Effects of Digital Transformation on Total Factor Productivity of Cultural Enterprises—Empirical Evidence from 251 Listed Cultural Enterprises in China

Abstract

By analyzing the sample data of 251 listed cultural enterprises’ annual reports from 2011 to 2019 in China, this study empirically analyzed the effects of digital transformation on the total factor productivity (TFP), paths of effects, and heterogeneity characteristics of cultural enterprises in China. The findings indicate that digital transformation can significantly increase the TFP of Chinese cultural enterprises. The results obtained through robustness tests and endogeneity questions verify the credibility of the findings. The paths of effects mainly include improving the efficiency of content creation, financing ability, and research and development (R&D) investment of cultural enterprises, and the boosting effect of digital transformation on TFP is more significant among enterprises that are smaller in scale, located in midwest China, and positioned within the upstream chain of a segmented industry.

1. Introduction

Digital transformation refers to the progressive upgrading process of enterprises in the digital mode of thinking, using big data and other digital technologies to change the traditional mode of production, change the old form of business organization, and produce digital products. As early as the 1980s, digital transformation, represented by Internet technology, penetrated most aspects of the American real economy [1]. Since Internet technology has continued to spread throughout the world, digitalization has rapidly become an indispensable and significant part of economic development [2]. However, academic debate has raged for a long time over how digital transformation affects industrial development. The classic Solow paradox was proposed by Robert Merton Solow in 1987, in which he explained that, although all sectors of society have been emphasizing the economic impact of the Internet, the real economy has not shown a significant increase in TFP [3]. Afterward, Gordon (2012) and Acemoglu et al. (2014), using American data to conduct empirical studies, found that emerging information technology did not increase real industries’ TFP. Thus, Solow’s economic paradox seems evident for a long time [4,5].

The academic community has attempted to clarify the content with respect to digital transformation. The complex relationship between digital transformation and economic performance, in particular total factor productivity (TFP), which measures the efficiency and sustainability of inputs and utilization of resources in the basic decision-making unit, must be clarified to reduce the risk of digital transformation in traditional enterprises [6,7,8,9], in which data elements can penetrate “information silos” within enterprises that influence their production and processing, business organization, and product sales and that enable enterprises to form close connections with consumers, other enterprises, governments, financial institutions, and other external entities. Given these factors, existing theories regarding digital transformation are inadequate for clarifying the path to such transformation for disruptive enterprises, and a new systematic analytical framework is required to analyze the effects of digital transformation on the TFP of enterprises.

Mainstream academics respond to Solow’s paradox as follows: On one hand, rapid information technology innovation has caused technological ruptures. However, the large investment in information capital has also squeezed out too much labor, so the indicators of TFP measured by the traditional method have shown insignificant growth. On the other hand, the rise of big data, artificial intelligence, and other new-generation Internet technologies has had a much more significant impact on the real economy, and the mechanism of TFP is much more complex, so it is no longer accurate to view the Internet solely as a means of improving communications infrastructure [10,11,12].

The digital transformation of cultural enterprises is profoundly shaping social development in China. On the whole, the Chinese cultural industry has maintained a high growth rate in the first two decades of the 21st century. Figure 1 shows that the Chinese cultural industry produced only CNY344 billion in 2004, accounting for only 2.13% of the national GDP. From 2004 to 2011, the total output value of the Chinese cultural industry increased steadily with an average annual growth rate of CNY143.414 billion, rising steadily from 2.13% to 2.76% of the GDP in that year. Since the digital transformation emerged in 2011, the annual output value of the Chinese cultural industry has rapidly accelerated from CNY134.79 billion in 2011 to CNY4501.6 billion in 2019, with an average annual growth rate of CNY384.929 billion.

Furthermore, digital transformation currently remains an emerging trend in the first two decades of the 21st century, and enterprises are still determining how they can change their production technology, organizational models, and product digitalization. The available practical data on the subject are insufficient for a detailed academic deconstruction. In addition, the meaningful conclusions that have been drawn from case studies are difficult to generalize to other business contexts because these studies have generally focused on individual enterprises. As a developing country with a large economy, digital transformation in China provides experiences for global enterprises, particularly for developing countries’ sustainable development. Cultural enterprises, which interpret human cultural heritage and implement digital transformations for the future, are closely related to sustainable development. Sustainable development after digital transformation requires enterprises to be more efficient in production, which is the key. Meanwhile, cultural enterprises are extremely important micro-components of modern economies, and with the help of digital elements, cultural enterprises can not only distribute their products more conveniently but also create more efficient content, which can propel the whole economy forward in a sustainable direction. Therefore, the present study investigated whether digital transformation can increase the TFP of Chinese cultural enterprises; how digital transformation affects the TFP of cultural enterprises if the answer to the preceding question is yes; the form of heterogeneity that the impact of digital transformation will exhibit after accounting for differences in enterprise size, geographic region, and industry segment; and the role of the external environment of cultural enterprises in regulating the effect of digital transformation.

To address the aforementioned questions, the present study used sample data from the 2011–2019 annual reports of 251 listed cultural enterprises in China to conduct theoretical analyses, mathematical modeling, and empirical tests to clarify the effects paths, and heterogeneity characteristics of digital transformation with respect to the TFP of cultural enterprises. The present study is innovative in this field because of the following reasons. (1) A unified theoretical framework is constructed for analyzing the effects of digital transformation on the TFP of cultural enterprises. The present study extends the classic Aghion–Howitt Schumpeterian innovation growth model by endogenizing data elements pertaining to government and financial institutions into the model system and analyzing the intrinsic mechanism by which digital transformation affects the TFP of cultural enterprises. (2) A general measurement method is constructed to quantify the digital transformation of cultural enterprises. Through the application of computer crawler technology, in-depth information on digital transformation is mined from the annual reports of listed cultural enterprises; thereafter, redundancy at the individual cultural enterprise level is removed, and the mined information is converted into quantitative assessment index data. Through data aggregation, the actual level of digital transformation of cultural enterprises can be determined, and therefore, the present study is a valuable addition to the current quantitative research on the economics of the cultural industry. (3) Through a quantitative analysis of how digital transformation affects the TFP of cultural enterprises in China, the present study obtains detailed empirical evidence that can guide future theoretical development.

The remainder of the present study is organized as follows. Section 2 introduces the conceptual model and hypothesis formulation, including theoretical framework and mathematical modeling. Section 3 describes the theoretical modeling and derivation process, including the construction of an econometric model, the measurement and description of variables, the data source, and the descriptive statistics. Section 4 covers the empirical tests, including full-sample regression, robustness tests, endogenous problem-solving, and heterogeneity analysis. Section 5 describes the tests conducted to clarify the mechanism of action. Section 6 presents the conclusions of the present study and various policy recommendations.

2. Theoretical Framework and Mathematical Modeling

2.1. Enterprise Digital Transformation

The China Statistics Bureau defines cultural enterprises as enterprises that offer cultural and culturally related products to the public in its latest Classification of Cultural and Related Industries (2018), including enterprises that provide news and information services, create and produce content, offer cultural communication channels, cultural entertainment, and leisure services, and manufacture cultural equipment and other related activities, which are classified into 9 major categories, 43 medium categories, and 146 subcategories. At the same time, this definition of cultural enterprises is global and consistent with international perspectives [13,14,15,16]. To select the research sample in the present study, cultural enterprises consistent with the China Bureau of Statistics’ definition of cultural enterprises were selected, such as Poly Culture Group Corporation and Bona Film Group Corporation.

Digital transformation refers to the progressive upgrading of enterprises in the digital mode of thinking, using big data and other digital technologies to change the traditional mode of production, change the old form of business organization, and produce digital products. With the help of emerging digital technology, enterprises can transform and store information from many sources, different periods, and different forms into unified data information, and then transform the former production mode, operation mode, and even the final product form using this new production factor [17,18,19].

2.2. Total Factor Productivity

Total Factor Productivity (TFP) is measured as the “residual growth that cannot be explained by changes in labour and capital inputs [20],” referring to the efficiency and sustainability of resource inputs and utilization in a basic Decision-Making Unit. TFP was first measured by Farrell (1957), who pioneered the idea of measuring production efficiency by the gap between actual output and optimal output [21]. In the same year, Solow offered the perspective on evaluating efficiency indicators by utilizing residual outputs that cannot be explained by inputs [22]. Several scholars have improved and created different TFP measurement methods based on this duo’s research since then. Farrell’s measurement idea was followed by some scholars, who continued to use linear programming knowledge in operations research to develop the non-parametric estimation of the data envelope analysis (DEA) method. Several scholars developed parametric stochastic frontier analysis (SFA) to improve Solow’s measurement concept. Afterward, subsequent scholars discovered that measurement accuracy is affected by whether the decision unit is a macro-dimensional or micro-dimensional concept. In 1996, Olley & Pakes developed the OP method in a semiparametric estimation framework to address the endogeneity problem in measuring micro-enterprises’ productivity [23]. Levinsohn & Petrin (2003) improved the method continuously to create the LP method [24]. By adding an intermediate goods indicator to the OP method, the LP method captures simultaneous adjustments with a more general indicator. However, as efficiency measure mentally penetrates the field of micro-data from the initial macro-data, scholars have found that neither the DEA method nor the SFA method can cope with the behavior of “simultaneous adjustment” in the production of the enterprise, and the estimation of efficiency values under traditional nonparametric and parametric frameworks has endogeneity problems, resulting in a significant reduction in accuracy. Due to the lack of research on OP and LP methods that are better suited to micro-scenarios, traditional measurements are somewhat less comparable and accepted.

2.3. Theoretical Modeling

The TFP of cultural enterprises in China has been estimated by several scholars using various efficiency evaluation tools and sample data of varying dimensions. Academics have identified three factors that influence as follows: financial constraints, government financial support, and regional economic development [25,26,27,28].

