With the revolution of information technology, digital technologies are being embedded into the process of products and services with unprecedented breadth and depth. It not only penetrates into the value chain of product design, manufacturing and marketing, but also promotes the digitization, networking and intelligent development of manufacturing industry, and fundamentally changes the innovation ecosystem of manufacturing enterprises. The integration of digital technology and industrial development has gone beyond the traditional innovation theory. However, there are relatively few studies discussing the innovation activities of manufacturing firms in the new scenarios of innovation ecosystem. Whether Chinese manufacturing enterprises can achieve a successful transformation by the application of digital technology? How does the digital technology affect enterprise innovation? And what is the mechanism of its effect? 
   With the increasing popularity of digital technology, a growing body of literature has analyzed the role of digital technology in manufacturing transformation and upgrading. Due to the limitation of data, these studies generally adopt the methods of qualitative discussion and case analysis, and there is no empirical evidence to validate the innovative effect of digital technology. Fortunately, the nationwide projects of smart city construction in China, which started in 2012, can be regarded as an exogenous event for manufacturing firms to apply digital technology. Therefore, we define a time dummy variable to indicate whether the time is before or after the intervention of smart city construction, and divide manufacturing enterprises into two groups according to the digital intensity. Then we design the quasi-natural experiment method to carry on the empirical analysis. 
   Specifically, this paper investigates the effects of digital technology application on enterprise′ R&D investment and innovation output by using the difference-in-difference (DID) and quasi difference-in-difference（QDID）method. The method can effectively identify causal relationships without suffering from selection bias. We employ an unbalance panel data of Chinese A-share listed manufacturing enterprises of the Shanghai and Shenzhen Stock Exchanges from 2007 to 2017. We use the variables R&D intensity and intangible assets ratio, which define as the ratio of R&D expenditures to the annual income of a firm and the ratio of intangible assets to total assets, represent the explained variables of R&D investment and innovation output. According to previous studies in the field of innovation, the control variables include enterprise size, enterprise age, return on assets, Tobin′s Q, the ratio of liability to total asset, the ratio of fixed assets to total assets, share of major shareholders and ownership. We also control the industry fixed effect and time fixed effect, which capture the time invariant characteristics of manufacturing industry and trend characteristics, respectively. Our financial data are collected from the China Stock Market and Accounting Research (CSMAR) database.
   Firstly, this paper displays descriptive statistics for two group of manufacturing enterprises. According to descriptive statistical, there are grouping differences of the two variables —— R&D intensity and intangible assets ratio —— between treatment enterprises and control enterprises. The grouping differences may be due to the exogenous impact of digital technology, or the differences of enterprise characteristics. We need to further investigate whether the exogenous event incurs the increase of R&D investment and intangible assets ratio of enterprises or not.
Secondly, we use the univariate DID estimation to test for the grouping differences of enterprise innovation. Specifically, the mean values of enterprise innovation are calculated across samples belonging to the treatment group versus the control group over the sample periods before versus after the intervention of smart city construction. Judging by the testing results for the DID estimation, the application of digital technology has a statistically significant effect on R&D intensity at the 5% significance level, while the effect on intangible assets ratio is insignificant at 1% level. In addition, the null hypothesis of no difference between the two groups is rejected, and thus may distort the evaluation of causality. 
   Thirdly, in order to identify the causality of digital technology and enterprise innovation, this paper introduces regression-based DID method to control enterprise characteristics. In terms of enterprises′ R&D intensity, the regression coefficients are significantly positive at the 1% level, indicating that the application of digital technology has a significant effect on enterprises′ R&D investment. In terms of enterprises′ intangible assets ratio, the coefficient of DID model is not significant and the T value is 1.38, while the coefficient of QDID model is significantly positive at the level of 10%. It indicates that the digital technology has significantly positive effect on the innovation output of treated enterprises, and this conclusion is not the same as univariate DID estimation. The robustness test shows that the above conclusion is still true. There is strong evidence that the digital technology application has significant effect on enterprise innovation.
   Fourthly, this paper divides the combination of digital technology and manufacturing industry into two types: the integration mode of machine coupling and substitution and the collaborative mode of group information interaction. We investigate the influence of the integration mode and the collaborative mode on enterprise innovation, respectively. For the integration mode, the effect of digital technology on R&D investment is significant, while the effect on innovation output is not significant. More time is needed to exert an effect on innovation output for intelligent manufacturing. For the collaborative mode, both the effects on R&D investment and innovation output are significant, the digital technology has promoted the prosperity of product innovation and business innovation. 
    Finally, this paper divides our sample firms into two groups of state-owned enterprises (SOEs) and non-state-owned enterprises (non-SOEs) and examine the treatment effect separately for each of the two subsamples based on DID regression. We also divide our sample firms into three groups according to firm size and investigate the heterogeneity effects. It can be concluded that there are heterogeneity effects of digital technology on enterprise innovation. Because of the advantages of digital resources, the treatment effects are significantly larger for SOEs and for large-sized enterprises.
    Our findings are enlightening for enacting better policies involving digital technology. For example, supporting policies for intelligent manufacturing, encouraging technological innovation of manufacturing enterprises, and reducing the digital divide of non-SOEs and private enterprises.