在创新越来越成为中国经济增长的新动能、产业政策被越来越广泛使用的今天，已有文献对政府补助对企业创新的影响并未形成明确结论。本文以全部A股上市公司2010-2021年的面板数据为样本，在剔除财务异常企业与金融类企业后开展实证研究，并创新性地以关键词筛选的方式提升政府补助数据质量，并将创新活动性质（利用性创新和探索性创新）纳入到研究当中。实证结果表明，从总量角度看，政府补助提升了企业的创新产出数量；从结构角度看，政府补助使得企业利用性创新成果占比提升，对于探索性创新成果占比则形成了压制。本文认为这与两类创新活动的特点以及企业开展创新活动的动机有关：一方面，利用性创新是可参考知识较多、确定性较强的创新，其产出与研发资源投入相关性更强，政府补助作为外部的资源流入，较好地促进了该类创新活动；探索性创新则是可参考知识较少、不确定性更大的创新，其产出受不可控因素影响大，除资源投入外更需机遇加持，政府补助作用相对有限。另一方面，企业也可能出于迎合政策部门与监管的动机而更多开展易取得成果的利用性创新活动，同样促进利用性创新占比上升而探索性创新占比下降。稳健性检验的结果与基准回归一致。本文针对遗漏变量、反向因果、样本选择三类可能的内生性问题进行了处理。针对遗漏变量问题，本文添加了更多控制变量与固定效应，实证结果未发生变化；针对反向因果问题，一方面本文使用滞后期的自变量予以缓解，另一方面也使用行业内其他公司接受政府补助的相对均值作为工具变量，通过两阶段最小二乘法予以缓解；针对样本选择问题，本文使用Heckman两阶段方法进行检验，结果表明样本存在一定的自选择偏误，但在修正后结果与基准回归模型保持了一致。 异质性分析的结果表明，政府补助同时促进了国有企业与民营企业创新产出总量的上升，但相对民营企业，国有企业迎合政策部门与监管的动机更强，从事高质量创新活动的激励相对较弱，表现为利用性创新的占比上升与探索性创新的占比下降，民营企业则未表现出此特点。本文提出的分析框架对强研发属性的企业解释力更强，而政府补助无论从总量还是结构来看，都对弱研发属性的企业无显著作用。 机制检验显示，政府补助降低了企业风险承担水平、提升了企业对创新失败的容忍度，这是其能够提升企业创新产出总量的一条内在机制。

In today‘s world where innovation is increasingly becoming a new driving force for China‘s economic growth and industrial policies are being widely used, there is no clear conclusion on the impact of government subsidies on corporate innovation in existing literature. This article takes panel data of all A-share listed companies from 2010 to 2021 as a sample, conducts empirical research after excluding financial abnormal enterprises and financial enterprises, and innovatively improves the quality of government subsidy data through keyword screening, and incorporates the nature of innovative activities (exploitative innovation and exploratory innovation) into the research.The empirical results indicate that from a total perspective, government subsidies have increased the number of innovative outputs of enterprises; From a structural perspective, government subsidies increase the proportion of exploitative innovation achievements in enterprises, while suppressing the proportion of exploratory innovation achievements. This article believes that this is related to the characteristics of two types of innovation activities and the motivation of enterprises to carry out innovation activities: on the one hand, exploitative innovation is an innovation with a lot of reference knowledge and strong certainty, and its output is more correlated with R&D resource investment. Government subsidies, as external resources, effectively promote this type of innovation activity; Exploratory innovation refers to innovation with less reference knowledge and greater uncertainty. Its output is greatly affected by uncontrollable factors, and it requires more opportunities in addition to resource investment. The role of government subsidies is relatively limited. On the other hand, enterprises may also engage in more exploitative innovation activities that are easy to achieve results in order to cater to policy departments and regulatory motivations, which also promotes an increase in the proportion of exploitative innovation and a decrease in the proportion of exploratory innovation. The results of the robustness test are consistent with the benchmark regression.This article addresses three possible endogeneity issues: omitted variables, reverse causality, and sample selection. Regarding the issue of omitted variables, this article added more control variables and fixed effects, and the empirical results remained unchanged; For the reverse causality problem, on the one hand, this article uses lagged independent variables to alleviate it, and on the other hand, it also uses the relative mean of government subsidies received by other companies in the industry as instrumental variables to alleviate it through two-stage least squares method; Regarding the issue of sample selection, this article uses the Heckman two-stage method for testing, and the results show that there is a certain degree of self selection bias in the samples. However, after correction, the results are consistent with the benchmark regression model. The results of heterogeneity analysis indicate that government subsidies simultaneously promote an increase in the total innovation output of both state-owned and private enterprises. However, compared to private enterprises, state-owned enterprises have a stronger motivation to cater to policy departments and regulation, and the incentives to engage in high-quality innovation activities are relatively weak. This is reflected in an increase in the proportion of exploitative innovation and a decrease in the proportion of exploratory innovation, while private enterprises do not exhibit this characteristic. The analytical framework proposed in this article has stronger explanatory power for enterprises with strong R&D attributes, while government subsidies have no significant effect on enterprises with weak R&D attributes, both in terms of total amount and structure. Mechanism testing shows that government subsidies reduce the level of enterprise risk-taking and increase the tolerance of enterprises for innovation failure, which is an inherent mechanism that can enhance the total innovation output of enterprises.