Under the theory of industrial integration, the phenomenon of cultural enterprises employing digital technology to alter their traditional production modes, business organizational structures, and production of digital products essentially represents an integration of the digital and cultural industries. Increasing industrial integration can enable cultural enterprises to acquire new and efficient production and innovation technologies and the technical means to convert positive externalities into real economic value on the cultural market, which is the process that increases enterprises’ TFP.

First, digital transformation can create additional economic value for an enterprise by introducing positive network externalities (network effects); improving its efficiency of factor use in production and operation, innovation and research and development (R&D), and market organization; and increasing its TFP. The production, operation, sales, innovation, and financing of cultural enterprises generate positive externalities. Through digital technology, content creation can be transformed from a traditional linear subproject model into a decentralized mesh platform structure [29]; this transformation can improve the efficiency of content creation and enhance the vitality of creative personnel and materials [30,31]. By establishing a larger market trading network, cultural enterprises can interact more extensively with cultural consumers and other enterprises and enhance their operational efficiency.

Second, Schumpeterian innovation theory emphasizes that R&D innovation drives enterprise growth. Through digital technology, cultural enterprises that focus on content creation can quickly convert their innovations into efficient means of production, thereby achieving higher market returns. Digital transformation encourages cultural enterprises to share their R&D knowledge externally, which enables them to contribute to the public knowledge pool and increase industry productivity [32].

Third, according to the new financial intermediation theory, banks and other financial institutions can provide enterprises with external financial support and can diversify the operational risks of financing enterprises by providing loans. The development of cultural enterprises is often limited by financing constraints. However, an open business model achievable through digital transformation can improve the standardization and transparency of the content creation process of cultural enterprises, reduce the difficulty of financing enterprises, and enhance financing efficiency.

Finally, regulation theory posits that if the market mechanism cannot effectively coordinate the long-term and short-term development goals of enterprises, the government must take proactive regulatory action to balance the long-term and short-term interests of enterprises. When cultural enterprises explore using digital technology, they must undertake the corresponding costs and risks associated with short-term changes; nevertheless, digital transformation can help them comprehensively achieve their long-term development and upgrading objectives. As cultural industry policies have been implemented, a considerable amount of social capital has been allocated to the digital transformation of cultural enterprises and the transformation of enterprises to attract government credit endorsements, which can in turn enhance the financing capacity and TFP of cultural enterprises.

In the present study, the current authors explored and analyzed how digital transformation affects cultural enterprises’ TFP by constructing a three-sector Schumpeterian innovation growth model. The present study was based on the Aghion–Howitt-type Schumpeterian innovation growth model proposed by Aghion & Howitt (1992) [33], with references to Levine (2005) and Aghion & Howitt (2008) to introduce financial institutions [34,35], while regarding Sun and Hou (2021) who introduced governmental subjects [36]. The logarithmic value of cultural enterprises’ TFP was estimated using two semiparametric estimation methods, LP and OP. The present study explored the micro level of cultural enterprises rather than the meso level, in light of previous studies that neglected industry heterogeneity and failed to observe the impact of digital technologies and evolving business models.

2.4. Mathematical Modeling

2.4.1. Production Function Setting

The present study developed a mathematical model for how digital transformation affects the TFP of cultural enterprises and proposes a hypothesis for testing. As a starting point, the abstract production function of the cultural enterprises is defined as follows:

$$Y_t=f\left(A_t,K_t,L_t\right)$$

(1)

where Y_t is the total amount of cultural products or services produced in period t, A_t is the TFP in period t, K_t is the total amount of capital invested in production, and L_t is the total amount of labor. In its linear form, this production function can be expressed as follows:

$$\ln Y_t={\alpha }_1\ln A_t+{\alpha }_2\ln K_t+{\alpha }_3\ln L_t$$

(2)

Technical efficiency is attached to labor in endogenous growth theory by Harrod’s neutral technical progress setting. By abstracting intermediate goods as a continuous unity model between [0, 1], the present study developed the micro-foundation of cultural enterprises based on the setting approach of Aghion & Howitt [33], and provided a C–D (Cobb–Douglas) form for the abstract production function for cultural enterprises. The equation is as follows:

$$Z_t={\displaystyle {\int }_0^1{\left(B_{it}{L}_{it}\right)}^{1-\alpha }{x}_{it}^{\alpha }}di$$

(3)

In the C–D production function, the subscript i indicates the ith cultural enterprise, Z_t is the sum of final products and intermediates produced by a cultural enterprise for consumption or scientific research; B_it is the technical efficiency under the production level at that time, which is a key parameter of TFP; x_it is the intermediates used in the current production; and α is the elasticity coefficient with a value ranging from 0 to 1.

Using the above Equation (3), cultural enterprises determine their overall output level by continuously integrating intermediate goods. The production function of the ith cultural enterprise is as follows:

$$Z_{it}={\left(B_{it}{L}_{it}\right)}^{1-\alpha }{x}_{it}^{\alpha }$$

(4)

Further, if the abstract production functions described in Equations (1) and (2) with the specific production functions described in Equations (3) and (4) are compared, in addition to the basic assumption that intermediate goods are transformed from inputs into final goods in a 1:1 ratio, there is a quantitative relationship between the total output of cultural enterprises (Y_t) and the production function (Z_t) [33,37]. The equation is as follows:

$$Y_t=Z_t-x_{\mathrm{t}}$$

(5)

In addition, the TFP (A_t), and technical efficiency level (B_it) can be interconverted. The equation is as follows:

$$A_t={\displaystyle {\int }_0^1{B}_{it}^{1-\alpha }}di-\frac{\partial Y_t}{\partial K_t}$$

(6)

By taking the first-order derivative of Equation (6), the derivative relationship can be obtained. As a result, the technical efficiency level parameter (B_it) positively determines the TFP (A_t). The equation is as follows:

$$\frac{d{A}_t}{d{B}_{it}}=\left(1-\alpha \right){\displaystyle {\int }_0^1{B}_{it}^{-\alpha }}di>0$$

(7)

2.4.2. R&D Innovation

R&D innovation within cultural enterprises is a risky process governed by certain laws. It is generally believed that the more an enterprise invests in R&D innovation, the higher the likelihood of R&D success; however, the more improvement it strives to achieve in R&D innovation, the more difficult R&D will be, and the lower the likelihood of success. According to the model setting, the equation is as follows:

$${\mu }_{it}=IN{T}_{it}\times g\left(\frac{R_{it}}{B_{it}^{\ast }}\right)$$

(8)

The abstract function g used in Equation (8) needs to satisfy the basic requirements of “g′ > 0” and “g″ < 0” at the same time, to ensure that the law of diminishing marginal returns is established. In Equation (8), μ_it represents the probability of success of R&D innovation in cultural enterprises, INT_it is the degree of utilization of the digital transformation by cultural enterprises, R_it is the amount of R&D innovation investment of cultural enterprises, and B_it is the level of technical efficiency that cultural enterprises are expected to improve through R&D innovation. It should be noted that Jones & Manuelli and Jensen pointed out that externalities in the economy can be transformed into a general form shaped like the AK model after being endogenized by the model [38,39]. The effect of INT_it is introduced by directly multiplying it by an abstract function g, drawing heavily on the AK model in the first generation of endogenous growth theory.

The model continues to give the abstract function g a basic C–D functional form to facilitate mathematical deductive reasoning and to ensure that the law of diminishing marginal returns holds. The equation is as follows:

$$g=\lambda \times {\left(\frac{R_{it}}{B_{it}^{\ast }}\right)}^{\sigma }$$

(9)

For convenience, the present study directly attributed the digital transformation’s improvement in innovation efficiency to parameter λ and calls it the research efficiency parameter. Therefore, the final specific form of Equation (8) is:

$${\mu }_{it}=\lambda \times {\left(\frac{R_{it}}{B_{it}^{\ast }}\right)}^{\sigma }$$

(10)

In Equation (10), λ is the efficiency parameter in the process of R&D innovation, which should satisfy the basic setting of λ > 0, according to the definition. The parameter will also be affected by the INT_it utilization by cultural enterprises; σ is the corresponding elasticity coefficient and obeys the 0 < σ < 1.

Further transforming Equation (10), it can be inferred that there is the following quantitative relationship between the amount of R&D investment in cultural enterprises (R_it), the probability of innovation success (μ_it), and the technical efficiency level (B_it). The relationship can be expressed as follows:

$$R_{it}=B_{it}^{\ast }{\left(\frac{{\mu }_{it}}{\lambda }\right)}^{\frac{1}{\sigma }}$$

(11)

Cultural enterprises often face high financing constraints, and the amount of capital that they can spend on R&D innovation is limited. Governmental and industrial policy support, large cultural enterprises with surplus funds, or loans from other financial institutions are all bases and funds available to cultural enterprises. To simplify the analysis step, the present study introduced the concept of financing costs. Cultural enterprises can obtain external funding support from financial institutions and government departments by paying the corresponding costs, for raising enough money for R&D innovation.

More specifically, the present study further made the following two key assumptions: First, it is assumed that cultural enterprises that undertake R&D innovation also need to raise financing from other cultural enterprises or financial institutions. Culture enterprises have to pay additional financing costs proportional to R&D input (R_it), denoted as θ₁R_it, where θ₁ is between 0 and 1. Meanwhile, since digital transformation can not only change the way cultural enterprises operate, affecting enterprises’ ability to raise external funds, but also reducing search costs within the cultural industry by reducing friction in the inter-enterprise lending market, the financing parameter θ₁ in the model is also closely related to the degree of digital transformation of the cultural enterprise (INT_it). Second, it is assumed that the government supports cultural enterprises. Cultural enterprises are rationed funding according to R&D innovation, denoted as θ₂R_it, where θ₂ is between 0 and 1. However, cultural enterprises can only use the τ part of these grants for R&D innovation, and the remaining part (1 − τ) is used to improve their TFP level to government support, of which is 0 < τ < 1. R&D innovation in cultural enterprises receives a final government support level of τθ₂R_i. Thus, the relevant support policies by the government are bound to favor actively implementing the digital transformation of cultural enterprises, so the parameter set (θ₂) should be shaped by the degree of digital transformation (INT_it).

