Criteria for Good Explanation in the
Positive Science of Economics
Bachelor Thesis
Joshua George Rook
10368744
Supervisor: Professor Marcel J. Boumans
University of Amsterdam, Faculty of Economics and Business
BSc Economics and Business, Economics Track
Statement of Originality:
This document is written by Joshua George Rook who declares to take full
responsibility for the contents of this document.
I declare that the text and the work presented in this document is original
and that no sources other than those mentioned in the text and its
Abstract
This paper is focused on identifying criteria that will provide an account for good explanation in the positive science of economics. There has been considerable research within the field of economic methodology with regard to what economists do and into how they should practice. This paper aims to bridge the gap between the topic of what
constitutes good explanation in science and how that can be applied in economics. The paper is in the form of a literature review regarding models of scientific explanation followed by an discussion regarding the applicability of these models to economics. It is found that, although philosophers of science are yet to prescribe a definitive account of good scientific explanation, the flaws in previous attempts to define good scientific
explanation illustrate key criteria that an economic explanation needs to meet in order to be considered good.
1 Introduction
Milton Friedman in his renowned 1953 essay ‘The Methodology of Positive Economics’ stated that:
‘The ultimate goal of a positive science is the development of a “theory” or “hypothesis” that yields valid and meaningful (i.e., not truistic) predictions about phenomena not yet observed.’ (Friedman, 1953, p. 148).
Whilst this may be the ultimate goal of a positive science, the role of explanation in science is a pre-empting factor to it. It is not possible to make non-lucky, valid and meaningful predictions about phenomena not yet observed unless phenomena belonging to the same class have previously been observed and explained to at least some extent. Explanations, being answers to why questions, are distinct from
descriptions, which are answers to what questions (Boumans & Davis, 2010, p. 3). It is possible to make predictions based upon a solely descriptive merit (Boumans & Davis, 2010, p. 54). However, description is a part of explanation. If I were to ask someone to ‘Explain this’, the response would invariably be ‘Explain what?’. The majority of
economists (other than followers of Professor Samuelson’s Descriptivism) aim to explain economic phenomena. The motives for providing explanations about such phenomena range from being useful to those who wish to provide a platform for prediction, and to subsequently use predictions to construct and impose policy
accordingly, to purely expanding human knowledge. Yet, there remains disagreement about the qualitative issue of what is, or more appropriately what should be, considered to be a good explanation in economics. The aim of this paper, then, is to provide
clarification on the matter of what it is that constitutes good explanations in economics. In addressing the central issue of the paper, it is necessary to address several sub-issues. These sub-issues are: what is considered to be an explanation in science?; what differences, if any, exist between phenomena studied in economics and those in other sciences?; determining these differences will aid, in part, in evaluating the credibility of applying models of scientific explanation to economics; finally, once this credibility has been established and being based upon models of scientific explanation, criteria should be given as to what constitutes a good explanation in economics.
In determining what it is that constitutes good economic explanations it is possible to approach the subject matter in two distinct manners. One is to descriptively discuss what, and how, economists explain and to then derive criteria that seemingly demarcate good explanations from bad ones. As economists and econometricians utilise statistical modelling for the purpose of explanation and prediction, these criteria could be in relation to how well an economic model provides foundation for prediction, or perhaps how well a statistical model accounts for the occurrence of a specific
phenomenon at a specific space, at a specific time according to some statistical criteria, or otherwise. The second approach is to provide an account of (good) explanation independent of what, and how, economists practice and subsequently apply this to the field. This second method is less susceptible to being subjective to what, and how, economists practice. Economists may practice their science with a fundamentally flawed view of explanation, so basing criteria of good explanation solely upon a
descriptively accurate account of economic practice would incorporate these flaws into an account of explanation. However, it is essential that, whatever the account of
explanation that arises from this method is, it must be applicable to the field and, therefore, some consideration as to the nature of economic phenomena must be made. Therefore, the descriptive adequacy of an account of explanation in economics is only of secondary importance. It is of far greater worth to focus on the epistemic and empirical adequacy of an account of explanation. For an account of explanation to be epistemically adequate, those things that it identifies as being an explanation should, in fact, be
2008). In order for an account of explanation to be empirically adequate, it should not identify anything as an explanation unless it is based on sufficient evidence (Reiss, 2008). Of course, should a model of scientific explanation first meet the criteria of being epistemically and empirically adequate, the level of descriptive adequacy to the field of economics will only increase its value.
