On the epistemic contribution of financial models

ABSTRACT

Financial modelling is an essential tool for studying the possibility of financial transactions. This paper argues that financial models are conventional tools widely used in formulating and establishing possibility claims about a prospective investment transaction, from a set of governing possibility assumptions. What is distinctive about financial models is that they articulate how a transaction possibly could occur in a non-actual investment scenario given a limited base of possibility conditions assumed in the model. For this reason, it is argued that the epistemic contribution of financial models is that of enabling the model users to envision exactly how a prospective investment could be achieved in various ways through a detailed understanding of the available transaction mechanisms. Thus, financial models provide information about the possibility of an investment scenario by showing how a specific transaction mechanism could result from a small set of initial possibility conditions assumed in the model.

KEYWORDS:

• Financial
• models
• epistemic
• possibility
• non-actual
• mechanisms

1. Introduction

Modelling is an essential tool for studying financial transactions’ possibility. Financial modellers construct models to learn about the prospects involved in carrying out an investment in a target company or group of companies. The financial modeller tries to solve a problem. Ultimately, similar to an engineer preparing for a project, the modeller asks: specifically, how a prospective investment transaction is possible.

This paper considers methodological and epistemological features that make financial models and modelling practice philosophically interesting. It asks: What type of models are financial models, and what sort of knowledge do financial modelling strategies aim to provide? Briefly, this paper argues that the modelling strategies in question centre on ascertaining what is possible in a non-actual investment scenario, by determining how it is possible, from a set of initial conditions leading to a specific transaction mechanism. More specifically, the modeller seeks to ascertain and determine a number of alternative mechanistic paths to making a prospective investment transaction possible.¹

The remainder of the paper proceeds as follows. The next section presents a common type of modelled transaction to highlight, through examples, important features of financial models in general. Section 3 surveys relevant philosophical literature concerned with scientific models and modelling of non-actual phenomena. Section 4 considers the epistemic import of models and proposes that financial modelling practice resides in formulating and establishing possibility claims from a limited base of epistemically constrained possibility assumptions.

What is distinctive about the work of financial modellers is that they are concerned with the possibility of non-actual phenomena. For this reason, it is proposed, financial models do not offer to describe actual investment transactions. Instead, the claims that these models support do not map on to any actual or actively ongoing investment transaction in the world and, thus, are most suitably characterised as pertaining to demonstration of possible non-actual transaction mechanisms, which enable model users to envision how, exactly, a possible investment in the relevant sense could be obtained in different non-actual scenarios.

2. Financial model-building in practice

Financial modelling, as the previous sketches show, is about solving problems related to a prospective (non-actual) investment transaction.² From this process of solving a more or less commonplace objective—working out an investment’s prospects—special modelling standards and techniques, and even aesthetic preferences, have emerged about what makes a good or useful model not just in practice but also in principle.

This section looks at some regular financial modelling features as they commonly appear in a model of a leveraged buyout, which is a type of investment transaction extensively modelled in the financial industry (Demaria, 2020). In a leveraged buyout, an investment fund³ acquires a controlling stake in a company using mostly debt from a financial institution to finance the transaction.

There are various formulations of such a model, but most normally have a leveraged buyout transaction modelled from nine components: (1) possibility assumptions, (2) income statement, (3) cash flow statement, (4) balance sheet adjustments, (5) depreciation schedule, (6) working capital, (7) balance sheet projections, (8) debt schedule, and (9) investor returns. The following explains the role each segment plays in the overall construction.

2.1. Possibility assumptions

Every financial model, whether of a leveraged buyout or any investment transaction, begins from a series of possibility assumptions. In a leveraged buyout, the assumptions set out the possible purchase price and other model parameters related to a plan targeting a particular company, or group of companies, whose acquisition (by an investment fund) the model concerns.

The modeller starts by obtaining a possible price for the acquisition target. Having arrived at that price, the modeller can begin setting out how the necessary funds would be used in completing the prospective transaction (‘uses of funds’), and how these funds would be sourced (‘sources of funds’). Most of the uses of funds would be for paying for the acquisition, but funds might also be appropriated to cover advisor fees, such as legal fees, incurred on the transaction (fees dedicated to compensating advisors could range from 2%–4% of the total transaction value). Thus, in specifying the sources of funds, the modeller determines what type of funds would be used, and in what proportion, to pay for the potential transaction. The uses of funds would clarify what the sources of funds intend to cover. The combined uses of funds must match the combined sources of funds for the model to be consistent.

Once the prospective uses and sources of funds are specified, the modeller successively makes further assumptions so that he/she can eventually establish a possible rate of return for investors. The most important of these is the assumption concerning the length of the investment period (‘holding period’)—how long the investment fund would own the acquired company. This possibility assumption may vary depending on a transaction’s details, but 5 years would normally be assumed in private equity transactions such as leveraged buyouts (Gompers et al., 2016).

