Relevance or irrelevance of retention for dividend policy irrelevance

In an interesting recent paper, DeAngelo and DeAngelo (2006) highlight that Miller and Modigliani's (1961) proof of dividend irrelevance is based on the assumption that the amount of dividends distributed to shareholders is equal or greater than the free cash flow generated by the fixed investment policy. They claim that, if retention is allowed, dividend policy is not irrelevant. This paper clarifies and reinterprets DeAngelo and DeAngelo's result: Retention itself has not to do with dividend irrelevance, which holds even in case of retention. The key assumption has to do with the NPV of the extra funds (either retained or raised): If NPV is zero, dividend irrelevance applies. Yet, the dichotomy retention/no-retention is useful, because if agency problems are present, managers tend to retain funds and invest them in negative-NPV projects, and therefore the zero-NPV assumption must be removed, so that dividend irrelevance does not apply any more