In view of development policy, capital formation is crucial for economic growth in a country. Since the reform in 1978, China has enjoyed high saving rates. However, in the late 1990s the saving rates rose to such a high level, therefore the economic policy to promote consumption in order to reduce savings was suggested,. This paper attempts explains this paradox of "excessively high saving rates" in relation to the capital accumulation mechanism in rural enterprise (TVEs: Township and Village Enterprises). The determinants of household savings in developing countries is one important topic of the development research. However most studies have ignored investment as one main determinant. In the famous Lewis' model, the savings of farm households for non-farm investment are not considered, because in most cases it is difficult to sort out capitalists from the rural households. The case of the capital formation of TVEs in China offers us an opportunity to test the non farm investment as a saving incentive, because there are no large or small capitalists and landlords available after the agriculture reform in China in 1978. With facts and statistical analyses, this paper has proved that the development of TVEs is an important factor of the high saving rates of rural households since the reform. The way of development of private TVEs and the process of the privatization of the collective showed that owing to the extreme financial discrimination, TVEs underwent an evolution into a financial institution. The demand of TVEs for private credits forced rural households to save and hence function as an informal financial market. The managers have the motivation to save because they want to run a collective TVE or to set up a private TVE. The workers have the motivation to save because they want to get a job or keep their job in TVEs. Using first hand data and statistical analyses, it was plausibly proved that households with manager in TVEs have higher saving rates than non-manager households. It is especially obvious in households with manager in shareholding and shareholding co-operative TVEs. Households with peasant worker in collective TVEs have higher saving rates than others. This implies that collective TVEs used many methods to force their workers to save.