Subsequently, a mathematical model considers cultural enterprises’ R&D innovation decisions. Since the probability of success of R&D innovation in a cultural enterprise is μ_it, when a cultural enterprise’s R&D innovation succeeds, it will monopolize the intermediate product; if it fails, it will lose its monopoly. For a cultural enterprise with a monopoly over an intermediate product, R&D innovation can benefit the enterprise, which is denoted as μ_itΠ_it. R&D innovation costs are both the added cost of R&D innovative inputs and additional outside funds. Thus, cultural enterprises make optimization decisions when choosing the amount of investment in R&D innovation based on the above two assumptions as follows:

$$\underset{R_{it}}{\max }{\mathsf{\Psi }}_{it}^{\mathrm{R}\&\mathrm{D}}={\mu }_{it}{\mathsf{\Pi }}_{it}^{\ast }-R_{it}-{\theta }_1{R}_{it}+\tau {\theta }_2{R}_{it}$$

(12)

The present study continued the optimization as follows: Substituting Equation (10) and the profit level under the market clearing state in Equation (6) into Equation (12), as well as taking the first-order derivative of the optimal decision function about R_it in this transformed form and making it zero. Organizing the first-order conditions further, it can be found that the optimal R&D innovation investment (R_it), and the technical efficiency level (B_it) to be achieved by enterprises must also satisfy the following quantitative relationship:

$${\left(\frac{R_{it}}{B_{it}^{\ast }}\right)}^{\sigma -1}=\frac{1+{\theta }_1-\tau {\theta }_2}{\lambda \sigma \left({\alpha }^{\frac{1+\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)L_{it}}$$

(13)

The optimum technical efficiency level of cultural enterprises will equal their desired technical efficiency level (B_it) after enterprises determine the optimal R&D innovation investment (R_it), which indicates that R_it will be finally defined as follows:

$$R_{it}={\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{1}{1-\sigma }}{\lambda }^{\frac{1}{1-\sigma }}{\left[\sigma \left({\alpha }^{\frac{1+\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)L_{it}\right]}^{\frac{1}{1-\sigma }}{B}_{it}$$

(14)

Further, by substituting Equation (14) into Equation (10) as a whole, the probability of success of R&D innovation in cultural enterprises can be obtained after eliminating the technical efficiency level (B_it). The equation is as follows:

$${\mu }_{it}={\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{\sigma }{1-\sigma }}{\lambda }^{\frac{1}{1-\sigma }}{\left[\sigma \left({\alpha }^{\frac{1+\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)L_{it}\right]}^{\frac{\sigma }{1-\sigma }}$$

(15)

According to Equation (15) and the specific form of R&D innovation probability of cultural enterprises, under the assumption that the labor supply is sufficient and exogenous, the model setting parameters completely determine the probability of cultural enterprises engaging in R&D innovation after optimal decision-making.

2.4.3. Overall Performance

The present study further summed up the equilibrium results of cultural enterprises regarding production scale decisions and R&D innovation decisions. Based on the model setup, cultural enterprises are distributed on the [0, 1] intermediate goods continuum. According to the study, cultural enterprises that succeed in R&D innovation are distributed in the interval [0, μ], and cultural enterprises that fail in R&D innovation are distributed in the interval [μ, 1]. The overall technical efficiency level of cultural enterprises (B_t) is as follows:

$$B_t={\displaystyle {\int }_0^{\mu }\gamma B_{i,t-1}{L}_{it}}di+{\displaystyle {\int }_{\mu }^1{B}_{i,t-1}{L}_{it}}di$$

(16)

The overall output of cultural enterprises is expressed as follows:

$$Y_t={\displaystyle {\int }_0^{\mu }\left({\alpha }^{\frac{2\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)\left(\gamma B_{i,t-1}{L}_{it}\right)}di+{\displaystyle {\int }_{\mu }^1\left({\alpha }^{\frac{2\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)\left(B_{i,t-1}{L}_{it}\right)}di$$

(17)

The present study assumed that cultural enterprises are uniformly distributed along the intermediate goods continuum and continuously summed from an intermediate goods perspective. Because the mean value of INT_it on the continuum of intermediate goods is INT_t, the mean value of B_i,t−1 on the continuum of intermediate goods is B_t−1, the mean value of L_it on the continuum of intermediate goods is L_t, and Equation (15) can be used to find the probability of cultural enterprises’ R&D innovation success (μ_it). The performance expectation means of cultural enterprises can be calculated as follows.

As the model reaches equilibrium, the performance expectation of the technology level is as follows:

$$\begin{array}{ll}{B}_t& =\left[1+\left(\gamma -1\right)\mu \right]B_{t-1}{L}_t\\ & =\left\{1+\left(\gamma -1\right)\left[c{\lambda }^{\frac{1}{1-\sigma }}{\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{\sigma }{1-\sigma }}\right]\right\}{B}_{t-1}{L}_t\end{array}$$

(18)

Upon reaching equilibrium, the total output of cultural enterprises (Y_t) should be as follows:

$$\begin{array}{ll}{Y}_t& =\left[1+\left(\gamma -1\right)\mu \right]\left({\alpha }^{\frac{2\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)B_{t-1}{L}_t\\ & =\left\{1+\left(\gamma -1\right)\left[c{\lambda }^{\frac{1}{1-\sigma }}{\left(1+{\theta }_1-\tau {\theta }_2\right)}^{\frac{\sigma }{1-\sigma }}\right]\right\}\left({\alpha }^{\frac{2\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)B_{t-1}{L}_t\end{array}$$

(19)

In Equation (18), when the model reaches an equilibrium state, it can be determined that the degree of digital transformation (INT_t), and the relationship between the three key parameters, λ, θ₁, and θ₂, determine how the technical efficiency level of cultural enterprises changes. The present study believed that the following quantitative relationship should exist—the higher the degree of digital transformation, the higher the innovation efficiency of the enterprises as follows:

$$\frac{d\lambda }{dIN{T}_{it}}>0$$

(20)

Since digital transformation can improve the financing efficiency of cultural enterprises mathematically, the financing cost of enterprises will decrease with an increase in the degree of digital transformation (θ₁), which would satisfy the derivative relationship as follows:

$$\frac{d{\theta }_1}{dIN{T}_{it}}<0$$

(21)

For the parameter of government support (θ₂), since the Chinese government is actively guiding the digital transformation of cultural enterprises, the following relationship should be satisfied between government support and the degree of digital transformation of cultural enterprises as follows:

$$\frac{d{\theta }_2}{dIN{T}_{it}}>0$$

(22)

A relationship between changes in parameters can be determined by summarizing the continuum of intermediate goods in the previous model. They are dλ/dINT_t > 0, dθ₁/dINT_t < 0, and dθ₂/dINT_t > 0. At this point, taking the derivative of INT_t for Equation (18) and simplifying shows that the relationship between the overall technical efficiency level (B_t) and the digital transformation level (INT_t) is expressed as follows:

$$\frac{d{B}_t}{dIN{T}_t}=\frac{1}{1-\sigma }c{m}_1\frac{d\lambda }{dIN{T}_t}-\frac{\sigma }{1-\sigma }c{m}_2\frac{d{\theta }_1}{dIN{T}_t}+\frac{\sigma }{1-\sigma }\tau c{m}_2\frac{d{\theta }_2}{dIN{T}_t}>0$$

(23)

The derivative equation abbreviates the mathematical expression terms of the parameters not affected by digitization as m₁ and m₂. Their specific forms indicate that they are always positive as follows:

$$\left\{\begin{array}{l}{m}_1=\left(\gamma -1\right){\lambda }^{\frac{\sigma }{1-\sigma }}{\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{\sigma }{1-\sigma }}{B}_{t-1}{L}_t>0\\ m_2=c\left(\gamma -1\right){\lambda }^{\frac{1}{1-\sigma }}{\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{1}{1-\sigma }}{B}_{t-1}{L}_t>0\end{array}\right.$$

(24)

Based on the quantitative relationship between TFP and technical efficiency level of cultural enterprises, which is dA_t/dB_t, as shown in Equation (7), the present study continued to rewrite Equation (23), and it could be seen that the first-order derivative of TFP in cultural enterprises is positively correlated with the degree of digital transformation as follows:

$$\frac{d{A}_t}{dIN{T}_t}=\frac{d{A}_t}{d{B}_t}\times \frac{d{B}_t}{dIN{T}_t}>0$$

(25)

According to the results of this comparative static analysis, as the degree of digital transformation of cultural enterprises increases, the level of TFP in cultural enterprises will rise.

On the basis of the aforementioned analysis, the present study proposed Hypothesis 1 as follows:

Hypothesis 1.

Digital transformation can improve the TFP of cultural enterprises in China.

2.5. Impact Paths

2.5.1. Digital Transformation Affects Content Creation Efficiency

The present study continued to derive hypotheses for the three previously explored impact paths presented in the theoretical analysis section.