This paper will be of interest to those practicing in the field of economic methodology, as well as philosophers of science in general, and should be viewed as part of a prescriptive methodology for economic practice. The research will be in the form of a literature review working from sources over the past several decades within the fields of philosophy of science and economic methodology to come to a conclusion about the role of explanation in economics and the criteria that an explanation needs to meet in order to be considered good. The structure of the paper will address the
aforementioned sub-issues: Firstly, a literature review is provided regarding the progression of models of scientific explanation from the field of the philosophy of science; Secondly, the applicability of these models to economics will be evaluated; Thirdly, criteria will be given as to what constitutes good economic explanation; Fourthly, a brief comment on prescriptive methodology has been provided.
2 The Progression (and Problems) of Models of Scientific Explanation
This section shall outline the progression of models of explanation in science; so as to illuminate factors that are necessary for scientific explanation, as well as to illustrate certain flaws in models of explanation, which may need to be overcome for defining explanation in general science and, as will later be argued, in economics.
The Deductive-Nomological Model of Scientific Explanation
The Deductive-Nomological (D-N) model of scientific explanation was formatted by Carl Hempel and Paul Oppenheim in their 1948 paper ‘Studies in the Logic of Explanation’. This account of scientific explanation states that the phenomena which are to be
explained, the explanandum, must be logically deducible from the class of true sentences that are deemed to account for the phenomena, the explanans (Hempel & Oppenheim, 1948). The explanans is subsequently divided into two sub-classes of sentences; namely, those that are antecedent conditions, and those that are representative of general laws (Hempel & Oppenheim, 1948). It is not a necessary condition to include antecedent
conditions in the explanans, in order for the explanatory argument to be considered sound in all circumstances, however, it is necessary to include, at least, one universal law (Hempel & Oppenheim, 1948). Finally, the explanans should contain empirical content. The model can be summarised as follows: 1
𝐶𝐶1, 𝐶𝐶2, 𝐶𝐶3, … , 𝐶𝐶𝑘𝑘 Initial conditions
𝐿𝐿1, 𝐿𝐿2, 𝐿𝐿3, … , 𝐿𝐿𝑘𝑘 Universal Laws
__________________
𝐸𝐸 Explanandum
An explanation, according to this model, is to be made after an event, 𝐸𝐸, has occurred (Boumans & Davis, 2010, p. 17). Explanation is to find a universal law and a set of initial conditions that will logically imply the outcome event. Prediction on the other hand is to have a set of initial conditions and a universal law to form an idea of an event that should occur according to these premises (Boumans & Davis, 2010, p. 17).
There are several complaints about this model of explanation. For example, the D-N model does not account for the existence of explanatory asymmetries. Sylvain Bromberger’s famous example of explaining an objects height and the length of its shadow brings this problem to light (Bromberger, 1966, pp. 92-93). The answer to ‘why is the shadow of this flagpole a certain length?’ can be answered by using the laws of electromagnetism, the angle of elevation of the sun, and the height of the flag pole to reduce the problem to a simple matter of trigonometry (Salmon, 1989, p. 47). This is explanatory according to the D-N model. When asked ‘why is the flagpole a certain height?’ one can also deduce the height of the pole by using the laws of
electromagnetism, the angle of elevation of the sun, and the length of the shadow. This also conforms to the D-N model, but few would claim that it is explanatory to say that the flagpole has its height because of the length of the shadow (Salmon, 1989, p. 47).
A second complaint regarding the D-N model is that it allows for explanatory irrelevancies; a well-known example of this kind being the case in which there exists a law that ‘all males who regularly take birth control pills will fail to become pregnant’, in conjunction with the antecedent condition that ‘John Jones is a male who regularly takes birth control pills’ (Van Fraassen, 1980, p. 106). The explanandum containing the
1 Credit for the visual summary of the Deductive-Nomological model is due to Boumans & Davis 2010
Explanans
sentence ‘John Jones does not become pregnant’ is deemed to have been explained according to the D-N model. However, to most, this should seem to be a rather
unsatisfactory explanation of why John Jones does not become pregnant (Van Fraassen, 1980, p. 106).