The modeller, having specified a holding period, proceeds to determine the acquired company’s potential price at exit: how much the entity would be sold for at the end of the holding period (upon the investment fund exiting the investment). A common way to express the exit price in a leveraged buyout model is as a multiple of projected earnings before interest, tax, depreciation, and amortisation (EBITDA). For example, the buyer may acquire a target company for 14× EBITDA. What this means is that the eventual buyer would pay 14× EBITDA for the company (i.e. 14× the projected EBITDA, wherein the projection would be made off an adjusted EBITDA baseline figure in the model). The modeller would generally assume the price paid both at entry and exit would be the same multiple of EBITDA.⁴

A crucial point to consider in this context is that the EBITDA that determines the acquisition price would not be the actual EBITDA the target company reports (i.e. the same figure quoted in the company’s annual report). Instead, the figure for the acquisition price multiple would be obtained from an adjusted (non-actual) EBITDA figure constructed in the model. The figure would be adjusted for the purpose of ‘normalising’ the EBITDA; however, differences in modelling methodology and transaction assumptions typically have deviations in adjustments made to the same baseline figure. Any adjusted figure, however, will inevitably diverge from that being adjusted from (that is, by definition).

In an ordinary case, the determined price at exit would not be based on a projection from any reported or unadjusted figure; rather, it would be based on an adjusted EBITDA in the model. The main point here is that the mechanism of the investment transaction in the model ultimately would be based on a possible (non-actual) figure.

After the core possibility assumptions (purchase price, uses and sources of funds, holding period, acquisition multiple, and exit multiple) are in place, the modeller would construct a 5-year projection (assuming a 5-year holding period is epistemically possible) purposing to eventually arrive at an exit price based on an exit multiple at the end of year five. The multiple would later be used to determine the potential rate of return to investors that can be expected or deduced from the modelled transaction.

In sum, the possibility assumptions are an important part of the model, as they set out the initial conditions or parameters for the modelled mechanism.

2.2. Income statement

This segment is for the purpose of constructing an income statement for the acquired entity derived from 3 years of historical financial data (‘company financials’). The modeller could then consider historical trends in formulating the growth rate assumption used in making projections over the subsequent holding period. As an example, the projected revenue growth rate may look as follows: projected growth in year one, 4%; year two, 4.5%; year three, 3.5%; year four, 3%; and year five, 3%. Important to observe, however, is that projections are not made directly from the actual or reported company financials, but rather from ‘adjusted’ (non-actual) income statement financials in the model. On this account, any projection expressing a possibility in a non-actual scenario would itself be based on, or derive from, one or more assumptions of possibility constraining the modelled transaction mechanism.

2.3. Cash flow statement

The general purpose in modelling the cash flow in this component is to track the inflows and outflows of cash, and make adjustments and projections with these in mind. The cash flow statement is aimed at clarifying where cash is being generated (cash inflow) and spent (cash outflow) over a projected holding period. Showing the possible (prospective) flow of funds is considered important in establishing how the company, if acquired under the initial possibility conditions specified in the model, would have sufficient cash on hand to pay expenses, purchase assets, and pay down debt to generate a certain investment return in different (non-actual) scenarios.

An important feature of the cash flow statement is that, unlike a company’s sales or revenue, cash flow could be negative. For example, a company might report a net profit (gain) in the income statement but have negative cash flow. Similarly, the modeller may learn that the positive cash flow figure reported in the target company’s actual cash flow statement is in fact negative once the relevant transaction adjustments are made (i.e. adjustments made based on the initially assumed possibility assumptions). For example, a company with a negative projected cash flow in the model may not be a strong candidate for a leveraged buyout transaction, in which strong cash flow and liquidity are typically important (Lehn & Poulsen, 1989; Pignataro, 2014). Conversely, a highly leveraged company with liquidity issues might be a strong candidate for various distressed debt investment strategies (Demaria, 2020; Whitman & Diz, 2009).

2.4. Balance sheet adjustments

This segment shows how the acquisition target’s balance sheet would appear after the acquisition. Briefly, a balance sheet is a financial statement showing company assets, liabilities, and equity at a given point in time. The modelled pre- and post-acquisition balance sheets would, thus, respectively reflect the entity’s financial position before and after an acquisition.

Note that the company’s total assets would inevitably equal (balance) the company’s total combined equity capital and liabilities in any possible scenario. The difference between the pre- and post-acquisition balance sheet would reside in how liabilities (external claims on company’s assets) and equity capital (internal claims on company assets) are adjusted (composed differently) because of the acquisition.