In terms of the content creation efficiency of cultural enterprises, Schumpeter’s innovation growth model abstracts real-world enterprises’ production process as processing intermediate goods to produce final goods, while culture enterprises’ core production is content creation. In the present study, the total consumption of intermediate goods by cultural enterprises (x_t) could be used to measure enterprises’ content creation efficiency. The intermediate goods consumption for each period (B_t) can be determined by substituting Equation (10) back into Equation (18), and in addition, taking its first-order derivative of INT_t, the equation is as follows:

$$\frac{d{x}_t}{dIN{T}_t}=\frac{1}{1-\sigma }c{m}_1{L}_t{\alpha }^{\frac{2}{1-\alpha }}\frac{d\lambda }{dIN{T}_t}-\frac{\sigma }{1-\sigma }c{m}_2{L}_t{\alpha }^{\frac{2}{1-\alpha }}\frac{d{\theta }_1}{dIN{T}_t}+\frac{\sigma }{1-\sigma }\tau c{m}_2{L}_t{\alpha }^{\frac{2}{1-\alpha }}\frac{d{\theta }_2}{dIN{T}_t}>0$$

(26)

According to this quantitative relationship, cultural enterprises consume more intermediate goods as INT_t increases. Based on the previous theoretical and mathematical model, the efficiency of inter-enterprise collaborative creation, as well as the efficiency of communication between the enterprise and the consumer through the positive externality brought about by the digital transformation, boosts the efficiency of the content creation process and accelerates the consumption of intermediates per unit of time [40,41]. Typically, these intermediates are intangible assets, such as copyrights (IP), exclusive creation technology, etc. The increase in consumption per unit of time not only means that digital transformation can contribute to these cultural enterprises developing more precipitated intangible assets that have not yet realized their economic value, but also assist cultural enterprises in transforming non-physical creative materials into more tangible benefits [42,43].

On the basis of the aforementioned analysis, the present study proposed Hypothesis 2a as follows:

Hypothesis 2a.

Digital transformation can improve the TFP of cultural enterprises by increasing their efficiency of content creation.

2.5.2. Digital Transformation Enhances Financing Ability and Efficiency

In terms of the financing ability and efficiency of cultural enterprises, by reducing the information asymmetry problem in the cultural financial market and introducing government credit to endorse cultural enterprises, digital transformation can reduce the financing costs of cultural enterprises [44,45,46]. Observing the cultural enterprise technical efficiency level parameter (B_t) at model equilibrium derived from Equation (18), it can be found that the financing cost parameter (θ₁) is located in the denominator, and the power index is positive, which suggests that B_t will increase with a decrease in θ₁.

On the basis of the aforementioned analysis, the present study proposed Hypothesis 2b as follows:

Hypothesis 2b.

Digital transformation can improve the TFP of cultural enterprises by increasing their financing ability.

2.5.3. Digital Transformation Improves R&D Innovation Efficiency and Attention

In terms of R&D innovation efficiency and attention of cultural enterprises, the quantitative relationship between the level of R&D innovation investment of cultural enterprises (R_it) and the technical efficiency level (B_it) expressed in Equation (14) indicates that the total level of R&D innovation investment of cultural enterprises (R_t) is expressed as follows:

$$R_t={\left(\frac{1}{1+{\theta }_1-\tau {\theta }_2}\right)}^{\frac{1}{1-\sigma }}{\lambda }^{\frac{1}{1-\sigma }}{\left[\sigma \left({\alpha }^{\frac{1+\alpha }{1-\alpha }}-{\alpha }^{\frac{2}{1-\alpha }}\right)L_t\right]}^{\frac{1}{1-\sigma }}{B}_t$$

(27)

To further simplify and organize the equation, the present study substituted Equation (23), which represents B_t at equilibrium for the cultural enterprises, into the above equation, thus taking the derivative concerning INT_t as follows:

$$\frac{d{R}_t}{dIN{T}_t}>0$$

(28)

The derivative relationship indicates that the level of R&D innovation investment (R_t) of the cultural enterprises at equilibrium will increase with the degree of digital transformation (INT_t). From Equation (23), the derivative relationship between the technical level (B_t), and the digital transformation degree (INT_t) indicates that the positive impact of the digital transformation on R&D efficiency will eventually result in a positive increase in cultural enterprises’ TFP. Moreover, higher resulting returns on R&D innovation investments increase the likelihood of cultural enterprises investing in R&D innovation [11,47].

On the basis of the aforementioned analysis, the present study proposes Hypothesis 2c as follows:

Hypothesis 2c.

Digital transformation can improve the TFP of cultural enterprises by facilitating their R&D innovation.

Based on the above models and hypotheses, the present study constructed a conceptual model as shown in Figure 2.

3. Theoretical Derivation

3.1. Econometric Modeling

The present study constructed the following two-way fixed effects model to study whether digital transformation can increase the TFP of cultural enterprises:

$$Eco{P}_{i,t}={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\displaystyle \sum }{\beta }_{j}Control{s}_{i,j,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}$$

(29)

In the model, i and t symbolize the individual enterprise and year, respectively; β is the estimated coefficient; j is the jth control variable; EcoP is the dependent variable indicating the TFP of cultural enterprises; lnINT is the core explanatory variable indicating the level of digital transformation of cultural enterprises; and Controls is a control variable. The present study referenced the research results of other studies and analyzed the following variables pertaining to cultural enterprises: finance expense rate (Finexr), fixed assets ratio (Fixr), return on assets (ROA), enterprise size (Size), board scale (Boars), leadership structure (Dual), average compensation of executives (EAC), age of enterprise (Age), enterprise ownership (SOE), and other related indicators that were controlled for; μ_i is the corporate fixed effect, ν_t is the yearly fixed effect, and ε_i,t is the classical randomized disturbance term.

The present study also constructed a benchmark indicator regression model, which can be expressed as Equation (30). Additionally, to ensure the robustness of the empirical conclusions of the present study, an auxiliary indicator regression model was constructed, as presented in Equation (31). That is, if the coefficient β₁ of the core explanatory variable (the level of digital transformation of cultural enterprises, lnINT) is significant, digital transformation can increase the TFP of cultural enterprises.

The benchmark indicator regression equation is expressed as follows:

$$\begin{array}{ll}\ln L{P}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Finex{r}_{i,t}+{\beta }_{3}Fix{r}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Siz{e}_{i,t}\\ & +{\beta }_{6}Boar{s}_{i,t}+{\beta }_{7}Dua{l}_{i,t}+{\beta }_{8}EA{C}_{i,t}+{\beta }_{9}Ag{e}_{i,t}+{\beta }_{10}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(30)

The auxiliary indicator regression equation is expressed as follows:

$$\begin{array}{ll}\ln O{P}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Finex{r}_{i,t}+{\beta }_{3}Fix{r}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Siz{e}_{i,t}\\ & +{\beta }_{6}Boar{s}_{i,t}+{\beta }_{7}Dua{l}_{i,t}+{\beta }_{8}EA{C}_{i,t}+{\beta }_{9}Ag{e}_{i,t}+{\beta }_{10}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(31)

3.2. Measurement and Description of Variables

1. TFP of cultural enterprises (lnLP and lnOP). Two semiparametric estimation methods, LP and OP, were used to estimate the logarithmic value of the TFP in cultural enterprises [31,48,49]. By adding an intermediate goods indicator to the OP method, the LP method captures simultaneous adjustments with a more generalized indicator. The TFP of cultural enterprises is a measure of the efficiency with which production factors are organized to create economic value in an enterprise that cannot be explained using input factors such as manpower and capital. A cultural enterprise’s TFP increases with its TFP value. In the present study, OP method indicators were used as auxiliary indicators alongside LP method indicators to ensure the validity of the empirical results.

2. Level of digital transformation (lnINT). By mining the text of the annual reports of listed cultural enterprises through the entropy weight Topsis method, the present authors were able to calculate an index of the level of digital transformation of enterprises. Unstructured text mining analysis technology has matured, providing new solutions and research support for measuring digital transformation indices [43,50,51,52,53]. After cleaning and de-redundancy processes were performed, the digital transformation indicators of cultural enterprises were constructed, with the lexicon method used to screen for and extract valid information from annual reports. Python software (3.8.3) was used to crawl annual reports from the Juchao Information Network (www.cninfo.com.cn 14 May 2023), designated as a centralized disclosure site for listed enterprises by the Securities and Exchange Commission, for all preliminary screening sample companies from 2011 to 2019. Over 3000 annual reports were reviewed for the scope of operation and main business content for the year. A total of 72,327 valid text statements were selected by combining programmed batch processing with manual screening one by one. Finally, the current authors retained the unbalanced panel data of 251 listed cultural enterprises with 1541 sample points. Based on the relevant policy documents about the digital development of cultural enterprises in China, keywords related to the digital transformation of cultural enterprises were selected from the relevant literature. The keywords were categorized and reviewed by a group of reviewers comprising five experts and scholars in the fields of culture and digital economy and five experts with practical experience in the cultural industry. The keyword list underwent several rounds of revisions, and whenever disagreements occurred, the reviewers discussed them and made decisions on the basis of the majority rule; subsequently, the keyword list was finalized with the approval of more than two-thirds of the reviewers. Figure 3 depicts the three-level indicator evaluation system employed in the present study and the finalized list of core keywords.

3. Control variables. The variable, Finexr, is the ratio of financial expenses to operating income for a cultural enterprise in a given year. Fixr is the ratio of net fixed assets to total assets for a cultural enterprise in a given year. ROA is calculated by dividing the net profit of a cultural enterprise by the balance of total assets in a given year. Size is the logarithm of a cultural enterprise’s total assets for a given year. Boars is the logarithm of the number of board members in a cultural enterprise. Dual is the dummy variable for determining whether the chairman and manager of the cultural enterprise are two positions in one. EAC is the average compensation of the executives of a cultural enterprise, calculated as the total remuneration and allowances paid to directors, supervisors, and senior executives divided by the number of directors and supervisors who received annual remuneration. Age is the age of an enterprise, calculated as the number of years since its establishment. SOE is a dummy variable indicating whether the government is the actual controller.

