Non-linear relationships are common in economic theory, and such relationships are also frequently tested empirically. We argue that the usual test of non-linear relationships is flawed, and derive the appropriate test for a U shaped relationship.Our test gives the exact necessary and sufficient conditions for the test of a U shape in both finite samples and for a large class of models. JEL: C12, C20

Countries rich in natural resources constitute both growth losers and growth winners. We claim that the main reason for these diverging experiences is differences in the quality of institutions. More natural resources push aggregate income down, when institutions are grabber friendly, while more resources raise income, when institutions are producer friendly. We test this theory building on Sachs and Warner's influential works on the resource curse. Our main hypothesis -that institutions are decisive for the resource curseis confirmed. Our results contrast the claims of Sachs and Warner that institutions do not play a role.One important finding in development economics is that natural resource abundant economies tend to grow slower than economies without substantial resources. For instance, growth losers, such as Nigeria, Zambia, Sierra Leone, Angola, Saudi Arabia and Venezuela, are all resource-rich, while the Asian tigers: Korea, Taiwan, Hong Kong and Singapore, are all resource-poor. On average resource abundant countries lag behind countries with less resources. 1 Yet we should not jump to the conclusion that all resource rich countries are cursed. Also many growth winners such as Botswana, Canada, Australia, and Norway are rich in resources. Moreover, of the 82 countries included in a World Bank study, five countries belong both to the top eight according to their natural capital wealth and to the top 15 according to per capita income (World Bank, 1994).To explain these diverging experiences this article investigates to what extent growth winners and growth losers differ systematically in their institutional arrangements. As a first take we plot in Figure 1 the average yearly economic growth from 1965 to 1990 versus resource abundance in countries that have more than 10 % of their GDP as resource exports. In our data set this group consists of 42 countries. Panel (a) is based on data from all 42 countries and the plot gives a strong indication that there is a resource curse. In panel (b) and (c), however, we have split the sample in two subsamples of equal size, according to the quality of institutions (a measure to be discussed below). Now the indication of a resource curse only appears for countries with inferior institutions -panel (b); while the indication of a resource curse vanishes for countries with better institutions -* We are grateful to two anonymous referees and editor Andrew Scott for their constructive suggestions. We also thank Jens Chr. Andvig, Carl-Johan Lars Dalgaard, James A. Robinson and a number of seminar participants for valuable comments.1 This is documented in Sachs and Warner (1995, 1997a, b), Auty (2001. See also Gelb (1988), Lane and Tornell (1996) and Gylfason et al. (1999). Stijns (2002), however, argues that these results are less robust than the authors claim. It should furthermore be noted that concerns about specialising in natural resource exports was raised by economists well before the recent resource curse literature. Notably, Raol Prebisch and Hans Singer argued more ...

1 We are grateful to two anonymous referees and editor Andrew Scott for their constructive suggestions. We also want to thank Jens Chr. Andvig, Carl-Johan Lars Dalgaard, James A. Robinson and a number of seminar participants for valuable comments.