Adjustments are made to the acquisition target’s original (actual) balance sheet statement to reflect the acquisition’s effects on the company’s assets and liabilities. The balance sheet shows, for instance, how the obligations and liabilities would appear after an acquisition has been completed in accordance with the ‘ground rules’ implicitly laid down in the possibility assumptions governing the transaction mechanism. This information would be considered important in, for instance, demonstrating how the acquired entity would be able to retain necessary funds to pay its current liabilities (debt and obligations owed by the company and due within 1 year) in different scenarios.

2.5. Depreciation schedule

The concept behind the depreciation schedule is to relate the depreciation’s projected impact on the acquired company’s assets. In modelling the depreciation schedule, the modeller would determine the depreciation’s possible impact on planned capital expenditure (improvements to property and equipment) in different non-actual scenarios. Because there are several allowable methods for depreciating a company’s assets, the schedule is constructed to show how different methods of depreciation change or influence other aspects in the model, such as the availability of cash to pay down debt and, resultantly, what returns investors can possibly anticipate.

Again, this is a matter of showing how different assumptions on depreciation variously affect or preclude certain possibilities in the modelled transaction. This segment of the model lets modellers test if the choice of possible depreciation methods leads to a material difference in the overall returns to investors. In comparing two allowable methods, the modeller, technically speaking, creates two models, wherein the method of depreciating assets (with ensuing effects on other aspects in the model) is the only difference.⁵

In practice, a ‘switch’ would be created in the model, which would allow the modeller to toggle between the two depreciation methods in any given scenario. In which regard the modeller may learn, for example, that the tax benefits are greater when choosing one possible depreciation method over another. On the whole, this experiment-like procedure, therefore, leads to a quantifiable difference in possible overall returns. Consequently, for example, if it is possible to choose a method to free up more cash in the short term, the availability of this (excess) sum might open a substantial number of other possibilities in the modelled mechanism.

Note that this demonstration of possibility—and any similar demonstration in the model—would expressly assume that the specific transaction conditions used in a prospective scenario are themselves possible or can be instantiated. There will often be a spread of possibilities presented in a model containing more than one non-actual investment scenario; however, this spread, standardly expressed as the best (management)-case, base-case, and worst-case scenarios, would be entirely determined by switching or changing the applied set of possibility assumptions.

2.6. Working capital

This component is modelled to determine how well the cash generated from the acquired entity’s current assets would be able to cover liabilities due within 12 months of the acquisition, under different possibility conditions. As such, it is for modelling the targets’ near-term liquidity after the acquisition (which, again, would entail projecting the acquired entity’s ability to meet its near-term obligations from the set of governing possibility assumptions regarding the modelled transaction). Thus, by showing how the company could meet its near-term obligations given different sets of assumptions, the working capital segment may provide assurance that the company would, under a varied set of possibility conditions, be able to meet its obligations on time.

In the model, assuming an increased working capital efficiency would normally have a correspondingly positive effect on the company’s profits or EBITDA. Projecting more efficient levels of working capital could, thus, further enhance returns to investors. The positive impact on returns would follow from the exit value of a company being determined from a multiple of the adjusted (non-actual) EBITDA figure, which, at the end of the holding period, would have increased, partly from the possibility assumption that working capital can be made more efficient.

2.7. Balance sheet projections

This segment projects the acquisition target’s balance sheet over the assumed holding period, with the adjusted (non-actual) balance sheet as the starting point (i.e. the modelled balance sheet, not the actual or reported pre-acquisition statement outside the model). The segment would be modelled to determine the projected cash flows’ effects on the company’s liabilities and assets over the holding period specified in the model.

2.8. Debt schedule

In a leveraged buyout, the acquiring entity would use debt financing to satisfy most of the acquisition cost. Normally, multiple debt instruments would be used in a transaction, with a set of conditions attached to each instrument. For instance, one type of debt instrument, commonly denoted as ‘senior,’ may come with a low interest rate, such as 2% (above the London Interbank Offered Rate [LIBOR], which, for example, would be assumed to rise by 0.3% per annum during the investment), and must be fully repaid by the end of the holding period. Other types of debt, denoted as ‘junior’ or ‘subordinate,’ may come with a high interest rate, perhaps up to four or five times the senior rate (but without LIBOR, in which case the rate, for example 10%, would be held constant), and not require that the principal be fully repaid by the end of the holding period.

A debt schedule would be constructed to show how the different financing conditions imposed by the various debt instruments in the prospective transaction could be accommodated. For this purpose, projections in the model would be linked to create a schedule that tracks debt and interest over the assumed holding period. Understanding how much cash could be made available to repay and service debt in different (non-actual) scenarios would be a critical step in the model because the strategy in a leveraged buyout usually involves using excess cash to pay down debt (as one way of enhancing returns to investors). Consequently, for instance, if the total amount of debt that the target company assumes to pay for its acquisition cannot be sustained in any prospective scenario, the possibility of the entire transaction may be called into question.