In both of these examples, it is the failure of the D-N model to identify the causal factors and the relationship between that which explains and that which is to be
explained. It is neither the shadow that causes the flagpoles height, nor is it the birth control pills that cause the prevention of John becoming pregnant. It seems, then, that an explanation must properly identify the cause(s) of a phenomenon and, possibly, detail the causal mechanism between the cause and effect.
These examples highlight that an explanation must contain some account of causal relevance. To have some conditions, an effect, and some law-like relationship between them is insufficient for the truth of causality, or the truth of having causal relevance, between the conditions and the effect. The D-N model of explanation needs to have some other semantic requirements that will account for causal relevance, however the various proposals of such requirements have, thus far, been unfruitful (Boumans and Davis, economic methodology p.17). The non-menial task for the philosopher of science, then, is to provide an account of causal relevance, which is based upon an account of causation. The D-N model could then be modified to account for causal effect.
An assumption of the Deductive-Nomological model is that phenomena are caused in a determined way. Many cases of explanation in economics, however, assume that phenomena are caused in a probabilistic manner (Hausman, Explanation and Diagnosis in Economics, 2001, p. 314). This could be a problem for applying the D-N model to economics. This assumption is problematic according to an individual’s metaphysical point of view (i.e. the non-determinist) (Van Fraassen, 1980, p. 105). The determinist can argue that the D-N model does not account for cases that appear ‘probabilistic’ simply because probabilistic cases do not exist and that an insufficient number of antecedent conditions have been included in the explanation. Admittedly, this does require the enquirer to have a large background knowledge of the subject matter under study, but would resolve this particular issue of applying the D-N model to the field of economics i.e. denying the existence of probabilistic causality.
Statistical Relevance and the Probabilistic-Causal Model of Scientific Explanation
Rather than to explain a phenomenon according to universal laws and certain boundary conditions, the Probabilistic-Causal model posits that explanation of phenomena
necessitates including information on their causes (Reiss, 2008). A phenomenon’s causes are probabilistic causes in the sense that there is no universal regularity between the cause and effect. These causes are identified according to the notion of statistical
relevance, which is defined as follows:
In determining the probability that 𝐸𝐸 has attribute 𝐵𝐵, where 𝐸𝐸 is an event and 𝐸𝐸 is ascribed to, and is a member of, a reference class 𝐴𝐴, a partitioning property or
characteristic 𝐶𝐶 of the reference class 𝐴𝐴 is statistically relevant to 𝐸𝐸 having attribute 𝐵𝐵 within the class of 𝐴𝐴 iff 𝑃𝑃(𝐴𝐴. 𝐶𝐶, 𝐵𝐵) ≠ 𝑃𝑃(𝐴𝐴, 𝐵𝐵) (Salmon, 1971, p. 42).
According to the Probabilistic-Causal model of explanation, an event 𝐶𝐶 is deemed to explain another event 𝐸𝐸 iff it is statistically relevant to 𝐸𝐸, and the probability of 𝐸𝐸 occurring in a population described by 𝑉𝑉 is different when 𝐶𝐶 is present than from when it is not (Reiss, 2008). This can be formally stated in terms of conditional probability:
𝑃𝑃(𝐸𝐸|𝐶𝐶, 𝑉𝑉) ≠ 𝑃𝑃(𝐸𝐸|~𝐶𝐶, 𝑉𝑉) = 𝑃𝑃(𝐸𝐸|𝑉𝑉)
So, the idea of statistical relevance improves upon the Deductive-Nomological model in the sense that it can identify certain causally irrelevant factors, such as the factor of taking birth control pills in explaining why a male fails to become pregnant:
𝑃𝑃(𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑃𝑃𝑃𝑃𝐵𝐵𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃|𝑃𝑃𝐵𝐵𝐵𝐵𝑃𝑃𝑃𝑃𝐵𝐵. 𝑀𝑀𝑃𝑃𝑃𝑃𝐵𝐵)
= 𝑃𝑃(𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵𝐵 𝑃𝑃𝑃𝑃𝐵𝐵𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃𝑃|𝑃𝑃𝐵𝐵𝐵𝐵𝑃𝑃𝑃𝑃𝐵𝐵. 𝑀𝑀𝑃𝑃𝑃𝑃𝐵𝐵. 𝑇𝑇𝑃𝑃𝑇𝑇𝐵𝐵𝐵𝐵 𝐵𝐵𝐵𝐵𝑃𝑃𝑃𝑃ℎ 𝐶𝐶𝐵𝐵𝑃𝑃𝑃𝑃𝑃𝑃𝐵𝐵𝑃𝑃 𝑃𝑃𝐵𝐵𝑃𝑃𝑃𝑃𝐵𝐵) = 0
A weakness in relying solely on statistical relevance is that there remains cases where one event can be statistically relevant to another but does not consist of any
explanatory relevance to the phenomenon to be explained because it is not causally relevant. For example, a properly functioning barometer will drop when there is a fall in
