In the Web2.0 era, with the rapid development of information technology and popularization of the internet, people are endowed with higher capabilities and more ways to create value. In the past, they were just “consumers” of Internet content. Now, they are also “producers”. In the internet platform, every user can display their original content and share it with others. This behavior is defined as “UGC”, which means user-generated content. UGC has developed rapidly by virtue of its advantages such as proximity to users, high efficiency of resource allocation, flexibility and autonomy, attracting a large number of enterprises and scholars. For every UGC platform, the user’s productivity determines the survival and development of the platform. However, in the platform, there is no employment relationship, and users are unable to get direct monetary and material reward most of the time ([12]). So why do users voluntarily invest time and effort in content production? What drives their behavior? What kinds of satisfaction do they get? Understanding these issues is critical to the sustainability of UGC platforms.

At present, most of the researches on user participation motivation of UGC platform are conducted through surveys/questionnaires, and the results are susceptible to subjectivity. Using the actual micro-level data from the Epinions, an online review community, this paper builds up a dynamic discrete choice model to delve into the participation motivations of users in UGC platform. For the solution of the model, we adopt the two-stage estimation method proposed by [2].

Our results suggest that users can obtain enjoyment and entertainment utility and information acquisition utility from posting and reading reviews in the Epinions. When users are trusted by others, or their posted reviews are recognized and praised, users will have a strong sense of identity. The higher the degree and frequency of users’ community contribution, the more they trust other users in the community, and the stronger their sense of community attachment will be.

Based on the existing literature ([2]; [11]; [6]; [5]; [14]), this study constructs a dynamic discrete choice model (DDCM) to characterize user participation behavior in Internet UGC society by combining satisfaction theory, achievement motivation theory and social identity theory. The dynamic discrete choice model is a microeconometric model that is widely used in economics and other social science research. For example, [16] applied this model to retirement decisions; [9] constructed a dynamic choice model about criminal behavior; It is based on utility maximization theory to study decision makers’ action choice problems, which is of great significance for understanding decision-making behavior under different situations. Its basic principles are as follows:

At each time t, t ∈ { 1 , 2 , ⋯ , T } , the decision maker has a s t , s t ∈ S , in which he chooses action a t from finite set A t and obtains utility U ( a t , s t , θ ) , where θ is a vector parameter. At the same time, because the decision-maker gains utility in the current time, his state s t will change and be updated to s t + 1 at the next time. The utility U ( a t , s t , θ ) of a decision-maker can be regarded as consisting of a deterministic component u ( a t , s t , θ ) and a random perturbation term γ t :

γ t satisfies independent and identical distribution for different t.

Because the utility of the decision maker at time t affects the state at time t + 1 and then affects the utility at the next time, the decision maker will consider the impact of the current action decision on the future utility and make a decision to maximize his total utility:

β is the utility discount factor.

The dynamic discrete choice model is a very effective and practical research tool for studying individual choice behavior. Starting from economic theory, it can simulate actual situations and restore individual choice processes by abstracting and describing realistic problems mathematically instead of simply fitting data. By analyzing the magnitude, sign, and significance of parameter θ estimation value of the dynamic discrete choice model, researchers can understand which factors influence decision-makers’ action choice and what the direction and importance of their influence are, which provides essential reference basis for understanding and revealing the behavior decision of individuals, enterprises or organizations. On the Internet UGC platform, user participation is a decision-making choice problem. Users can adjust their behavior decisions dynamically better to meet their psychological, economic, and social needs and maximize their total utility. Therefore, the dynamic discrete selection model can be applied to this research.

Figure 1. Two-stage estimation process.

To solve the structural model, we employ the two-stage estimation method (BBL) proposed by [2]. This method, developed from the work of [14] and [7], is based on the Markov Perfect Equilibrium (MPE) concept proposed by [5]. It assumes that the behavior of decision-makers in the model follows Markov processes, where each decision-maker’s behavior is determined by the current state and random disturbances. The two-stage estimation method, illustrated in Figure 1, involves two stages. In the first stage, we solve the decision equation, the distribution, and transition probability of the state variables, and the utility function of the decision maker at equilibrium. In the second stage, we estimate the model parameters by combining the results of the first stage with the equilibrium inequalities of the model. This method has been proven effective for solving various dynamic choice models, including discrete, continuous, or single multi-agent problems ([2]).

The two-stage estimation process is shown in Figure 1.