The definitions and calculations of the variables for the above indicators are summarized in

Table 1. Definitions and calculations of the main variables.

Variable Type	Definition	Symbolic	Calculation
Explanatory Variable	TFP of cultural enterprises (Baseline indicators)	lnLP	Logarithmic TFP of enterprises measured using the LP method
	TFP of cultural enterprises (Auxiliary indicators)	lnOP	Logarithmic TFP of enterprises measured using the OP method
Core Explanatory Variables	Level of digital transformation	lnINT	ln(Number of keywords related to annual reports +1)
Control Variables	Finance expense rate	Finexr	Financial expenses/Operating income
	Fixed assets ratio	Fixr	Net fixed assets/Total assets
	Return on assets	ROA	Net profit/Balance of total assets
	Enterprise size	Size	Logarithm of total assets
	Board scale	Boars	Logarithm of the number of board members
	Leadership structure	Dual	Whether chairman and manager are two positions in one, 1 for yes, 0 for no
	Average compensation of executives	EAC	Logarithm of the average compensation among executives
	Age of enterprise	Age	Years of existence in cultural enterprise
	Enterprise ownership	SOE	Whether the actual controller of the enterprise is central or local government, 1 for yes, 0 for no

.

3.3. Measurement and Description of Variables

The data sets used in this study were primarily retrieved from the publicly available reports of listed cultural enterprises. Financial data and enterprise information about listed cultural enterprises were retrieved from the CSMAR and Wind databases. Furthermore, the variables used in the estimation process, such as the gross enterprise output index and the asset index, are analyzed by the corresponding price deflators, and all of these deflators are taken from China Economic Network’s Statistics Database (https://db.cei.cn/jsps/Home 8 June 2023).

Table 2. Descriptive statistics for examined variables.

Variable	Observed Value	Mean	Minimum	Median	Maximum	Variance
lnLP	1541	1.924	1.666	1.923	2.128	0.089
lnOP	1541	1.631	0.422	1.635	1.963	0.116
lnINT	1541	3.121	0	3.296	5.394	1.377
Finexr	1541	0.012	−0.055	0.004	0.172	0.034
Fixr	1541	0.183	0.001	0.129	0.648	0.164
ROA	1541	0.033	−0.602	0.046	0.225	0.105
Size	1541	21.841	19.625	21.785	24.304	1.030
Boars	1541	2.123	1.609	2.197	2.708	0.205
Dual	1541	0.330	0	0	1	0.470
EAC	1541	12.547	11.115	12.545	14.290	0.648
Age	1541	18.910	4	18	54	6.411
SOE	1541	0.352	0	0	1	0.478

lists the descriptive statistics for the examined variables.

3.4. Data Representativeness and Limitations

Early on, government-owned cultural institutions mixed with government functions were included in Chinese cultural enterprises, since they were inherited and developed from the former planned economy’s cultural sector. In response, the Chinese government introduced a series of enterprise restructuring policies, regaining market management rights for these large cultural institutions. The monopoly power due to their administrative privileges through government regulation was also reduced.

Data in the present study were collected from listed cultural enterprises in China, eliminating the possibility of mixed data about cultural enterprises with government functions, whereas, to avoid potential risks, the current authors assessed the TFP, as economic performance of cultural enterprises by digital transformation, rather than their social performance.

4. Empirical Tests

4.1. Full-Sample Regression Results

Table 3. Full-sample regression results for effects of digital transformation on cultural enterprises’ TFP.

Variable	(1)	(2)	(3)	(4)
	lnOP	lnLP	lnOP	lnLP
lnINT	0.008 **	0.009 ***	0.007 **	0.006 ***
	(2.30)	(4.11)	(2.02)	(2.67)
Finexr			−0.782 ***	−0.465 ***
			(−9.14)	(−8.42)
Fixr			0.044 *	0.039 **
			(1.69)	(2.29)
ROA			−0.043 **	−0.013
			(−2.09)	(−1.01)
Size			−0.003	0.015 ***
			(−0.57)	(4.37)
Boars			0.008	−0.003
			(0.42)	(−0.27)
Dual			0.006	0.006
			(1.04)	(1.53)
EAC			0.035 ***	0.022 ***
			(5.78)	(5.71)
Age			0.003 **	0.001 *
			(2.23)	(1.65)
SOE			0.016	0.014
			(0.84)	(1.11)
Intercept	1.595 ***	1.881 ***	1.177 ***	1.275 ***
	(164.99)	(300.33)	(10.03)	(16.85)
Year Fixed Effects	Yes	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes	Yes
N	1541	1541	1541	1541
R2	0.064	0.117	0.148	0.204

***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the full-sample regression results for the effects of digital transformation on the TFP of cultural enterprises.

Columns (1) and (3) list the regression estimation results obtained using Equation (32) and the auxiliary explanatory variable indicator lnOP, and columns (2) and (4) list the regression estimation results obtained using Equation (30) and the baseline explanatory variable indicator lnLP. A univariate regression was also conducted using only the core explanatory variables (lnINT); this step was performed to address the potential interference effects of the control variables on the estimated coefficients of the core explanatory variables. The multivariate regression results reveal a goodness of fit of 15%, and given that enterprise microdata and TFP were used as explanatory variables in the present study, an R² of this level indicates that the model exhibited a high level of explanatory power.

The regression results presented in columns (1) to (4) reveal that the estimated coefficients for the core explanatory variables are positive at least at the 5% level of significance. They also indicate that digital transformation has significant positive effects on the TFP of cultural enterprises, regardless of whether control variables are considered. Column (4) presents the benchmark indicator regression results for the control variables, and they reveal a lnINT coefficient of 0.006 at a significance level of 1%. When the individual characteristics of the cultural enterprises and macro-environmental factors were fully considered, digital transformation significantly increased the TFP of these cultural enterprises. Therefore, Hypothesis 1 is supported.

As for the control variables, the regression results from column (4) estimated by the benchmark indicator are used as the basis for the analysis, and the regression results from column (3) estimated by the auxiliary indicator serve as the robustness corroboration. First, the estimated coefficient on the financial expense rate (Finexr) is −0.465 at a significance level of 1%, and the indicator coefficient on the share of fixed assets ratio (Fixr) is positive at 5% significance, demonstrating robust results in both cross-sectional comparisons. Two perspectives are reflected in this result: the lower the cost of financing, the higher the TFP of cultural enterprises; the lower the difficulty of financing, the higher the TFP of cultural enterprises. Since cultural production relies on the content creation without a specific physical form, most cultural enterprises in the industry will not hold a large amount of fixed equipment to operate. However, banks and other traditional financial institutions require fixed assets as a guarantee when granting credits to enterprises, so cultural enterprises often cannot utilize these external financial supports. In some cases, cultural enterprises have to give up immediate business opportunities due to financial constraints. Therefore, they are unable to scale production to achieve a higher level of TFP in enterprises. In addition to being limited by high financing costs, cultural enterprises that must raise funds from outside will also be limited by high-interest rates, which suggests that “difficult and expensive financing” is one of the major factors limiting Chinese cultural enterprises’ development.

Moreover, the coefficient of the average compensation of executives (EAC) indicator is 0.022, with a significance level of more than 1%, and the cross-sectional comparison is also robust. Cultural enterprises’ TFP performs better economically when their executive compensation is higher. From an economic perspective, more experienced and forward-looking managers tend to create more economic value for the enterprises they work for, so enterprises often need to pay a higher salary to hire them. In addition, high compensation motivates cultural enterprise executives to work harder to improve the enterprises’ TFP. The coefficient of the enterprises’ age (Age) indicator of cultural enterprises is positive at a 10% significance level, and the comparison results remain robust, suggesting that the longer that cultural enterprises survive, the higher their TFP would be. What contributes to this result is that the longer a cultural enterprise has existed, the more closely it is connected to the community, and the better it can grasp current trends. A key trend in Chinese cultural enterprises is the integration of the Internet and cultural enterprises. As cultural enterprises become more experienced in digital transformation, they will be able to use Internet technology to transform their own production and operation mode, ultimately achieving higher levels of TFP.

4.2. Robustness Tests

4.2.1. Robustness Test Involving Removal of First-Tier-City and Megacity Samples

A city’s economic environment affects the development of cultural enterprises. Given that a megacity typically has a comprehensive economic structure and abundant cultural production resources, whether digital transformation increases the TFP of cultural enterprises exclusively in megacities and whether the local effect is sufficiently strong to render nonsignificant the effect of digital transformation on the TFP of cultural enterprises in other types of cities warrant exploration. Therefore, the sample was regressed after the exclusion of cultural enterprises based in first-tier cities [54], such as Beijing, Shanghai, and Guangzhou, and in megacities, such as Shenzhen and Chengdu.

Table 4. Regression results after the exclusion of cultural enterprises from first-tier cities and megacities.

Variable	(1)	(2)	(3)	(4)
	Excluding the Samples in First-Tier Cities		Excluding the Samples in Megacities
	lnOP	lnLP	lnOP	lnLP
lnINT	0.011 ***	0.008 ***	0.012 ***	0.009 ***
	(3.09)	(3.29)	(3.26)	(3.37)
Finexr	−0.628 ***	−0.430 ***	−0.613 ***	−0.422 ***
	(−6.40)	(−5.94)	(−6.18)	(−5.78)
Fixr	0.037	0.041 **	0.036	0.043 **
	(1.38)	(2.04)	(1.31)	(2.12)
ROA	0.014	0.006	0.007	−0.000
	(0.55)	(0.30)	(0.29)	(−0.02)
Size	−0.016 **	0.004	−0.017 **	0.004
	(−2.39)	(0.82)	(−2.45)	(0.82)
Boars	0.045 **	0.034 **	0.050 **	0.037 **
	(2.26)	(2.27)	(2.45)	(2.48)
Dual	0.010	0.008	0.011	0.009 *
	(1.47)	(1.60)	(1.48)	(1.66)
EAC	0.040 ***	0.028 ***	0.040 ***	0.029 ***
	(5.57)	(5.25)	(5.48)	(5.25)
Age	0.003 *	0.001	0.003 *	0.002
	(1.79)	(1.40)	(1.93)	(1.45)
SOE	0.012	0.011	0.014	0.012
	(0.59)	(0.73)	(0.68)	(0.79)
Intercept	1.298 ***	1.352 ***	1.290 ***	1.328 ***
	(8.77)	(12.39)	(8.35)	(11.66)
Year Fixed Effects	Yes	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes	Yes
N	896	896	874	874
R2	0.191	0.200	0.198	0.208

***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

reports the results of this regression.

