The Five Essentials of a Modeling Paper

1. Introduction

Formal methods sparked the scientific revolution, starting in physics (Newton 1687). Over the centuries, they have played an increasingly important role in disciplines ranging from chemistry and biology to economics and sociology. Strategic management is no exception. Many seminal works have had a model at their core, covering a wide variety of topics, including the resource-based view (Lippman and Rumelt 1982), exploration versus exploitation (March 1991), commitment (Ghemawat 1991), value-based strategy (Brandenburger and Stuart 1996), and organizational adaptation (Levinthal 1997). The key benefits of models for developing management theory are their precision, logical consistency, and derivation of nonobvious implications (Adner et al. 2009, Knudsen et al. 2019, Csaszar 2020, Hannah et al. 2021, Makadok 2022). The hallmarks of good models are that they are not unnecessarily complicated and lead to relevant results that challenge the conventional wisdom (Knudsen et al. 2019, Csaszar 2020).

Although these prior works provide a good overview of the “why” and “what” of modeling, they are less clear about the “how.” For instance, Csaszar (2020) provides an excellent and comprehensive analysis of why we need models and what constitutes a good model but is less specific about the most common issues modeling papers face in practice and how to exactly address them. Although Makadok (2022) does cover the “how,” it is very specific in its focus to make models accessible for the Academy of Management Review audience. What in our view is missing is a more general practical guide for authors on how to address the most common issues facing modeling papers.

With this editorial, we hope to provide such a guide. Having edited and reviewed many manuscripts, we find that almost all rejections are caused by a relatively small number of issues. As authors, we had to learn how to overcome these issues the hard way, often by being rejected ourselves. With this editorial, we hope to spare future modelers the same fate. We identified five issues that we have come across again and again and provide guidance on how to overcome them. Because what constitutes a good model has been extensively covered in the earlier works cited above, in this piece, we focus squarely on the “how to” of publication. Needless to say, however, that following the advice contained in this editorial does not guarantee publication in Strategy Science or another journal. Some scholars consider the ultimate test for publishability the “insight-to-ink-ratio,” and this will ultimately be subjective and based on reviewers’ and editors’ scholarly preferences. What this editorial aims to do is to enable authors to present the insights they want to convey by means of a model in the most effective way. That way, readers can focus on the content and the depth of the insights rather than having to work their way through an opaque presentation.

Our guidance applies to the many different potential forms of (formal) modeling. For us, the key defining element is that the reasoning is based on formal assumptions, usually in the form of mathematical equations or algorithmic procedures. Such assumptions allow for highly reproducible theoretical reasoning, which is one of the most prominent benefits of using a model to develop a theory. We thus include analytical models with formal proofs (Lippman and Rumelt 1982); computational models in which key results are calculated rather than analytically derived (Wibbens et al. 2025); simulations based on random draws of (initial) conditions (Siggelkow and Rivkin 2006); and combinations of these. Models can serve as stand-alone theory pieces (Adner et al. 2009) or be used to guide the theory of empirical papers (Hannah et al. 2021). This editorial covers both.

Table 1 shows the key three elements of a model. The differences across the different types of models will be most salient in step 2, the inference. Still, most modeling papers will rely on a combination of formal derivation and computation. For instance, simulation models often use some mathematics to derive intermediate results needed for computation. Conversely, analytical papers relying mostly on formal proofs often use some numerical calculations to illustrate results. Ultimately, all models yield insights how certain sets of parameters (i.e., contextual conditions) lead to various outcomes, using highly reproducible inference.

Table 1. Elements of a Formal Model

Although we believe that the five “essentials” discussed in Section 2 are broadly applicable, there are two classes of models that perhaps require a different emphasis. The first class consists of simulation models that aspire to capture real life as realistically as possible by including many parameters and their interactions, the second of models as part of empirical papers. We briefly discuss them in Section 3.

2. The Five Essentials

2.1. Contribution to Management

Models are no exception to the requirement that papers in Strategy Science need to “[enhance], in some meaningful manner, our understanding of some substantive issues in the strategy domain.”¹ Importantly, the sole fact that something has not been studied before is a poor motivation. Rather, a good paper needs to help solve a pertinent issue in strategic management.

The best way to motivate a paper is by identifying a key issue in an active conversation in the literature. The paper’s introduction needs to clearly articulate the current state of the literature (sometimes called the “null”) and its novel contribution (the “delta”). This requires an in-depth knowledge of the literature you want to speak to.

