Andrin Pelican,Bryan S. Graham
Abstract:Consider a setting where $N$ players, partitioned into $K$ observable types, form a directed network. Agents' preferences over the form of the network consist of an arbitrary network benefit function (e.g., agents may have preferences over their network centrality) and a private component which is additively separable in own links. This latter component allows for unobserved heterogeneity in the costs of sending and receiving links across agents (respectively out- and in- degree heterogeneity) as well as homophily/heterophily across the $K$ types of agents. In contrast, the network benefit function allows agents' preferences over links to vary with the presence or absence of links elsewhere in the network (and hence with the link formation behavior of their peers). In the null model which excludes the network benefit function, links form independently across dyads in the manner described by \cite{Charbonneau_EJ17}. Under the alternative there is interdependence across linking decisions (i.e., strategic interaction). We show how to test the null with power optimized in specific directions. These alternative directions include many common models of strategic network formation (e.g., "connections" models, "structural hole" models etc.). Our random utility specification induces an exponential family structure under the null which we exploit to construct a similar test which exactly controls size (despite the the null being a composite one with many nuisance parameters). We further show how to construct locally best tests for specific alternatives without making any assumptions about equilibrium selection. To make our tests feasible we introduce a new MCMC algorithm for simulating the null distributions of our test statistics.
What problem does this paper attempt to address?
This paper aims to address the question of whether there is strategic interaction in the process of network formation. Specifically, the author attempts to construct a statistical test to detect whether, in network formation, individuals establish links independently based solely on their own preferences, or whether their decisions are influenced by the behavior of other individuals, that is, there is strategic interaction.
### Background and Objectives of the Paper
In the study of social and economic network formation, individuals (or agents) may choose to establish connections with other individuals based on multiple factors. These factors can include individual characteristics (such as gender, occupation, etc.), position in the network (such as centrality), and relationships with other individuals (such as homophily). However, a key question is whether individuals will consider the behavior of other individuals when deciding to establish links, that is, whether there is strategic interaction.
### Paper Methods
To detect the existence of such strategic interaction, the author proposes an optimal test method. This method is based on the following assumptions:
- **Null Hypothesis** (\(H_0\)): There is no strategic interaction, that is, the link - decision between individuals is independent.
- **Alternative Hypothesis** (\(H_1\)): There is strategic interaction, that is, an individual's link - decision is influenced by the behavior of other individuals.
### Model Specification
1. **Network Structure**:
- Consider \(N\) individuals, each belonging to one of \(K\) observable types.
- Directed links can be formed between individuals, forming a directed graph \(G(V, A)\), where \(V\) is the set of individuals and \(A\) is the set of directed links.
2. **Utility Function**:
- The utility function of individual \(i\) consists of two parts:
\[
\nu_i(d_i, d_{-i}; \theta, U_i)=\gamma_0 g_i(d)-\sum_j d_{ij} c_{ij}(X_i, X_j; \delta, U_{ij})
\]
where:
- \(\gamma_0 g_i(d)\) represents the benefits brought by the network, which may include the benefits of direct and indirect connections.
- \(\sum_j d_{ij} c_{ij}(X_i, X_j; \delta, U_{ij})\) represents the cost for individual \(i\) to maintain links, including the attributes \(X_i, X_j\) of itself and the other party and the unobserved heterogeneity \(U_{ij}\).
3. **Cost Function**:
- The link cost function \(c_{ij}(X_i, X_j; \delta, U_{ij})\) is defined as:
\[
c_{ij}(X_i, X_j; \delta, U_{ij})=-[A_i + B_j+X_i'\Lambda_0 X_j - U_{ij}]
\]
where:
- \(A_i\) and \(B_j\) respectively represent the unobserved heterogeneity of individual \(i\) initiating the link and individual \(j\) receiving the link.
- \(X_i'\Lambda_0 X_j\) represents the systematic benefits brought by the attribute matching between individuals \(i\) and \(j\).
- \(U_{ij}\) represents the random preference shock of individual \(i\) for link \(ij\).
### Test Methods
1. **Distribution under the Null Hypothesis**:
- Under the null hypothesis, link - decisions are independent and can be described by the binary log - regression model proposed by Charbonneau (2017).
- This model can match any in - degree and out - degree sequences and homophily matching patterns in the actual network.
2. **Similarity Test**:
- The author uses the exponential family structure and the conditional score test method to construct a similarity test that can precisely control the size of the test under the null hypothesis.
- Calculate the critical value by simulating the reference distribution under the null hypothesis through the MCMC algorithm.
3. **Locally Optimal Test**:
- The author also derives the locally optimal test statistic under a specific alternative hypothesis.