The lnINT coefficients listed in columns (1) to (4) are all positive at the 1% level of significance. This indicates that the sample estimation results support the empirical assertion of the present study, regardless of whether the sample group of cultural enterprises includes those from first-tier cities (e.g., Beijing, Shanghai, and Guangzhou) or megacities (e.g., Shenzhen and Chengdu). Overall, this finding indicates that the boosting effect of digital transformation on the TFP of cultural enterprises is not affected by the presence or absence of first-tier-city and megacity samples and can be generalized across China. Thus, the robustness of the empirical findings is confirmed.

4.2.2. Robustness Tests for Removing Potential Systematic Measurement Error Disturbances in Core Explanatory Variables

Although the effects of variables were controlled for at multiple levels in the full-sample regressions, systematic bias in the measurement of the core explanatory variable (lnINT), could have led to biased regression estimates. Therefore, to address potential biases, the indicators of the digital transformation of the cultural enterprises were reconstructed, and the regressions obtained using Equations (30) and (31) were estimated again. The new indicator (rINT) deemphasizes the information regarding the number of appearances of relevant keywords in a given annual report. Instead, it focuses only on an enterprise’s ranking with respect to its level of digital transformation among all cultural enterprises in the enterprise’s city in a given year, as listed in China Economic Network’s Statistics Database (https://db.cei.cn/jsps/Home). This method reduces the influence of systematic measurement errors.

Table 5. Regression results after replacement of core explanatory variables with proxies.

Variable	(1)	(2)
	lnOP	lnLP
rINT	0.002 ***	0.001 ***
	(3.54)	(2.82)
Intercept	1.223 ***	1.280 ***
	(10.44)	(16.88)
Year Fixed Effects	Yes	Yes
Firm Fixed Effects	Yes	Yes
N	1541	1541
R2	0.154	0.205

*** indicates significance at the 1% levels; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the estimation results obtained after replacing the core explanatory variable indicators. For brevity, the control variable results will not be reported as part of the empirical results in the present study.

Table 5 reveals positive results at the 1% level of significance for the coefficients of the new indicator (rINT) for the level of digital transformation of cultural enterprises; notably, these results are unaffected by whether regression is performed using the baseline indicator (lnLP) or the auxiliary indicator (lnOP), indicating that they are unaffected by systematic measurement bias and confirming the reliability and robustness of the findings.

4.3. Endogenous Problem-Solving: A Retest Based on Matching Treatment Effect Estimates

Although benchmark regression analysis was used to analyze and control for the effects of measurement factors at various levels, endogenous interference must be considered in the present study. Figure 4 reveals that the digital transformation of Chinese cultural enterprises during the study period (2011–2019) is characterized by dynamic development.

The results reveal clear shifts in the trend of digital transformation among Chinese cultural enterprises. With 2015 serving as a time node, the digital transformation of cultural enterprises in China shifted from an early period of rapid expansion to a period of dynamic adjustment and deepening. The years 2011–2015 marked the peak of digital upgrading for Chinese cultural enterprises, whereas the years 2015–2019 were characterized by a high level of volatility with respect to the digital transformation indicators of Chinese cultural enterprises. Therefore, a reverse causality relationship might exist between digital transformation and the TFP of Chinese cultural enterprises, and the TFP of these enterprises in specific years could have influenced the overall effect of digital transformation. Therefore, the digital transformation degree indicator (lnINT) of cultural enterprises was treated as a dummy variable (DID) in this subsection to address endogeneity-related concerns [55]. Matching treatment effect estimation was used to address potential endogenous disturbances. The average level of digital transformation of Chinese cultural enterprises in the 3 years before and after 2015 were calculated separately. Subsequently, the average values were reassessed after differential treatment. Chinese cultural enterprises whose level of digital transformation exceeded the aforementioned average value were classified as the experimental group. By contrast, cultural enterprises whose level of digital transformation under the average value were classified as the control group. The year 2015 was used as the start time of the intervention, and the control variables were used as matching conditions for estimating treatment effects. To reduce estimation interference caused by the use of specific matching methods, four matching methods were sequentially developed, namely put-back propensity score matching (PSM matching with put-back), PSM matching without put-back, kernel density function matching, and Mahalanobis Distance Matching.

Table 6. Treatment effect model estimates based on four matching methods.

Variable	PSM Matching with Put-Back	PSM Matching without Put-Back	Kernel Density Function Matching	Mahalanobis Distance Matching
	(1)	(2)	(5)	(7)
DID	0.148 ***	0.147 **	0.153 ***	0.138 ***
	(2.88)	(2.41)	(3.00)	(2.75)
DID-1	0.104	0.094	0.092	0.084
	(1.12)	(0.90)	(1.00)	(0.93)
Intercept	1.725 *	1.969 *	2.513 ***	2.255 **
	(1.76)	(1.72)	(2.64)	(2.41)
Year Fixed Effects	Yes	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes	Yes
N	698	452	692	506
R2	0.253	0.345	0.251	0.249

***, **, and * indicate significance at the 1%, 5%, and 10% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the results.

Table 6 reports the matching treatment effect model estimates. The effects of the estimated coefficients on the treatment effect indicator (DID) were positive at least at the 5% level of significance, indicating that the full-sample regression results remained valid after the endogeneity problem was addressed. This finding confirms the reliability and robustness of the previously reported findings.

4.4. Heterogeneity Analysis

4.4.1. Enterprise Size Heterogeneity

In the sample of cultural enterprises, those of different sizes differed significantly in terms of their productivity, their main goals, and development constraints, and these factors could have affected the effects of digital transformation in increasing the TFP of the enterprises. To measure the differences among enterprises of different sizes, as third quartile employed as a boundary to divide the sample into subsamples comprising larger and smaller cultural enterprises based on their median sizes [56,57]. Equations (30) and (31) were applied for regression estimation, and the results are presented in

Table 7. Estimated regression results based on sizes of cultural enterprises.

Variable	(1)	(2)	(3)	(4)
	Supporting Indicators Measured by the OP Method		Benchmark Indicators Measured by the LP Method
	Larger Enterprises	Smaller Enterprises	Larger Enterprises	Smaller Enterprises
lnINT	0.003	0.010 *	0.003	0.008 ***
	(0.69)	(1.91)	(1.07)	(2.63)
Intercept	1.000 ***	0.792 ***	1.113 ***	1.047 ***
	(4.99)	(3.51)	(8.10)	(7.50)
Year Fixed Effects	Yes	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes	Yes
N	770	771	770	771
R2	0.181	0.202	0.218	0.213

*** and * indicate significance at the 1% and 10% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

.

An analysis of the regression results presented in columns (3) and (4) was performed using the benchmark indicator lnLP, and it revealed that the estimated coefficients of lnINT were nonsignificant in the subsample of larger-sized enterprises. By contrast, the estimated lnINT coefficients were significantly positive in the subsample of smaller-sized enterprises. A cross-sectional comparison of the regression results of columns (1) and (2) revealed that this estimation was stable, indicating that the boosting effect of digital transformation on the TFP of cultural enterprises was more pronounced among the smaller-sized enterprises. The TFP potential of the larger-sized cultural enterprises was limited relative to that of the smaller-sized ones. The smaller-sized cultural enterprises tended to be more incentivized to implement reforms and digital transformation to overcome their financing constraints and lower productivity, which hindered their development.

4.4.2. Regional Heterogeneity

In China, the level of economic development and sophistication of economic structures differs widely by region, with the eastern region in China exhibiting substantially higher levels of economic development relative to the midwestern region.

Table 8. Regression estimation results stratified by region of cultural enterprise.

Variable	(1)	(2)	(3)	(4)
	Supporting Indicators Measured by the OP Method		Benchmark Indicators Measured by the LP Method
	East	Midwest	East	Midwest
lnINT	0.004	0.012 **	0.004	0.009 **
	(0.97)	(2.29)	(1.46)	(2.51)
Intercept	1.223 ***	1.353 ***	1.326 ***	1.326 ***
	(8.78)	(6.31)	(15.17)	(8.44)
Year Fixed Effects	Yes	Yes	Yes	Yes
Firm Fixed Effects	Yes	Yes	Yes	Yes
N	1177	364	1177	364
R2	0.173	0.215	0.239	0.226

*** and ** indicate significance at the 1% and 5% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the results of regression estimation stratified by geographic location. The regression results presented in columns (3) and (4), which were obtained using the benchmark indicator (lnLP) as the explanatory variable, reveal that the lnINT coefficient was positive at the 10% level of significance when using the subsample of cultural enterprises based in midwest China. However, when the subsample of cultural enterprises based in east China was used, the lnINT coefficient was nonsignificant. A cross-sectional comparison of the regression results in columns (1) and (2), which were obtained using the auxiliary indicator lnOP, indicated that they are fundamentally consistent. The cultural enterprises based in midwest China were more likely to benefit from digital transformation than those based in east China. These results indicate that relative to midwest China, east China was more economically developed and more suitable for the development of cultural enterprises in terms of its financial services, talent supply, and policy environment; that is, digital transformation in east China has limited room for further improvement, whose current level of digital transformation is at a high level. By contrast, the cultural enterprises in midwest China can benefit from digital transformation by converting their abundant regional cultural resources into cultural products and generating high economic returns, which can increase the TFP of these enterprises to a greater degree than it would in east China.