Especially for authors from other disciplines such as economics, it is critical to actively engage with the scholarly community in strategic management. Although this may almost seem too obvious to state explicitly, we regularly see papers as editors and reviewers that fail to do this beyond a few cursory citations to age-old strategy papers. Before submission therefore, you should seek feedback from colleagues in strategic management and present your paper, for instance, at the annual conferences of Strategy Science, the Academy of Management, or the Strategic Management Society. Presenting helps ensure that the contribution you seek to make is truly novel and important, as well as that you motivate your study in a language that appeals to your target audience. Verbal framing—still essential, even in modeling pieces—often evokes different reactions from different people. Feedback from your audience is critical to ensure that your framing evokes the intended understanding.

You also need to describe the implications of the model for managers. Some of the most impactful modeling papers directly challenge the conventional wisdom for managerial practice, yet are grounded in evidence or real-world plausibility and/or resolve a puzzle in the literature. Basically, generating “counterintuitive” results that no practitioner finds plausible or that work off a “strawman” argument from outdated literature is unlikely to generate much interest. To see how models can generate immediate and novel insights, consider, for example, Siggelkow and Rivkin (2006), who use an agent-based simulation to show that exploration at low hierarchical levels of the organization can backfire. This theoretical finding has direct practical implications for how managers should organize for exploration. As another example, Wibbens et al. (2025) use a dynamic-game model to show that under intense competition resource redeployability might backfire for corporations due to their lack of credible commitment to a specific market. This implies that multibusiness corporations should be cautious about entering winner-takes-all markets because they might be outcompeted by more committed focused rivals. Finally, Giustiziero et al. (2023) show that resource scalability can lead to the joint phenomenon of “hyperscaling” (firms scaling rapidly) and “hyperspecialization” (firms focusing on just a single activity while scaling), contradicting the established wisdom that firms will need to diversify to grow. To be sure, Strategy Science is not the only journal that will look for managerial relevance in (all) their papers, but it is key that they refer to plausible managerial choices and relevant outcomes, even in a modeling paper.

2.2. Plausible Mechanism

A model is only as good as its assumptions. In any formal analysis, the assumptions will fully determine the results. There is no room to hide behind the fuzzy or imprecise reasoning from which verbal theories sometimes suffer. This is commonly called the “garbage-in garbage-out principle.” Poor assumptions will render the model useless. In our experience there are three key issues the assumptions can suffer from:

1. The assumptions are not constraining enough;

2. The key results are trivially baked into the assumptions; or

3. The key results follow from implausible assumptions.

Although issue 1 is usually relatively straightforward to diagnose, issues 2 and 3 are more subtle. Ultimately any formal reasoning is a tautology (or wrong, which is worse). The question is whether the reasoning to get to the results is enlightening or trivial. Moreover, many (indeed, most) models will have implausible assumptions. In fact, to counter issue 1, you must make constraining assumptions, which might not be realistic.² As the aphorism attributed to George Box goes, “all models are wrong, but some are useful.” The question is whether the key results follow from assumptions that plausibly capture the essence of the problem at hand.

An important way to address these issues is to focus on mechanisms. They should describe how the assumptions lead to the results in a nontrivial way, countering issue 2. Moreover, the mechanism should be plausible. To counter issue 3, ideally, you can illustrate the mechanism and associated outcomes with real-world examples. The main contribution of many models is to introduce a previously overlooked mechanism, leading to results that challenge the conventional wisdom.

For instance, the mechanism that Siggelkow and Rivkin (2006) identify is that a lower-level manager engaging in extensive exploration is very likely to find the best-performing organizational configuration for its own department and recommend that to their superiors. However, interdependencies might cause the optimal solution for this specific manager’s department to be inferior for the organization as a whole. By contrast, a manager engaging in more limited exploration is more likely to sometimes recommend alternatives that, although suboptimal for their own department, might be superior for the entire organization. As a result, extensive low-level exploration can backfire for the organization as a whole. The simulation model shows how this overlooked mechanism upends the conventional wisdom on how organizations should engage in exploration.

The great advantage of models over other methods is that they allow scrutinizing the mechanism to any wanted degree. Unlike with empirical analyses, in a modeling study you can always look deeper into the mechanism and do more calculations, until you understand the mechanisms as deeply as needed. For instance, Siggelkow and Rivkin (2006) provide very detailed and specific simulated examples if low-level managers engaging in exploration. Wibbens et al. (2025) provide detailed examples of the market evolution for the competition between focused and diversified firms, illustrating the interplay between competitive intensity, investment incentives, resource positions, and profits. Both articles use extensive charts to illustrate the uncovered mechanisms.

The ability to extend the analysis to any wanted degree is also a reason why regressions are often considered a poor method to analyze, or verify, models. The primary purpose of regressions is to perform statistical inference using inherently limited data from an unknown data generation process. Because in a model there is no inherent limitation to the data and the data generation process is known, regressions are not needed. Moreover, regressions often simply cannot “test” the validity of models and their results at the required level of detail.³

2.3. Keeping It Simple

Einstein allegedly contended that “everything should be made as simple as possible, but no simpler than that.” This version of Occam’s razor is the key counter to issue 1 identified in the previous section. In a modeling study this means that in the main version of the model you should remove all elements of the model that are not needed to establish the main mechanisms and results of the paper. If useful, you can extend the model later in the paper to explore your results under a wider range of conditions.

