產出與政府支出彼此間的關係是很密切的。它們之間的因果關係，無論在理論或實證上都是值得學者與執政當局的重視。在理論上，當政府支出會影響產出時，文獻稱其為凱因斯理論；而當產出會影響政府支出時，文獻上則稱為華格納假說。然而既存文獻的實證結果卻歧異不一，這可能是因為所使用的方法、代理變數，或研究對象不同所致。本文利用實質人均國內生產毛額(real per capita GDP)作為產出的代理變數，政府消費支出作為政府支出的代理變數，研究對象為台灣、香港、新加坡、南韓，研究期間除了南韓是自1970年至2007年之外，台灣、香港、新加坡皆自1961年至2007年。在研究方法上，首先應用單根檢定法，來檢定變數是否為恆定的序列，然後再進行Johansen共整合檢定，以檢定變數間是否有長期均衡關係，接著進行Granger因果關係檢定。 實證結果顯示：所有變數皆為I(1)序列，除了香港沒有共整合之外，台灣、新加坡、南韓皆存在一個共整合的長期均衡關係。因此台灣、新加坡、南韓以VECM模型進行Granger因果檢定，實證結果顯示，台灣、新加坡皆為雙向因果關係；南韓則支持凱因斯假說。香港則以VAR模型進行Granger因果檢定，其結果為華格納法則成立 The relationship between output and government expenditure is very close. For researchers and policy makers, their causal relationship is very important, both theoretically as well as empirically. In theory, when the causation runs from government expenditure to output, the issue known as the Keynesian theory；when the causation runs from output to government expenditure, the issue known as the Wagner’s law. However, the empirical results are discrepancy in the existing literature. The major contributory factor to these conflicting findings is due to methodological differences, different kind of proxy variable and so on. This article applies the technigue of unit root, cointegration and Granger Causality for the countries of Taiwan, Hong Kong, Singapore and South Korea using annual data. The study period is from 1961 to 2007 for Taiwan, Hong Kong and Singpore and 1970-2007 for South Korea. The empirical results show that all the variable series are I (1) for four countries. The Johansen’s cointegration test indicates that there exist at least one cointegrating relationship between per capita GDP and government consumption expenditure in Taiwan, Singapore and South Korea. We carry out Granger-Causality test by vector error correction model (VECM) for these three countries. The results of the Granger-causality tests indicate that there exists an interdependency between the per capita GDP and government consumption expenditure with a bi-directional causal relationship for Taiwan and Singapore and the output of South Korea is supported the Keynesian’s principle. Because there is not cointegration relationship between per capita GDP and government consumption expenditure for Hong Kong, so we carry out Granger-causality test by vector autogression (VAR) model. The result of Granger-causality test indicates that the Wagner’s law exists in Hong Kong.