2.9. Investor returns

This final segment involves calculating investment returns for investors, imputed or surmised, from the projections and assumptions made thus far in the model. In many modelled transactions, investor returns are expressed by an internal rate of return (IRR) metric, and often further supplemented by a money-on-money multiple (M-o-M) metric.

The IRR metric shows the percentage rate of return that investors’ initial investment would be expected to generate over the 5-year holding period, given the model assumptions. Many investors often see the M-o-M metric as more intuitive because it simply adds up all the cash paid out to investors (money gone out of the fund) and divides this amount by the total amount received from investors (money gone into the fund). Despite this viewpoint of it being more intuitive, the metric does not, in contrast to the IRR, consider the length of the holding period (i.e. makes no assumption concerning the time value of money, thus avoiding the use of a discount function). In any event, the specific metric, or set of metrics, used in financial models to express returns could ultimately vary, and the choice in many cases would be determined by the type of transaction modelled.

Generally, then, the overall construction works to determine what returns investors could anticipate from the presumed transaction mechanism; that is, provided the investors are willing to believe or accept a set of possibility conditions or assumptions governing the model. In other words, the modelled transaction could achieve a certain IRR if investors have a justifiable belief that the underlying possibility assumptions and the set-up conditions are themselves possible and capable of being instantiated. For example, they believe the acquired company would be able to increase its sales at a given percentage each year during the holding period (and therefore increase its profits, by further assuming, for instance, the proportion of EBITDA to sales would remain constant). As another example, they believe the acquired company would be able to maintain more efficient levels of working capital (‘optimise cash on hand’) by, for instance, collecting receivables faster, reducing inventory cycles, and extending payment terms. Perhaps even more importantly in the context of a prospective leveraged buyout, investors are willing to accept that it would be possible to sell the acquired company for at least the same multiple as by which it would be acquired, and be able to pay down debt (initially used to fund the transaction) at a certain pace, as the model requires.

3. Models and modelling in the philosophy of science

Contributors to the literature on scientific models have suggested that many models do not purport to describe actual phenomena and therefore apparently lack any epistemic value, whereas others have pointed out that the ability to describe actual phenomena is not a necessary condition for models to have epistemic value (Grüne-Yanoff & Verreault-Julien, 2021). Instead, it has been suggested that many such models’ epistemic contribution lies in their ability to provide knowledge of possibilities (Aydinonat, 2018; Aydinonat & Köksal, 2019; Bokulich, 2014; Brainard, 2020; Forber, 2010; Grüne-Yanoff, 2013; Massimi, 2019; Reiss, 2013; Verreault-Julien, 2019; Weisberg, 2013). This section considers the characteristics of financial models in relation to the philosophical literature concerned with the epistemic value of scientific models and modelling of possible (non-actual) phenomena.⁶

3.1. Models of non-actual phenomena as ‘mere’ heuristics

In the literature, several authors have suggested that how-possibly models serve a purely heuristic rather than an epistemic function (Alexandrova, 2008; Alexandrova & Northcott, 2013) as they apparently lack the ability to describe actual mechanisms.

However, many considerations oppose such a view. To begin with, modellers often seek to learn something about structures that are not actively caused or sustained at present (though they possibly could or will be). One may likewise observe that it is too strong to suggest from a purely descriptive point of view that modellers do not seek to acquire knowledge about possible (non-actual) economic or financial mechanisms.

For example, consider the vast amount of time and effort diligently spent on building extensive mathematical models and operating with different scenarios and experimenting with eventualities if only for the sake of being heuristically useful or generating a hypothesis as such. As Reiss (2013) points out in a reply to Alexandrova and Northcott (2013): ‘Why do economists build complex, mathematically sophisticated models rather than, say, resort to creativity and intuition, crystal balls, hypothesis-generating algorithms or consciousness-enhancing drugs? All of these sources of inspiration would be a lot easier to come by, and some of these would be more fun, than doing the hard work of constructing and solving a model’ (Reiss, 2013, p. 282). The heuristic characterisation creates a still further problem according to Reiss in that it leaves out a pivotal aspect of economic model-building, namely, the epistemic advantage that models are justifiably expected to procure: ‘To warrant their existence, models must do more than to provide hypotheses. They must have some genuine epistemic benefit’ (p. 282).

Others have similarly argued that the heuristic justification of scientific models is weak. For example, Grüne-Yanoff (2013) notes that the heuristic justification ‘places the use of such models in the same category as taking a walk, reading the newspaper, or whatever scientists do in order to inspire themselves to further theory development’ (p. 851). In a similar manner, Gelfert (2019) points out that exploratory models do not merely serve the heuristic function of stimulating further research but also ‘explicitly aim at identifying how-possibly explanations or otherwise delineate the space of possibilities’ (p. 15). Thus, he goes on to argue, considerations about possibilities play an important role in the ‘proof of principle’ (p. 16) that exploratory models provide.