Keeping it simple is important for your readers to understand the model; the worst are models with a “letter soup”—overly extensive formulas with so many variables and symbols that they become very hard to grasp. Moreover, the more your hands are tied, the more convincing your results are. Finally, having few parameters allows you to explore the parameter space more comprehensively. As a rule of thumb, models with three moving parts allow for the analysis of the relationship between two parts (e.g., innovative effort and market share) with the third kept constant (e.g., the degree of intellectual property protection). More parameters imply more contingent results (e.g., four parameters require two of them to be held constant), and as the number of combinations grows exponentially with the number of parameters, a comprehensive analysis becomes impossible in more complicated models with many parameters.

2.4. Anchoring in Prior Models

You can be almost sure that an issue similar to the one you are interested in has been studied before. Many brilliant researchers over the past centuries have sought to understand the world around us, so chances are very high that modeling frameworks already exist that are applicable to your research question. You should use them to anchor your model firmly in prior literature. This anchoring based on prior models is important for three reasons.

First, like the “keeping it simple” mantra, it aids your readers. Prior models serve as a solid foundation. Any assumptions and reasoning that have been established before might be already known and need minimal scrutiny.

Second, anchoring helps establish your contribution. Your paper’s “delta” should follow from those elements of the model that you change vis-à-vis prior literature. Ideally, these changes are minimal but critical. That way you can clearly point to what important point prior literature has overlooked and to the mechanism by which it leads to outcomes that challenge the conventional wisdom.

Third, anchoring helps establish a cumulative literature. If each paper tries to reinvent the wheel, science is unlikely to progress. By contrast, a stream of papers that continuously enhances prior models is much more likely to build a useful edifice enhancing our understanding.

Finally, note that the literature (or multiple literatures) in which the model is anchored might be very different from the literature to which the main contribution is made. In fact, some of the most influential models are built on modeling frameworks outside of management. For instance, the seminal Levinthal (1997) paper introduced NK models from evolutionary biology to study organizational adaptation in complex environments.

2.5. Expositional Clarity

As a final point, we want to provide some guidance to improve expositional clarity. A clear paper is more likely to be reviewed positively and to make an impact once published. Reviewers and other readers who do not quickly grasp what you are trying to say are likely to put your paper aside.

First, to improve clarity it is important where to focus your effort in the exposition. Although most of your work has likely been on the inference (step 2 in Table 1), this is usually of least interest to the reader. If you did your inference well, it should logically follow from the assumptions that you made. Therefore, your assumptions and their anchoring in the literature are much more important than the proofs or computational procedures. Of course, they need to be included so they can be scrutinized, but usually an (online) appendix suffices for that. If you use code, it should be included as supporting information so the reader can download it, replicate your results, and extend the model as needed. The main paper only needs a short description of the proofs or computational procedures, pointing to the supporting information for further details.

Second, because the assumptions are so important, we usually find it helpful to include tables with all parameters and variables included in the model, with a short description. This greatly aids the reader in keeping track of all the moving parts throughout the paper.

Third, we (and, more broadly, the typical Strategy Science readership) find graphical representations (charts, graphs) of the paper’s key results very helpful. They are easier to understand and can concisely convey much more information than tables or text. Ideally, all main results and perhaps even the main assumptions should be clear from the figures.⁴ Moreover, figures offer an opportunity to present results across the entire parameter space rather than just a narrow set of parameter values. It greatly helps if they are self-contained: The figures are so clear that their key insights can be understood without reading the main text of the paper. You might need to include notes accompanying your figures to achieve this.

These are just a few short guides. Please take inspiration from other modeling papers about what works well (and what does not) to enhance clarity. Further, Makadok (2022) offers excellent guidance in making modeling papers accessible also to nonmodelers—which is critical to reach a broad audience and corresponding impact. Finally, and perhaps most importantly, present your insights to colleagues outside your discipline. This is the best way to evade the “curse of knowledge”: What is self-obvious to you might be incomprehensible to others. You want to make sure to address these issues before submitting your paper. Reviewers and editors that struggle to comprehend your paper are likely to reject it.

3. Two Special Cases

Although we believe that the principles outlined above apply to most, if not all models, it is important to keep in mind what the purpose of a model in a particular paper is. To illustrate this, we briefly discuss two classes of models that may, on the surface, not be subject to the criteria outlined in the previous section.