The data sample is from the Epinions, a product review community owned by eBay. It covers a total of 91,338 users, including their home page, trust list, trusted list, list of posted reviews, and comments of other users. For research purposes, we constructed this set of data into a set of panel data, with a period of one month. We selected 210 ordinary users and obtained a total data of 24 periods.

Suppose there are i = 1 , 2 , ⋯ , I users in Epinions, they need to make posting and reading decisions in each period t = 1 , 2 , ⋯ , T . Let a i t ∈ A i denotes the action of user i at period t , a i t = ( n i p t , n i r t ) , among them, n i p t is the number of reviews that user i posts at time t , n i r t is the number of reviews that user i reads at time t .

Based on the use and satisfaction theory, we believe that the reason why users voluntarily participate in UGC platform is that in this process, users’ enjoyment and entertainment needs, identity needs, community attachment needs and information acquisition needs can be satisfied. Next, this paper will build up a dynamic discrete choice model around these four motives.

1)Enjoyment and entertainment

Enjoyment and entertainment motivation refers to the user’s desire for fun and recreation in the process of generating content, which is an important motivation for users to participate in the UGC platform ([1]). Users interact with each other on topics they are interested in, sharing their experiences and insights, from which they can obtain emotional comfort and pleasure, and relieve the pressure brought by life. This part of the utility is represented by a function φ i t :

Here, n i t p is the number of reviews posted by user i at period t . The quadratic function is used to take into account that within a certain time, each additional review posted by users brings a sense of amusement with diminishing marginal utility. In addition, enjoyment and entertainment motivation is an intrinsic motivation of users, but now many platforms will set monetary and material rewards to incentivize active participation. When users post too many reviews due to these external incentives, their enjoyment and sense of entertainment will be squeeze out ([13]; [10]).

2)Identity

Identity refers to the confidence and sense of value that users gain from others’ recognition and appreciation. This motivation is included in virtually all researches on UGC platform ([15]; [3]). By sharing their knowledge, experience and talents that others are interested in, users can obtain recognition and appreciation, thus enhancing their sense of achievement and personal identity. In the utility function, we use ω i t to express the identity part:

Among them, P i t and T i t respectively represent the number of praises and trusters that user i has accumulated as of period t , r b i t represents whether the user’s post is approved by the platform, and the three are state variables:

p i t and t b i t respectively represent the number of praises and trusters that user i obtained in the period t . δ and μ are the utility attenuation rates and we set them to be 0.8.

3)Community attachment

According to social identity theory, individuals belonging to a particular social group will perceive the emotion and value gained as members of that group. Users with a strong sense of community are more loyal to the platform, so they are more active in content contribution ([15]; [8]). The main factors that affect the community attachment include users’ emotional identification with the community and other members, as well as users’ degree of participation. The specific form of community attachment utility function is as follows:

Here, T T i t represents the cumulative number of people that user i trust to as of period t , K S i t is the amount of knowledge that has been shared by user i as of period t , h i t represents the number of periods since the user last post a review, and the three are state variables:

t t i t and n i t p are the number of users trusted and the number of reviews posted in period t , respectively. δ and μ are the utility attenuation rates and we set them to be 0.8.

4)Information acquisition

As a public information gathering center, UGC platforms provide users with valuable information and resources. When users participate in UGC platform, they can learn about things they are interested in, broaden their knowledge scope and improve their cognition level ([4]). We use χ i t to express the utility of this part:

Here, K A i t is the amount of information that has been acquired by user i as of period t . We represent users’ information acquisition by the number of reviews they read. K A i t is a state variable:

n i t r indicates the number of reviews that user i read in period t , α is the utility attenuation rate of the acquired knowledge, and we set it to be 0.8.

5)Cost

At period t , the total cost for user i to post and read reviews is:

The cost here mainly refers to the time and effort spent by users in posting and reading reviews.

To sum up, the utility of user i in period t is given by U i t ：

Here, ν i t is the random disturbance of utility and it obeys an i.i.d normal distribution.

This research assumes that users are forward-looking and they make decisions to maximize the sum of the discounted values of utility for each period, so their goal is:

Here, β is the discount rate, we set β = 0.8 to balance short-term benefits and long-term utility.

Parameters that need to be estimated are: θ = ( θ 1 , θ 2 , θ 3 , θ 4 , θ 5 , θ 6 , θ 7 , θ 8 , θ 9 , θ 10 , θ 11 ) .