4.4.3. Industry Segment Heterogeneity

Based on their production characteristics, the full sample of Chinese cultural enterprises was further divided by industry segment into the three categories of cultural creation enterprises, cultural communication and entertainment enterprises, and culture-related manufacturing enterprises.

Table 9. Regression estimation results stratified by industry segment of cultural enterprise.

Variable	(1)	(2)	(3)	(4)	(5)	(6)
	Supporting Indicators Measured by the OP Method			Benchmark Indicators Measured by the LP Method
	Cultural Creation Industry	Cultural Communication and Entertainment Industry	Culture-Related Manufacturing Industry	Cultural Creation Industry	Cultural Communication and Entertainment Industry	Culture-Related Manufacturing Industry
lnINT	0.012 **	0.026 ***	−0.005	0.010 ***	0.019 ***	−0.002
	(2.00)	(3.39)	(−1.34)	(2.69)	(3.37)	(−1.01)
Intercept	1.475 ***	1.832 ***	1.116 ***	1.469 ***	1.665 ***	1.149 ***
	(8.11)	(4.63)	(6.40)	(12.65)	(5.70)	(10.21)
N	729	238	574	729	238	574
R2	0.188	0.339	0.296	0.250	0.306	0.309

*** and ** indicate significance at the 1% and 5% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the regression estimation results stratified by industry segment.

Columns (4)–(6) present the regression estimation results obtained using lnLP as the explanatory variable and reveal that the lnINT coefficient of the cultural creation industry subsample was 0.010 at a 1% level of significance; by contrast, the lnINT coefficient of the cultural communication and entertainment industry subsample was 0.019 at a 1% level of significance, and the lnINT coefficient of the culture-related manufacturing industry subsample was nonsignificant. A cross-sectional comparison involving the corresponding subsample regression results in columns (1)–(3), which were estimated using the auxiliary explanatory variable indicator (lnOP), revealed that the observed phenomenon was stable. This result indicates that digital transformation was most effective in increasing the TFP of the cultural enterprises in the cultural creation industry, followed by those in the cultural communication and entertainment industry and those in the culture-related manufacturing industry.

The following presents a reasonable explanation for this result. In the cultural industry, cultural creation enterprises and cultural communication and entertainment enterprises are positioned in the upstream chain of the industry and entirely focused on content creation. That is, these enterprises mainly rely on the productivity of creative personnel, and consequently, they focus more on innovation aimed at improving the soft power of their creation works and are less focused on changing professional instruments and equipment during the process of transformation. Because of these factors, cultural creation enterprises and cultural communication and entertainment enterprises may find it easier to implement changes and achieve results through digital transformation relative to the other type of cultural enterprises. By contrast, culture-related manufacturing enterprises primarily manufacture auxiliary cultural equipment; thus, they rely more on using physical assets as collateral to secure financing from traditional financial institutions, which enables them to implement digital transformation with fewer financing constraints. For cultural enterprises, the success of their digital transformation is strongly affected by their ability to address the problem of severe financing constraints associated with light-asset operations. Thus, specifically addressing the urgent financing needs of cultural creation enterprises and cultural communication and entertainment enterprises can significantly increase the effectiveness of digital transformation in enhancing the TFP of upstream chain enterprises in segmented industries.

5. Mechanism-of-Action Testing

5.1. Content Creation Efficiency Testing

Intangible assets, such as intellectual property rights and copyrights, are processed by cultural enterprises to create cultural products. Digital transformation can increase the efficiency with which cultural enterprises transform their intangible assets into final cultural goods, that is, their content creation efficiency. On a digital content creation platform, the production process that cultural enterprises employ is standardized and more efficient, and collaboration between cultural enterprises is closer. Digital transformation enables a producer to engage in faster and more detailed communication of specific demand-side requirements, enabling cultural enterprises to adjust their creative content and transform their intangible assets more quickly into cultural products and services that meet the needs of customers. By increasing the efficiency of content creation, digital transformation can enhance the TFP of cultural enterprises.

This subsection describes the mediation effect model constructed to test the influence mechanism in the present study. For the selection of proxy indicator, a cultural enterprise that exhibited higher content creation efficiency had fewer intangible assets that were not yet converted into cultural products by the end of a given year. The number of intangible assets held at the end of the year, as disclosed in the annual report of a listed cultural enterprise, was standardized on the basis of the total corporate assets of the enterprise in a given year to construct an intangible assets ratio indicator (Intar), which served as a proxy variable for the efficiency with which a cultural enterprise transforms its intangible assets. For the empirical model setup, the current authors employed the stepwise method used in other studies to conduct mediation effect tests [58,59], building on the total effect model in Equation (30) to construct the mediation effect test model outlined in Equations (32) and (33).

The benchmark indicator regression equation is expressed as follows:

$$\begin{array}{ll}Inta{r}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Finex{r}_{i,t}+{\beta }_{3}Fix{r}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Siz{e}_{i,t}\\ & +{\beta }_{6}Boar{s}_{i,t}+{\beta }_{7}Dua{l}_{i,t}+{\beta }_{8}EA{C}_{i,t}+{\beta }_{9}Ag{e}_{i,t}+{\beta }_{10}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(32)

$$\begin{array}{ll}\ln L{P}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Inta{r}_{i,t}+{\beta }_{3}Finex{r}_{i,t}+{\beta }_{4}Fix{r}_{i,t}+{\beta }_{5}RO{A}_{i,t}+{\beta }_{6}Siz{e}_{i,t}\\ & +{\beta }_{7}Boar{s}_{i,t}+{\beta }_{8}Dua{l}_{i,t}+{\beta }_{9}EA{C}_{i,t}+{\beta }_{10}Ag{e}_{i,t}+{\beta }_{11}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(33)

If the verified mechanism exists, β₁ in the estimation result of Equation (32) is significantly negative, and β₂ in the estimation result of Equation (33) is significantly negative.

Table 10. Regression estimation results in content creation efficiency improvement mechanism.

Variable	(1)	(2)
	Intar	lnLP
lnINT	−0.004 ***	0.005 **
	(−2.60)	(2.29)
Intar		−0.218 ***
		(−5.64)
Intercept	0.087	1.294 ***
	(1.61)	(17.29)
Year Fixed Effects	Yes	Yes
Firm Fixed Effects	Yes	Yes
N	1541	1541
R2	0.029	0.223

*** and ** indicate significance at the 1% and 5% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the results of the mechanism test performed to evaluate the digital transformation implemented by cultural enterprises to improve their content creation efficiency.

Column (1) presents the regression estimation results obtained using Equation (32); the estimated coefficient of the indicator of the level of digital transformation in cultural enterprises (lnINT) was −0.004 at the 1% level of significance. Column (2) presents the regression estimation results obtained using Equation (33); the estimated coefficient of the indicator of the proportion of intangible assets in cultural enterprises (Intar) was −0.218 at the 1% level of significance. These results indicate that a cultural enterprise with a higher level of digital transformation has a smaller backlog of intangible assets that have yet to be converted into cultural products and is more efficient at transforming intangible assets. This increase in the efficiency of content creation can significantly contribute to the enhancement of the TFP of cultural enterprises. Thus, Hypothesis 2a is confirmed.

5.2. Financing Ability Testing

Although traditional financial institutions, such as banks, tend to focus on the value of an enterprise’s tangible assets when considering whether to lend to it, cultural enterprises are primarily involved in content creation, which does not require a large amount of special equipment. This production characteristic has generally led to high financing constraints for cultural enterprises. Digital transformation can alleviate this problem. A digital content creation platform increases the standardization and transparency of cultural enterprises’ production process, and it provides lenders with an efficient means of monitoring cultural enterprises’ internal production status, thereby reducing credit risk and, ultimately, increasing the cultural enterprises’ financing ability. Moreover, digital transformation can generate considerable benefits for cultural enterprises and can serve as a new direction for government development in China; that is, digital transformation can not only help cultural enterprises attract more capital from the capital market but also reduce the risk of external investment by attracting government credit endorsement. Therefore, digital transformation can improve the financing ability of cultural enterprises and thereby increase their TFP.

This subsection describes the empirical testing of the influence mechanism, which involved continuing the construction of the mediation effect model. For the selection of a proxy indicator, the tangible asset–liability ratio indicator (TDR) was used to proxy an enterprise’s financing ability [60]. The intangible assets of a cultural enterprise were deducted from its total assets to calculate the tangible asset gearing ratio, which reflects the cultural enterprise’s real financing capacity. Cultural enterprises with a higher tangible asset gearing ratio are more likely to be capable of financing their operations. Two additional regression equations were constructed for mediation effect testing as follows:

$$\begin{array}{ll}TD{R}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Finex{r}_{i,t}+{\beta }_{3}Fix{r}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Siz{e}_{i,t}\\ & +{\beta }_{6}Boar{s}_{i,t}+{\beta }_{7}Dua{l}_{i,t}+{\beta }_{8}EA{C}_{i,t}+{\beta }_{9}Ag{e}_{i,t}+{\beta }_{10}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(34)

$$\begin{array}{ll}\ln L{P}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}TD{R}_{i,t}+{\beta }_{3}Finex{r}_{i,t}+{\beta }_{4}Fix{r}_{i,t}+{\beta }_{5}RO{A}_{i,t}+{\beta }_{6}Siz{e}_{i,t}\\ & +{\beta }_{7}Boar{s}_{i,t}+{\beta }_{8}Dua{l}_{i,t}+{\beta }_{9}EA{C}_{i,t}+{\beta }_{10}Ag{e}_{i,t}+{\beta }_{11}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(35)

If the mechanism is verified to exist, β₁ is significantly positive in the estimation results of Equation (34), and β₂ is significantly positive in the estimation results of Equation (35).