3.1. Models in Empirical Papers

First, consider models that are paired with an empirical analysis in the same paper. The purpose of the model is often to provide a framework to guide the empirical analysis.

Such models face a specific tension between keeping it simple and corresponding with the empirics. In fact, simplicity is even more important than in stand-alone modeling papers, because less space is available for the model. At the same time, mechanisms that cannot be tested with the data at hand, for example, because they operate at a lower level of aggregation, may create a disconnect between model and empirics and leave readers unsatisfied. Therefore, the model must be crafted such that it corresponds well with the empirics, while keeping it sufficiently simple that it can be easily understood in a shorter section of the paper. Somewhat analogously, model parameters crucial for the model results but not present in the data create a mismatch because the model results do not correspond to the empirical findings we observe.

Overall, we believe that theoretical models to support and guide empirical analysis will add most value when they uncover nontrivial interactions between two (measurable) variables or over time. Direct effects are often “baked into” the assumptions and as such do not require a formal model to be presented clearly, whereas interaction effects between variables typically need to rely on additional insights about the underlying payoff function and what determines the marginal returns (i.e., the linear effect) of a variable and how these returns might be affected by changes in another variable. For example, a model “showing” that team incentives increase firm performance (a direct effect) might not add much value over and above verbal theorizing, whereas a formal model outlining the market conditions (e.g., the level of local competition) under which incentives have the biggest positive impact on performance (a contingent effect) can helpfully outline the likely underlying forces (Khashabi et al. 2021). Also, models can be helpful in capturing and predicting dynamic effects, for example, how compounding and/or path dependence matter, that can then subsequently be taken to the data.

Hence, even models matched with empirical analysis should be nontrivial, plausible, and as simple as possible. Having said that, an important criterion for such a model is how well it informs and matches the empirical analysis that follows.

3.2. “Realistic” Simulation Models

The second class of models we discuss are models that are intended to be realistic by capturing as many parameters found in the real world as possible. At first glance, this might conflict with the demand of keeping it simple outlined in Section 2. How can a model with dozens of parameters be “simple”? Again, it is important to consider the purpose of the model. “Realistic” simulations try to relax the restriction common to all models that they are an abstraction of reality. In situations in which a model is supposed to mirror reality (i.e., data found in the real world) as closely as possible, the goal is often to derive parameter combinations that allow the modeler to replicate patterns found in the real world. This places high demands on a model in terms of capturing functional forms and allowing for interactions among parameters, not necessarily with the ultimate goal of uncovering mechanisms, but rather to replicate aggregate patterns through a series of lower-level mechanisms.

We might think of such models as the domain of weather forecasts or macroeconomic models, but system dynamics models that include a wide range of parameters and their interactions can also shed light on operational problems like quality maintenance in the service industry (Oliva and Sterman 2001) or the dynamics of demand with social bandwagoning for cultural products (Crossland and Smith 2002). In both cases, a large number of parameters are used to calibrate a system dynamics model to the real data of single firms: a large UK bank in Oliva and Sterman (2001) and an artisanal fine porcelain producer in Crossland and Smith (2002).⁵

Again, the purpose of the model determines the way in which the recommendations outlined in this editorial should be interpreted and implemented. “Simple” in this context means any parameters that might have nontrivial interactions with others, “plausibility” will predominantly refer to the functional form and quantitative effect of certain variables, and “expositional clarity” typically stems from the visual representation of model outcomes rather than formal statements or results tables indicating goodness-of-fit.

3.3. The Five Essentials’ Boundary Conditions

These two model classes also highlight some of the boundary conditions of our recommendations. For a simplified model in an empirical paper, there may not be a clear “model blueprint” (see Section 2.4, anchoring in prior models) because the model is highly context specific, and equally, some of the outcomes of a system dynamics model may be a direct result of the assumptions (see Section 2.2, criticism 2, results baked into the assumptions) simply because the researcher knows this to be an empirical reality. Similarly, obvious plausibility of the mechanism might be less relevant, because presumably it will be tested in the paper. Nevertheless, we strongly believe that readers will feel most inspired by models that have the capacity to surprise them and that do not seem tailormade for just one singular context without any potential for learning something more general, so the principles we outline still hold, but possibly to a different extent and with different emphasis.

4. Conclusion

In this short editorial, we have provided guidance addressing the key issues that we often come across when editing and reviewing modeling papers. We have written what we would have liked to have in our hands when we embarked on writing modeling papers ourselves. From the journal’s inception, Strategy Science has encouraged modeling papers. They align very well with its editorial mission to publish outstanding research that pushes the field forward using a high level of rigor. With this editorial, we wish to facilitate this mission and encourage the submission of high-quality modeling papers. We hope it provides inspiration and guidance to authors aspiring to publish models in Strategy Science. We look forward to receiving your submissions.