CHINA S MANAGERIAL LABOR MARKET performing firms to be auctioned off, but also subjected them to more frequent review and larger security deposits. Conversely performers were given more autonomy, in the sense of ha longer-term managerial contracts and smaller security deposits ver, it is interesting to note that poorly performing firms by this measure did not have a disproportionate number of their managers demoted poor performance was not a significant explanatory vari able in explaining demotions In addition to examining the firms'ex ante performance relative to industrial averages and relating this performance to various mana genial selection variables, we compared each firm's performance be fore and after the change of manager. Significant improvement of firm performance after a managerial change can provide information about the former manager's performance that is not apparent by looking directly at the performance of the firm under the former manager. Poor performance of a firm, relative to the industry, for example, may be the result of bad luck or poor market conditions, a poor physical plant, low-quality labor and other inputs, or poor nanagement. However, an improvement in performance may reveal he existence of unfulfilled potential of the firm prior to the manage rial change. Thus ex post improvement is potential evidence of ex nte poor managerial performance. While ex ante poor performance may not be observable by outside observers(such as ourselves)view- ing only simple indicators of performance, it is plausibly observable by industrial bureaus, which have significant local knowledge of their Thus it is reasonable to suppose that an improvement in performance accompanying a change of manager is perhaps partly the result of the bureau's observance that the firm is performing poorly, relative to its potential (as perceived by the bureau), and then th ment of a more competent or energetic manager However, since a managerial change, as we measure it, is accom panied by a new managerial contract, a significant improvement in performance might also be expected when the old manager is reap ointed or even promoted, since managerial incentives are presum ably improved under the new contract. Thus, although ex ante poor performance is not associated with demotion, if the hypothesis is true that managers were demoted be- cause their performance was poor relative to their firm's potential then we would expect to see performance improve after the change of managers for these firms. This hypothesis is not contradicted by our data. Table 2 shows that a significant increase in performance occurred under a new manager if the previous manager was de moted. In fact, relative performance increased under a new manager