Table 11. Regression estimation results for the financing ability mechanism.

Variable	(1)	(2)
	TDR	lnLP
lnINT	0.016 **	0.005 **
	(2.37)	(2.50)
TDR		0.023 ***
		(2.67)
Intercept	−2.321 ***	1.329 ***
	(−9.50)	(17.01)
Year Fixed Effects	Yes	Yes
Firm Fixed Effects	Yes	Yes
N	1541	1541
R2	0.352	0.208

*** and ** indicate significance at the 1% and 5% levels, respectively; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

presents the regression estimation results.

Column (1) presents the regression estimation results obtained using Equation (34); the estimated coefficient of the indicator of the level of digital transformation in cultural enterprises (lnINT) was 0.016 at a 5% level of significance. Column (2) presents the regression estimation results obtained using Equation (35); the estimated coefficient of the indicator of the level of digital transformation in cultural enterprises (TDR) was 0.023 at the 1% level of significance. These results indicate that digital transformation improves the financing ability of cultural enterprises, and this increase in financing ability can significantly increase cultural enterprises’ TFP. Thus, Hypothesis 2b is confirmed.

5.3. R&D Innovation Testing

R&D innovation is crucial for cultural enterprises which are primarily focused on content creation. Through digital transformation, cultural enterprises can improve their innovation efficiency, promote R&D innovation activities, and acquire competitive advantages quickly through the creation of high-quality products, and they can generate much more economic profit by rapidly outperforming their competitors. Various innovation elements within cultural enterprises can be stimulated through digital transformation, including the acquisition of positive externalities and reduction of costs associated with achieving each unit increment in innovation results. Furthermore, digital transformation alleviates the excessive financing constraints experienced by cultural enterprises and influences their investment decisions regarding R&D innovation. Therefore, digital transformation can increase the TFP of cultural enterprises by encouraging them to invest in R&D innovation.

This subsection describes the construction of the mediation effect model for R&D innovation testing. For the selection of proxy indicator, a higher level of R&D innovation efficiency is associated with greater focus of a cultural enterprise on R&D activities, which leads to greater investment in R&D innovation. The R&D innovation investment intensity indicator (RDI) is a proxy variable for measuring the focus of a cultural enterprise on R&D innovation after it is standardized based on its total assets. A higher value for this indicator indicates that a cultural enterprise is investing more in R&D innovation in a given year, indicating a higher level of innovation and R&D focus. In the empirical model setup, two mediation effect test equations were constructed, namely Equations (36) and (37):

$$\begin{array}{ll}RD{I}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}Finex{r}_{i,t}+{\beta }_{3}Fix{r}_{i,t}+{\beta }_{4}RO{A}_{i,t}+{\beta }_{5}Siz{e}_{i,t}\\ & +{\beta }_{6}Boar{s}_{i,t}+{\beta }_{7}Dua{l}_{i,t}+{\beta }_{8}EA{C}_{i,t}+{\beta }_{9}Ag{e}_{i,t}+{\beta }_{10}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(36)

$$\begin{array}{ll}\ln L{P}_{i,t}& ={\beta }_0+{\beta }_1\ln IN{T}_{i,t}+{\beta }_{2}RD{I}_{i,t}+{\beta }_{3}Finex{r}_{i,t}+{\beta }_{4}Fix{r}_{i,t}+{\beta }_{5}RO{A}_{i,t}+{\beta }_{6}Siz{e}_{i,t}\\ & +{\beta }_{7}Boar{s}_{i,t}+{\beta }_{8}Dua{l}_{i,t}+{\beta }_{9}EA{C}_{i,t}+{\beta }_{10}Ag{e}_{i,t}+{\beta }_{11}SO{E}_{i,t}+{\mu }_i+{\nu }_t+{\epsilon }_{i,t}\end{array}$$

(37)

If the mechanism is verified to exist, β₁ is significantly positive in the estimation results of Equation (36), and β₂ is significantly positive in the estimation results of Equation (37).

Table 12. Regression estimation results for R&D innovation enhancement mechanisms.

Variable	(1)	(2)
	RDI	lnLP
lnINT	0.001 **	0.005 **
	(2.40)	(2.53)
RDI		0.262 **
		(2.06)
Year Fixed Effects	Yes	Yes
Firm Fixed Effects	Yes	Yes
N	1541	1541
R2	0.137	0.207

** indicates significance at the 5% levels; Cluster-robust t-estimates are parenthesized; N is the sample size used for estimations, and R² is the goodness of fit; “Yes” indicates that a given type of fixed effect was controlled for.

reports the regression estimation results.

Column (1) presents the regression estimation results obtained using Equation (36); the indicator estimated coefficient of the level of digital transformation in cultural enterprises (lnINT) was 0.001 at the 5% level of significance. Column (2) presents the regression estimation results obtained using Equation (37); the indicator estimated coefficient of the level of investment in R&D innovation (RDI) was 0.262 at the 5% level of significance. These results indicate that a higher level of digital transformation is associated with cultural enterprises focusing more on improving their R&D innovation abilities. This increase in R&D innovation can significantly promote the TFP of cultural enterprises. Thus, Hypothesis 2c is verified.

6. Research Conclusions and Policy Recommendations

6.1. Research Findings

Using sample data from the 2011–2019 annual reports of 251 listed cultural enterprises in China, the current authors empirically analyzed the effects of digital transformation on the TFP of cultural enterprises and their paths of action and arrived at three key conclusions. (1) Digital transformation can significantly increase the TFP of Chinese cultural enterprises; this finding provides a theoretical foundation for promoting the efficient transition from a traditional economic model to a digital economic model in such enterprises. This conclusion is supported by the findings of a robustness test, and the endogeneity problem was addressed. (2) Digital transformation has heterogeneous effects on the TFP of cultural enterprises. Notably, the effects of digital transformation were more significant on enterprises that were smaller in size, located in midwest China, and positioned in the upstream chain of segmented industries. (3) Digital transformation can enhance the TFP of cultural enterprises by enhancing their content creation efficiency, financing ability, and R&D innovation.

Specifically, the finding in (2) above that the effects of digital transformation were more significant on enterprises that were smaller in size, and located in midwest China, is the novelty of the present study. There are some views that are consistent with this finding [61,62] and some that are inconsistent with [63,64]. The current authors argue that the digital transformation of cultural enterprises in underdeveloped regions of midwest China and which are smaller in size can maximize the “refurbishing” effect. That is, digital transformation can greatly alleviate the current developmental challenges faced by cultural enterprises of smaller size and in undeveloped regions by improving the efficiency of content creation and financing capabilities.

6.2. Policy Recommendations

China is undergoing a transitionary period in which it is promoting the digital development of cultural enterprises, and academics should elucidate the digital transformation process of cultural enterprises to enable traditional enterprises to successfully implement digital upgrading and conclude experiences for global enterprises. Based on a sample of 251 listed cultural enterprises in China, the authors empirically examine the mechanism and effect of digital transformation on TFP. In addition to being useful for understanding how digital elements within enterprises enable them to become more efficient and enter into sustainable development modes at the micro level, the results of this study are also useful for the government to improve regulations in order to enhance enterprise data use and transform into a self-propelled and sustainable development mode. The following recommendations are proposed on the basis of both the theoretical analysis and empirical findings of the present study.

First, to achieve high-quality development goals, Chinese cultural enterprises should continue to deepen their digital transformation. They can employ digital technology more widely to facilitate creation, increase their innovation in digital business models to improve their operational efficiency, create more and higher-quality digital cultural products, and further explore the economic value of traditional cultural resources. Cultural management departments should continue to implement relevant regulations and policies to support cultural enterprises through their technical transition and to assist them in achieving digital transformation and upgrading.

Second, the Chinese government should coordinate the development of various types of cultural enterprises. To identify cultural enterprises that would implement digital transformation, relevant management departments of the Chinese government should coordinate the development of cultural enterprises of varying sizes and from various industry segments. As cultural enterprises vary in size, government policies should be adapted to actual circumstances, and funds should be more precisely injected into cultural enterprises that are exploring and deepening digital transformation to facilitate this process.

Third, the Chinese government should coordinate and guide cultural enterprises in east and midwest China in strengthening their economic exchanges. Introducing appropriate incentives can help cultural enterprises based in east China to share their advanced digital development experience and knowledge with cultural enterprises based in midwest China. Such exchanges can promote in-depth cooperation between the two regions and create opportunities for them to jointly develop and use the rich cultural resources of midwest China.

The present study has several limitations that should be addressed in future research. First, because of data limitations, this study uses total factor productivity (TFP) to measure the economic performance of enterprises. However, it is unable to explore the effects of digital transformation on the social performance of cultural enterprises. In follow-up studies, inbound interviews and tracking surveys should be conducted to collect data on indicators that can be used to measure listed enterprises’ subjective willingness to engage in social responsibility. Based on this information, the present study findings can be expanded. Second, the present study only discussed how digital transformation affects the TFP of Chinese cultural enterprises. However, Digital factors are increasingly interconnecting and interacting with each other, and this trend is likely to lead to more extensive and in-depth data exchanges between China and other countries. In a follow-up study, the present authors will collect data pertaining to the export product structure and export trade volume of Chinese cultural enterprises, and these data will be explored in a detailed and comprehensive manner to determine how digital transformation affects cross-border trade and cross-border services from the perspective of Chinese cultural enterprises.