It is important to note in this context that a ‘proof of principle’ goes beyond the mere postulating of a hypothesis; it must articulate how a prospective phenomenon could possibly be brought about, for example, by specifying the relevant (non-actual) mechanism through which the phenomenon could operate. In this regard a financial model of an investment transaction needs to accomplish more than simply proposing a hypothesis (i.e. suggesting that an investment transaction might possibly come to pass). The model must provide knowledge about how the non-actual situation could transpire by showing how the transaction mechanisms explored in the model could lead to a relevant set of possibilities.

3.2. An example of financial models’ ability to determine the possibility of a non-actual investment transaction

As shown in Section 2, financial models are not typically constructed to have a heuristic function only, but purposely made in an attempt to learn something about a prospective investment transaction. To further illustrate how this could work in practice, consider a financial model of a merger scenario seeking to demonstrably provide evidence for the possibility of a prospective (non-actual) investment mechanism.

In this non-actual situation, a merger is taking place between two previously separate entities ‘Amazon’ and ‘Netflix’, each owning a substantial share of the global video streaming market. In such an admittedly non-actual scenario, a model of a possible transaction mechanism would seek to determine how certain components of ‘Amazon’ and ‘Netflix’ would function in a newly created entity, denoted as ‘Amaflix’.

To provide such information, the modeller needs to demonstrate how this particular merger would take place by way of a number of different (non-actual) transaction mechanisms, all assumingly distinct from the actual.⁷ At this stage, the modeller has a number of potential mechanisms to consider, for instance, what would happen if Amaflix eschewed most and contained only a few of the components or organisation of the original entities; how could this be achieved? What if Amazon and Netflix were first made private (de-listed from the stock market) and consequently merged; how could this affect the prospective enterprise, and in what way would it be possible for a merger to take place while both entities were still publicly listed? These are just a few examples of the problems that the modeller may need to solve.

In any event, the modeller will seek information relating to what the possible (non-actual) constitution and behaviour would look like in a newly merged entity. Above all, the modeller seeks to acquire information regarding a possible (non-actual) investment mechanism. For example, he or she may ask what if only a specific segment of Amazon’s operations were divested (sold off), such as the company’s online streaming component, leaving the rest of its operations intact. In such a scenario, only this segment would be combined with Netflix, which opens new possibilities regarding the remaining segments. The modeller can now try to find a solution: what if at the same time Amazon’s remaining operations (i.e. excluding the streaming segment) were divested in the same investment transaction. All these requests pertain to the acquisition of knowledge regarding the possible operations that various components in different non-actual setups could perform.

As such, if the modeller is justified in believing that certain initial conditions are (epistemically) possible, and then proceeds to determine what else is possible in a non-actual situation given these possibility conditions, the amount of information that is potentially gained could be impressive. For instance, the information potentially gained in modelling the initial steps in a transaction could be used to fund a jarringly orchestrated symphony of acquisitions, taking its course with ‘Disney’ and that company’s video streaming segment (‘Disney+’).⁸

At this point, the modeller can consider the possibility of repeating the previous strategy with Amazon. He or she could consider, for instance, how selling off Disney’s non-streaming related operations, such as the company’s customer product and theme park segments, could be used (again, assuming this would be epistemically possible) to fund either a part of the original costs associated with the de-listing and merging of Amazon and Netflix or to fund additional acquisitions. In any event, the newly consolidated investment mechanism (‘Amaflix+’) would still result from the amalgamated operations and components from at least three distinct entities, all previously large video streaming competitors battling for market share.

At this stage, the modeller could consider the prospects of Amaflix + in the absence of any major competition. For instance, what would Amaflix+’s operating margins look like without the need to divert funds to protect or expand its market share? In the absence of considerable competition, would Amaflix+’s services need to be as competitively priced as Amazon, Netflix, or Disney + were prior to the merger. To give a final example, what if the prospective cash flow (in what appears to become an ever more slightly complex merger transaction) from the newly consolidated Amaflix + were to be used to purchase any remaining large streaming service providers to create an even less competitive environment (thus in effect potentially bringing about an Amaflix + streaming service monopoly). These are only a fraction of the type of questions that modellers of actual phenomena can neither ponder nor pretend to answer.

The process of modelling Amaflix + in effect entails concurrently mapping out and determining the non-actual behaviour of a newly minted (legal, financial, causal) entity as opposed to simply bundling together the combined operations of the three previous entities (i.e. knowing how Amazon, Netflix and Disney individually operate does not automatically translate to knowledge about how Amaflix + would operate because Amaflix+’s prospective operations could be established through a variety of different mechanisms that work only as a result of an entirely new set of circumstances, such as the absence of any actual competition).

