Using overall institutional holdings as the proxy for tax-exempt investors, many existing literature has found mostly unfavorable evidence against dividend clientele theories. This paper categorizes institutions into two types, tax-exempt and non-tax-exempt institutions, to test dividend clientele theories. By examining the changes of institutional holdings after extreme dividend policy changes, we find that both tax-exempt and non-tax-exempt institutions tend to increase (decrease) holdings after firms initiate (omit) dividends. Moreover, when firms substantially decrease dividends, tax-exempt institutions are likely to reduce their holdings, while non-tax-exempt institutions tend to increase holdings. Our findings suggest that the traditional method using all institutional holdings as a proxy of tax-exempt institutional ownership may be problematic. In addition, because non-tax-exempt institutions prefer dividend-paying stocks to non-dividend-paying stocks, but favor stocks paying fewer dividends to more, so the traditional cross-sectional regressions of firm dividends neglect the fact that different dividend levels can have opposite impacts on non-tax-exempt institutions' preferences. This may partly generate the current unfavorable empirical evidence of the dividend clientele theory.