We use the two-stage estimation method proposed by [2] to estimate the model. It assumes that users’ behavior is consistent with Markov perfect equilibrium (MPE), which means each user’s action decisions are determined only by the current state and random disturbance.

The two-stage estimation process is shown in the above Figure 1.

In the first stage, we estimate the action function and the state transition, then we can get the utility function in equilibrium. According to our hypothesis, the action strategy of user i is determined by the current state s i t and the random disturbance ν i t , that is a i = σ i ( s i t , ν i ) . Therefore, the optimal decision can be estimated as long as the distribution of action strategy under a certain state s and random shock ν is known, which can be obtained from the existing data. To estimate the state transition, we can use the observed value of state variables.

Let V i ( s ; σ ; θ ) represents the utility function of user i in state s , and assume that the action strategy σ i is in line with the Markov strategy, there are:

when the state transition and decision equation has been estimated, the utility function V i ( s ; σ ; θ ) can be obtained by forward simulation.

In the second stage, we use the estimation results of the first stage and the equilibrium conditions of the model to estimate parameters. According to the hypothesis, users’ action strategy is consistent with MPE, so for any user i in any state s , there are:

Let x represents a group of ( i , s , σ ′ i ) ,we define the function:

where H is the distribution of the set x .

As θ 0 is the true value of θ , then Q ( θ 0 ) = 0 . We can minimize the function Q ( θ ) to estimate the value of parameter θ .

It should be noted that both θ 2 n i t p and θ 10 n i t p consist of n i t p , so we combine these two terms and estimate them.

Based on the above results, the main conclusions are summarized as follows.

First, users who post articles on Epinions get enjoyment and entertainment utility. This hedonic utility is “inverted U,” which increases with the number of user posts at first. However, the hedonic utility decreases when the number of user posts reaches a specific value. This is mainly because external incentives such as money and material gain have a “crowding out effect” on intrinsic motivation.

Second, users gain a sense of identity when they publish articles that are well-received or gain the trust and attention of others. However, if this kind of praise and trust comes from “junk accounts” such as “marketing numbers”, it will not help the user’s utility improvement and even have a negative impact.

Third, when users publish more comment articles in the community, participate heavily, and trust other users in the community, they will gain a strong sense of community belonging. They will form stronger emotional ties with the platform, and the stronger the sense of dependence and loyalty to the platform, the more motivated they will be to continue content production and construction on the platform. If users post articles at long intervals, they will also reduce their participation and, thus, their sense of community.

Fourth, the Epinions community can provide users with rich knowledge and experience sharing to meet their information acquisition needs. However, due to the large number of users and the large amount of information on the platform, users need to waste more time searching, filtering, and identifying the information they are interested in and need, which also causes greater utility loss.

Finally, this paper puts forward the following guiding suggestions for the Internet UGC platform to encourage users to participate actively and continuously contribute: 1. When designing incentive mechanisms, we should pay attention to the mutual influence and balance among various mechanisms and fully consider the possibility of external incentives such as money to “squeeze out” users’ intrinsic motivation. 2. Feedback mechanisms can be used to strengthen the theory of user management to achieve clear rewards and punishments and improve the quality of the platform’s content contribution. Maintaining network order and information security, strengthening content supervision, establishing strict user screening mechanisms and scientific and practical reporting mechanisms, strictly eliminating illegal information, and creating a safe and healthy network environment. Develop more standardized collaborative production models and systems, improve convenience and intelligence, and reduce user content production and information acquisition costs. In addition, the Internet UGC platform should form its own unique platform culture and positioning, choose the direction suitable for its development, provide users with more innovative and personalized experiences, and obtain long-term development momentum.

The main contribution of this study lies in constructing a structured model based on classical theory and objective data to analyze the participation behavior of users on the Internet UGC platform, explaining the motivation and utility of users’ participation without apparent employment relationship and usually unable to obtain monetary reward. This overcomes the problems of previous studies affected by factors such as self-cognitive bias. The conclusions obtained have specific guiding significance for developing and constructing the UGC platform. The author hopes that more scholars will conduct relevant research on this basis in the future, put forward more targeted and constructive opinions, promote the vigorous development of the Internet UGC platform, and achieve more significant value creation.

To reduce the difficulty of model solving, this study sets the utility discount factor and attenuation rate to a known constant of 0.8. A more accurate approach would be to estimate them as unknown parameters, but this would lead to excessive computational complexity and require the development of new algorithms. Future research will focus on improvements and enhancements in this regard.

*Corresponding author.