2 | 6358 | 1997.2 | 18697 | 2002.02 | 27965 | 2007.02 | 61243 |
1992.3 | 7119 | 1997.3 | 19148 | 2002.03 | 29716 | 2007.03 | 64102.2 |
1992.4 | 8472 | 1997.4 | 24871 | 2002.04 | 37276 | 2007.04 | 85709.2 |
1993.1 | 6500 | 1998.1 | 17501 | 2003.01 | 28861.8 | 2008.01 | 66283.8 |
1993.2 | 8044 | 1998.2 | 19722 | 2003.02 | 31007.1 | 2008.02 | 74194 |
1993.3 | 9048 | 1998.3 | 20372 | 2003.03 | 33460.4 | 2008.03 | 76548.3 |
1993.4 | 11742 | 1998.4 | 26807 | 2003.04 | 42493.5 | 2008.04 | 97019.3 |
1994.1 | 9065 | 1999.1 | 18790 | 2004.01 | 33420.6 | 2009.01 | 69816.9 |
1994.2 | 11085 | 1999.2 | 20765 | 2004.02 | 36985.3 | 2009.02 | 78386.7 |
1994.3 | 12447 | 1999.3 | 21859 | 2004.03 | 39561.7 | 2009.03 | 83099.7 |
1994.4 | 15601 | 1999.4 | 28263 | 2004.04 | 49910.7 | 2009.04 | 109599.5 |
1995.1 | 11858 | 2000.1 | 20647 | 2005.01 | 39117.4 | 2010.01 | 82613.4 |
1995.2 | 14110 | 2000.2 | 23101 | 2005.02 | 42795.2 | 2010.02 | 92265.4 |
1995.3 | 15535 | 2000.3 | 24340 | 2005.03 | 44744.4 | 2010.03 | 97747.9 |
1995.4 | 19291 | 2000.4 | 31127 | 2005.04 | 58280.4 | 2010.04 | 128886.1 |
1996.1 | 14261 | 2001.01 | 23300 | 2006.01 | 45315.8 | | |
1996.2 | 16601 | 2001.02 | 25651 | 2006.02 | 50112.7 | | |
1996.3 | 17671 | 2001.03 | 26867 | 2006.03 | 51912.8 | | |
| | | | | | | |