The admittedly colourful Amaflix + example may seem somewhat forced or artificial. Nevertheless, in modelling financial transactions involving mergers and acquisitions (and more specifically in the case of ‘Amaflix+’ in models pertaining to market-consolidation strategies), we find many ‘real-life’ examples with similar structures that arise naturally in a financial modelling context. In any case, models of possible non-actual phenomena appear to be financial modellers’ preferred method, not because they demonstrate the possibility of actual transactions, but because they enable adequate demonstration of prospective (non-actual or currently inactive) investment mechanisms, which models of actual phenomena cannot possibly do.

Certainly, the financial modeller may learn that some of the investigated transaction mechanisms, especially in seemingly complex transactions such as the Amaflix + merger, would turn out to be impossible in all or most attempts to model. Thus, the modeller may not find a modelling solution to the problem of obtaining a possible investment transaction, but would nevertheless gain valuable insight into which component parts make the transaction impossible. He or she may then rely on this information in modelling related scenarios (perhaps the impossible components can be omitted or made possible in the context of a different mechanism). At any rate, the arguments and examples provided in this section will hopefully clarify the sheer breadth and depth of relevant detail that could be obtained from the modelling of non-actual transaction mechanisms.

4. Financial models contribute to knowledge about possible non-actual transaction mechanisms

Financial modellers, like engineers tasked with the construction of a bridge, are typically prompted to establish the possibility of a particular design for their project. Establishing the possibility of a non-actual mechanism in the case of a financial model is a standard prerequisite for attempting an investment transaction. What this section will argue is that financial models contribute epistemically to knowledge of financial transactions by establishing how exactly a specific investment could be achieved from a limited base of initial possibility conditions assumed at the outset in the model.

As shown earlier, financial models proceed from a set of epistemically constrained possibility assumptions about the situation presented within the model. Relatedly, we noted that modellers do not always seek to convey information about the possible causes that sustain actual phenomena. Rather, many model demonstrations proceed by showing how a phenomenon is possibly caused in a non-actual situation. At the very least, this characterisation would seem to be a descriptively more correct take on financial models, which, in general, examine ways in which modelled phenomena could be financially or economically different, as opposed to trying to model actual circumstances underlying actual phenomena—the latter being a standard textbook view and default position on the epistemic contribution of financial models (to be discussed below). Finally, modellers also appear to consider the epistemic community’s perspective of what the physically or economically possible conditions are in a situation (cf. Grüne-Yanoff & Verreault-Julien, 2021; see also, Sjölin Wirling & Grüne-Yanoff, 2021). Likewise, this appertains to the justifying of any possibility conditions, from which financial modellers would carefully proceed to demonstrate how exactly initial possibilities through a mechanism could lead to a specific change within a particular investment transaction, such as a certain IRR.

Taking the above considerations together, we find that the epistemic contribution of financial modelling is well captured by models’ ability to provide information regarding the possible non-actual components of a transaction mechanism based on an initial set of epistemically constrained possibility conditions. Most regularly, a financial model will attempt to demonstrate how a certain investment transaction and rate of return are obtainable and, therefore, possible (or, conversely, if clearly unobtainable given certain assumptions, impossible in a non-actual situation). As mentioned, a modeller may, for instance, attempt to determine how to progressively consolidate a market or successfully turn around a financially distressed company by determining under which, if any, possibility conditions or circumstances such a non-actual scenario could be achieved.

The definition in Oxford University Press’ A Dictionary of Finance and Banking, however, indicates financial modelling is primarily aimed at simulating actual circumstances:

Financial modelling [is] the construction and use of planning and decision models based on financial data to simulate actual circumstances in order to facilitate decision making within an organization (Law, 2014, emphasis added).

The above reference in the dictionary definition to the goal of simulating ‘actual circumstances’ is clearly at odds, or incompatible, with the present analysis. This is because the characterisation exceedingly implies that models are constructed and aim at some level of descriptive accuracy in depicting actual circumstances pertaining to the actual mechanism of a phenomenon. Similarly, a textbook on financial modelling states that financial models ‘are at best a simplistic representation of how the firm actually creates value for its shareholders’ (DePamphilis, 2015, p. 317, emphasis added).

As already suggested, however, it appears that the more fundamental insight that modellers attempt to provide when purposefully examining almost any kind of investment transaction, is to carefully articulate how exactly that transaction mechanism is possible in one or more non-actual scenarios. In other words, the crucial feature relevant to epistemic import is that such models do not propose to identify how a company is actually operating or how (in retrospect) a transaction was actually performed. Rather, they seek to determine how an investment mechanism would presumably operate in the event of a scenario wherein certain assumptions and conditions could be instantiated, as shown in a model of a non-actual investment transaction. This view is arguably closer to the described modelling practice in that it allows modellers to anticipate possible investment returns in a prospective transaction scenario from a limited base of possibility conditions assumed in the model.