atmospheric pressure. Storms, given other certain preconditions, will also occur when there is a fall in atmospheric pressure. The fall in the level of the barometer is
statistically relevant to the occurrence of a storm, though does not explain the
occurrence of the storm, it is the fall in atmospheric pressure that explains both (Jeffrey, 1969). This is the problem of a spurious correlation, and so statistical relevance does not necessarily provide explanatorily relevant factors in explanation. Where 𝑆𝑆 is the event of a storm, and 𝐵𝐵 is the event of a fall in the level of mercury in a barometer:
𝑃𝑃(𝑆𝑆|𝐵𝐵) > 𝑃𝑃(𝑆𝑆|~𝐵𝐵)
The undesirable aspect of statistical relevance, being susceptible to the problem of spurious correlation, can be corrected by means of screening off (Reichenbach, 1956, p. 189). However, this then requires a large amount of background knowledge regarding the overriding causes of a phenomenon.
The Probabilistic-Causal model also improves upon the Deductive-Nomological model as it allows for a causal factor, as identified by statistical relevance, to be deemed as explanatory simply by having an influence on the probability of the effect event occurring, without requiring the effect to occur without exception. That is to say it takes into account that there will often be occurrences of imperfect regularities. Although, as previously noted, a defender of the D-N model has the possibility of disregarding the problem of imperfect regularities as it may be resolved by increasing the number of initial conditions to correct for the insufficiency.
The Probabilistic-Causal model does not, however, deal with the problem of asymmetry. According to conditional probability both the (presumed) cause and the (presumed) effect are statistically relevant to each other:
𝑃𝑃(𝐸𝐸|𝐶𝐶) > 𝑃𝑃(𝐸𝐸| ∼ 𝐶𝐶) 𝐵𝐵𝑖𝑖𝑖𝑖 𝑃𝑃(𝐶𝐶|𝐸𝐸) > 𝑃𝑃(𝐶𝐶| ∼ 𝐸𝐸)
The sufficiency of statistical relevance to causal explanation remains a problem for the Probabilistic-Causal model. Nancy Cartwright provides the example of spraying weed killer on weeds that is 90% effective in killing weeds. Concerning the 90% of weeds sprayed with such a product that suffer from it, when answering the ‘question of
with weed killer’. However, when asking of the 10% of weeds that survive ‘why is this weed alive?’, it is not explanatory to answer in the same fashion despite the fact that the probability that ‘the weed is alive given that it has been sprayed by weed killer’ is
different from the probability that ‘the weed is alive’ (Cartwright, how the laws of physics lie, p.28). Therefore, causally relevant factors need to be properly identified in explanation and this identification must be consistent with statistical relevance. To be explanatorily relevant, it must also provide some further account that would link the causally and statistically relevant factors to the outcome of the phenomenon. The problem that remains is that we currently have the definition of an event being explanatorily relevant to another event iff it is statistically relevant and causally relevant, without having defined what it is to be causally relevant, and without having provided a further link between the causes and the effect.
The Causal-Mechanical Model of Scientific Explanation
The Causal-Mechanical model of explanation employs a different notion of causation from the Deductive-Nomological and Probabilistic Causal models of explanation. Its idea of causation is based on three aspects. The first aspect of this is that it describes
causation according to causal processes (Woodward, 1989, p. 357). These causal processes have the feature of transmitting a mark or its own structure in a spatio-temporally continuous way (Woodward, 1989, p. 357). A mark is an adjustment to the structure of a process. If some thing or object is not capable of transmitting a mark through space-time, then any process it may be involved in is not a causal one but is instead is a pseudo-process (Woodward, 1989, p. 357). The second aspect is that of causal interaction. A causal interaction is when two or more causal processes intersect and subsequently produce a modification in their respective structures (Woodward, 1989, p. 357). Finally, conjunctive forks are used to describe the correlations that exist between two spatio-temporally distinct effects of a common cause (Woodward, 1989, p. 357).
