**3. The Model**

### *3.1. The History-Friendly Model*

As an emerging industry in China, there are limited statistics of the fuel ethanol industry. Therefore, it is difficult to analyze the mechanisms and factors affecting the fuel ethanol industry using a statistical model. In order to explore the coevolutionary relationships between the fuel ethanol industry, the technology system, and the market system, this paper employs a history-friendly evolutionary model which has been applied to many industries, including computers, DRAM chips, pharmaceuticals, semiconductors, synthetic dyes, and mobile phones and memory chips [41–46]. Scholars studying industrial dynamics generally rely quite heavily on the appreciative theory which is a body of verbal

arguments representing causal explanations of observed patterns of economic phenomena [47,48]. Although the appreciative theory is an appropriate tool to characterize the main mechanisms at work, it is di fficult to verify the logical consistency of the theory due to its complexity and the lack of precision of the verbal language [47]. The history-friendly model is a formal model of the appreciative theory and can overcome the above limitations of the appreciative theory [41]. It aims to analyze, in a more formal form, the influential factors and their influencing mechanisms in industry evolution, technological progress, and institutional change that have been confirmed by appreciative theory [41,49].

The construction of the history-friendly model consists of three important steps [47]. The first step is the selection of the stylized facts deserving attention from theoretical perspectives. These stylized facts mainly refer to the history and evolution of the fuel ethanol industry such as specific institutions, technological change, and market characteristics. Second, there is the choice of how to represent the selected phenomena. In this respect, the history-friendly model adopts the same basic representations used in evolutionary models. All reported models are built around four main blocks: firm behavior, technological change, market demand, and industry dynamics [50,51]. The creation, entry, exit, and technological change of the business firms a ffect the performance of the industry and further impact industry evolution. The third step is the manipulation and implementation of the model designed in the second step.

The history-friendly model is a type of agent-based simulation model dealing with the complexity of the economic system [52]. A typical history-friendly model has many variables and parameters. Under a wide range of parameter settings, some of the parameter settings will lead to the replication of the industry history being modelled. Importantly, "replication" here mainly refers to qualitative reproduction, not quantitative reproduction [53]. Once the model is built, there is room for wider applications such as policy analysis. The history-friendly nature is threefold. Firstly, in the process of model construction, stylized facts in industrial development are fully considered, and the relationship between variables is constructed on this basis. Secondly, the initial values of variables in the model are set based on the true values of industrial history. Thirdly, the selection of the parameters' values can qualitatively reproduce the stylized facts in industrial history.

There are two compelling reasons for using a history-friendly model in this paper. First, a history-friendly model helps us better explore the causal mechanisms in the evolution of the fuel ethanol industry. As a formal model, all the logic that drives model outcomes is explicitly represented in a history-friendly model [27]. In addition, the mechanisms built into the model are transparent which means that if the model does not work as expected, the analyst can adjust the settings of the model until the model is able to qualitatively capture the stylized facts in the appreciative theory [54]. Developing and working through a history-friendly model could bring to mind mechanisms, factors, and constraints of the industry evolution [41]. Therefore, compared with the appreciative theory, the history-friendly model is conducive to analyzing the causal mechanism of the fuel ethanol industry. Second, the model setting of the history-friendly model is transparent rather than arbitrary, so, it serves as a good starting point for further policy analysis. Comparing the influence of di fferent systems and policy arrangements on industry evolution can provide a deep understanding of the influence mechanism of the above factors and provide a basis for further policy selection and institutional arrangemen<sup>t</sup> [41,55].

### *3.2. The Model Specification*

The basic model is presented in this section. Given the complexity of the history-friendly model, it is di fficult to lay out all the details of all the equations without confusing the reader and obscuring the basic logic of the model [41]. Therefore, we have tried to present only the equations which can reflect the stylized facts of the fuel ethanol industry and put other related equations in the Appendix A. Just like most of the other history-friendly models [50,51], our model is also built around four main building blocks: firm, technology progress, market demand, and industrial dynamics. In selecting the stylized facts to investigate, we considered their relevance on the selection of the variables and the

relationships among these variables, which a ffect the model specification [47]. These stylized facts are put at the beginning of each block. In addition, given that our model involves many subjects which are further divided into di fferent types, we use a lot of superscripts to minimize the number of variables used in the model. The variables with superscripts *b* and *f* represent the variables associated with the properties of fuel ethanol and fossil fuels, respectively. The variables with superscripts *tf*, *nf*, and *rd* represent the variables associated with the properties of traditional technology firms, new technology firms, and R&D firms, respectively. The variables with superscripts *m* and *t* represent the variables associated with the properties of materials and traditional materials, respectively.
