In recent years, China has been plagued by overcapacity problem and zombie enterprises, which are seriously harmful to sustainable economic development and the stability of financial system. More importantly, it slows down the process of economic restructuring. But in fact, the problem of zombie enterprises is not a unique phenomenon to China and is also not a new word. The Japanese economy had experienced rapid growth of zombie firms in the last decade of the 1990s. This circumstance aroused widespread interest of scholars, leading to a growing number of researches. Caballero, Hoshi and Kashyap(2008)proposed CHK” model by calculating a hypothetical risk free interest payment”. Later, Fukuda and Nakamura(2011)developed a more accurate FN-CHK” model by adding two indicators, namely profitability criterion” and evergreen lending criterion”. However, FN-CHK” model has two defects if it is directly applied to China. On the one hand, it totally overlooks the significant role of government subsidies, which is especially serious in China. On the other hand, FN-CHK” model regards the interest rate of convertible bonds as the unique criterion when calculating the hypothetical risk free interest payment”. As a consequence, this model has a tendency to underestimate the number of zombie firms. This paper attempts to improve FN-CHK” model in several aspects, for instance, improving calculation method of the risk free interest payment”, and taking into account the dependence of government subsidies, and then identify zombie enterprises by using the data of Chinese listed companies from 2010 to 2016. This paper shows that, the rate of zombie companies was around 3.3% in 2011 and 2012, while it rose to 5% after 2013 and slowed down in 2016. This tendency could be related with different features of the early and late periods of the prolonged recessions. In the earlier stage(between 2011 and 2013), economy experienced an unexpected downturn and fewer companies could adjust their management tactics. As a result, the number of zombie enterprises was growing. In the later period(between 2013 and 2016), however, the economic decline was much more modest and more firms had adjusted their management strategies. Consequently, the number of zombie enterprises declined. Compared with our improved recognition model, FN-CHK” model misses 16.59% of actual zombie firms. Using unique indicator when calculating the risk free interest payment” and ignoring government subsidies have equal weighting in missing samples. It demonstrates that, our proposed model can improve the accuracy of recognition effectively, while directly using FN-CHK” model might lead to error. We also find that, the proportion of zombie companies in western region is the highest, followed by central and eastern regions. The rates of zombie companies in traditional industries, like iron industry and papermaking industry, are the highest. It is remarkable that the rates of zombie companies in these traditional industries decreased the most in 2016. It might be attributed to a big rise in commodity prices and the success of dealing with zombie enterprise problem in iron industry in 2016. Previous studies show that the increasingly serious zombie problem might be blamed on the overcapacity caused by the 4 trillion yuan stimulus plan”. Using statistical analysis and DID” regression method, this paper confirms that, the over-rapid growth of investment in fixed assets caused by 4 trillion yuan stimulus plan” might be responsible to the increasing rate of zombie enterprises after 2013. Besides, we also find that, the proportion of zombie state-owned companies is much higher than others. Moreover, this proportion is much higher in the areas where the state seriously intervenes the market. The most frequently used method of receiving financial assistance is evergreen lending, which is followed by interest discount and government subsidies. Because of the harmfulness of zombie firms, precise prediction is extremely important in prevention. We attempt to predict zombie enterprises by using Panel Logit model, the correct forecast rates in sample and out of sample are 88.57% and 96.58%. These results have significance to preventing and controlling zombie enterprises.