Moreover, it is often the case that investors believe a certain transaction to be possible yet are unable to conceive exactly how (even if they already believe a transaction mechanism could be obtained in a scenario, model users may still fail to envision how it could be obtained). Thus, the model not only gives assurance to individual sceptical investors that a transaction mechanism is indeed possible given a set of conditions, but also articulates precisely how it is possible across multiple scenarios or conditions.

Accordingly, modellers may attempt to show how an investment in a seemingly nonintuitive investment strategy can be possible given certain possibility assumptions even in scenarios that would be considered as extraordinary. For example, ‘Sula Vineyards (A): Indian Wine - Ce n’est pas possible!’ provides a case study of modelling challenges related to demonstrating to investors how an investment in an Indian winemaker was possible (Zeisberger et al., 2017). To this end, the model probed assumptions on how fast the Indian wine market would grow, and on the acquired company’s ability to keep or grow its current market share across a range of non-actual circumstances.

The model further addressed questions of how the Indian winemaker would remain competitive and retain its position in different, more or less extreme, scenarios. In one such (non-actual) scenario, for instance, the Indian government introduced reduced tariffs on wine imports and a rapidly growing market attracted increased competition from producers in established wine-producing countries. In general, then, financial model information obtained through comparing different non-actual scenarios will convey some form of insight about financial possibility regarding a particular investment transaction.

By and large, a central aim of financial modelling—whether considering a prospective merger, leveraged buyout or any other type of non-actual investment transaction—can be articulated as exhibiting different possible mechanistic paths to obtaining a transaction. To achieve this aim, the model first needs to address the transaction itself—what needs to be (epistemically) possible for the transaction to occur. The bulk of the model-building exercise, however, is concerned with establishing how it is possible for the investment transaction to transpire in a non-actual scenario from an initial set of possibility conditions imposed at set-up.

As most of these conditions would contain implicit or explicit reference to possibility claims, the demonstration proceeding from such claims would attempt to convey relevant insight about scenarios that have not been instantiated but are possible. In another sense, to relate possibilities about non-actual phenomena, such demonstration must correspond with the possibility assumptions constraining or regulating the initial conditions from which a transaction mechanism may be modelled. Thus, models of prospective transactions demonstrate how to progress from specific (epistemically constrained) possibility conditions at set-up to a possible (non-actual) investment mechanism, which may then be further constrained, for example, by an IRR above a specific ‘hurdle rate’.⁹

Conclusion

This paper aimed to characterise financial modelling practice and, based on this, consider (1) what type of models they are and (2) what their purported epistemic value and contribution to knowledge are. The consideration of what sort of contribution to knowledge or understanding financial models provided was informed by recent developments and discussions in the literature on scientific models and modelling.

Regarding modelling transactions, this paper showed modellers will seek to explore and understand the possibilities involved in carrying out an investment in a target company or group of companies. Financial modellers try to solve a problem. Ultimately, they specifically ask how exactly a (non-actual) investment transaction is possible and what exactly the model needs to assume for the transaction mechanisms to be possible. As shown, the possibility assumptions provide an important consideration in epistemically constraining what is achievable in any given investment scenario.

Considerations about possible non-actual mechanisms, therefore, appear to play a considerable role in conveying different possibilities in a financial model of a prospective investment transaction. As shown in the described modelling practice, the goal in modelling different non-actual components of prospective investment mechanisms entails specifying, delineating, and determining what is necessary, possible, or impossible given a set of possibility conditions. If an assumption in the model expresses a necessary condition, such as the amount of funds required to pay for the transaction, it effectively sets a limit on what is assumable in subsequent transaction assumptions. If a further assumption suggests it is possible to fund the acquisition with any combination of debt and equity, then different combinations of debt and equity will be possible if the necessary condition is fulfilled. That is, different proportions of debt and equity can be used if the total amount of debt and equity match the total amount of funds necessary to complete the transaction.

It was further suggested that the epistemic contribution of financial models appears to be the opportunity that they provide to visualise at length (or grasp) precisely how a non-actual transaction could be obtained. Often, investors may already be convinced on a point—for example, that a transaction could be obtained—without quite being able to establish how exactly the pertinent investment mechanism from a specified set of conditions could be achieved. Yet without a financial model for the possibility of a prospective investment mechanism, model users would be unable to perceive how the possibility would play out from the initial set-up conditions imposed on the model. In other words, the information about possible transaction mechanisms potentially gained from a financial model would allow users to perceive more fully the investment situation they perhaps already perfectly knew or suspected to be possible.

For example, model users or investors may know that buying a company for an assumed acquisition multiple would be possible (the same applies to other assumed transaction possibilities in the initial set-up, such as the amount of debt taken on that would need to be repaid at exiting the investment). However, investors may not understand (just from their knowledge of these possibilities) how the initial possibilities, such as the possibility of acquiring a target company at a certain multiple of EBITDA, could practicably be fashioned into a specific investment return produced by a specific investment mechanism. Above all else, knowledge that this specific return could be achieved would require a detailed demonstration of how exactly the prospective investment transaction could be enacted across several possible (non-actual) scenarios in the model.

Accordingly, this paper showed that models are chiefly tools for thinking about prospective transaction mechanisms, establishing how something is possible in a non-actual investment situation (which, ultimately, guide real-world financial decisions). This, hopefully, provides a compelling starting point for further work on the epistemic significance of financial models and modelling practice. Of particular note is that more methodological and philosophical attention is required to better understand how knowledge claims provided in the form of initial possibility assumptions (the set-up conditions of non-actual investment mechanisms) should be evaluated, appraised, and justified.

Acknowledgements

The author would like to thank Till Grüne-Yanoff for his many excellent suggestions for improvement in preparing the manuscript for submission. Thanks also go to the anonymous reviewers who read through all the changes at each stage of revision, offering invariably intelligent and helpful advice on all aspects of the paper. Further corrections were suggested by Eva-Charlotta Mebius, and the author is grateful for these.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Open access funding provided by Royal Institute of Technology.

Notes on contributors

Alexander Mebius

Alexander Mebius’s scholarly work is focused on questions traditionally raised in the philosophy of science but pertaining to less explored topics in highly specialised and rapidly expanding areas within biology, epidemiology, economics, finance, and medicine. His most recent academic appointments were as a postdoctoral fellow at the University of Oxford and, prior to that, in an equivalent position at the Ludwig-Maximilian University in Munich.

Notes

1 Note that financial modellers are rarely decision makers. The decisions made based on possibility claims in these models, however, have a profound overall impact on the economy at large. The total estimated value of investment transactions, regularly decided on such models, exceeded US$3.6 trillion in 2020 (estimate derived from Aliaj et al., 2020).

2 This paper assumes that the financial models presented in this section are almost always models of prospective (non-actual) investment transactions. As such, any perspective construing the use of such models as not generally being intended to demonstrate the possibility of prospective transactions will fall outside the scope of this paper. In addition, it is worth noting that the present treatment is specifically tied to modelling practice in private equity (and mergers and acquisitions in general). The analysis may not apply to other areas of finance, for instance, where models are primarily used to price derivatives and measure risk.

3 In the context of a leveraged buyout, the fund would be held by a private equity firm. This is an investment company that manages one or several such funds (pools of capital) for a group of investors (who act as limited partners in these funds and can be anything from a pension fund or university endowment down to a private individual), primarily investing in private, as opposed to publicly held, companies.

4 Consider, for example, a company sold for the very same multiple at which it was originally acquired: if the EBITDA figure had increased by the end of the holding period, the price paid at exit would be higher than at entry, but the multiple would remain the same. Accordingly, the cause of the higher price at exit would be ascribed to the EBITDA figure increase, as opposed to a higher multiple of EBITDA paid at exit.

5 In this respect, the work of the financial modeller is similar to that of the economist in that both repeatedly ‘try to make sense of the world with multiple models’ (Aydinonat & Köksal, 2019, p. 900; cf. Ylikoski & Aydinonat, 2014).

6 Consequently, performativity and strong social constructivist accounts of the epistemic function of financial models, or the purported lack thereof, will not be discussed in this paper. Readers interested in performativity discussions of modelling are advised to consult MacKenzie (2006), who presents one of the more influential performativity accounts related to modelling. See also Mäki’s (2013) compelling critique and cf. Peled (2020) for a response. Brisset (2018) and Walter (2016) offer more general accounts that, like MacKenzie (2006), are partially predicated on the theory of speech acts, following Austin (1975), but are nevertheless distinct from the ‘performativist’ sociology criticised by Mäki. As an anonymous reviewer has pointed out, it seems reasonable that certain procedures in the described modelling practice could be analysed as speech acts. For instance, an extension of this kind might shed some light on the practical significance of illocutionary acts (‘making promises’) that modellers conventionally rely upon in establishing and justifying certain initial conditions assumed in a prospective transaction.

7 Technically speaking, an actual merger between ‘Amazon’ and ‘Netflix’ has never taken place, so a request for information regarding an actual merger would clearly be out of place, but we can safely ignore this for the purpose of illustration.

8 Needless to say, this presents a highly optimistic view of the ability of models to provide knowledge about possible transaction mechanisms. At the same time, it is important to point out that these models are built on the presumption that the initial set of governing assumptions are genuine (epistemic) possibilities. As such, an entire model can collapse if only a single assumption turns out to be unjustified.

9 The hurdle rate, or the minimum rate of return on an investment, would allow model users to qualify a set of transaction scenarios that can be pursued that would only include those above the set hurdle as alternatives to decide from in pursuing an investment (where, in general, the resulting decision might come down to the highest possible IRR and a low perceived risk).