An explanation following the Causal-Mechanistic model explains an event, 𝐸𝐸, by citing the initial causes of the event, 𝐶𝐶1, 𝐶𝐶2, … , 𝐶𝐶𝑛𝑛, and then describing the mechanism
that connects them (Woodward, 1989, p. 358). In other words, it will illustrate the causal processes and causal interactions preceding an event as well as the causal processes and causal interactions of that same event.
Does the Causal-Mechanistic model of explanation then provide an explanatory account of phenomena? At first glance it appears to have good criteria for identifying causal factors of phenomena, that is according to causal processes and causal
interactions with no action at a distance. However, there are still problems with this model of explanation. It is still possible, for example, that cases where there exist two marks of a causal process, and where there is a causal interaction with another causal process in which these marks are transmitted, the model does not provide a good account of how the transmitted marks on the new causal process are causally
dependent on their respective, presumably relevant, marks on the prior causal process (Hitchcock, 1995, p. 310).
The Causal-Mechanistic model improves upon the Probabilistic-Causal model as, although it provides a different approach to identifying the causes of a phenomenon, it requires an explanation to provide a mechanism between the causes of the
phenomenon and its occurrence. This provision of a type of further link between the causes and effects is what the Probabilistic-Causal model lacked.
In this section, it has been found that it is necessary for a scientific explanation to provide some account of the causes of a phenomenon. Differing models of scientific explanation attempt to identify the causes of a phenomenon by different methods and each face certain problems in their respective accounts of explanation. However, to simply cite the causes of a phenomenon, in whichever way they’re identified, is
evidently insufficient for explanation. Therefore, in order for the scientific enquirer to provide an explanation of a phenomenon they must provide information on its causes as well as providing a further account on the connection between the causal and effect phenomena, and their explanatory relevancy to satisfy the condition of sufficiency.
3 Economic Phenomena and the Applicability of Models of Scientific Explanation
to Economics
The preceding discussion has reviewed some of the progress in, and problems with, models of explanation of science within the realm of philosophy of science. It remains to be seen what relevance this has to the field of economics. At this point it will be
motivated that those concerned with the explanation of economic phenomena should also be concerned with issues such as models of explanation. Furthermore, models of explanation in science have typically been formed with the intention of being applied to
the natural sciences. So, to provide a foundation for applying models of scientific
explanation to economics, it can either be established that the extent to which economic phenomena differ from those that models of scientific explanation are designed for is acceptable, or that economic phenomena do not constitute any form of relevance in applying models of scientific explanation to economics.
Positive economics can be considered similar to the natural sciences in the sense that it is concerned with the description, explanation and (subsequent) prediction of phenomena i.e. the goals of economics are identical to those of physics. This is
irrespective of the motivation for doing such activities (Hausman, Explanation and Diagnosis in Economics, 2001, p. 314). Given these common goals, it follows that the economist, as much as any other scientist, should be concerned with the definition of explanation and the varying models of (good) explanation; as well as those of
description and prediction. The scientific enquirer should be concerned with models of explanation to provide a basic foundation and understanding of what it is that they are trying to achieve (i.e. explaining phenomena). For should they claim to have an
explanation of some class of phenomena without understanding what explanation is, their explanation would, at best, be lucky.
Where economics differs from the natural sciences regards the nature of the phenomena under study, and the extent to which they can be observed and theorised about independently of other phenomena. The economist is faced with the problem of explaining economic phenomena according to their causes without having the ability to isolate the desired factors from other influences (Haavelmo, 1944, p. 18). It is not possible to observe many types of economic phenomena in a controlled environment. This often leaves the economist with only passive observations available (Haavelmo, 1944, p. 18). This does not affect the common goals of economics with the other sciences. However, the relevancy of the nature of economic phenomena to the
discussion of models of explanation is found with regard to the nature of phenomena these models are designed to explain. If the nature of economic phenomena are
different from those that a specific model of explanation is designed to account for, then the credibility of applying this specific model to economics is greatly lowered.
As previously discussed, the Deductive-Nomological model is a regularity theory based model of explanation; the phenomena that it deems to explain follow, without exception, from laws and initial conditions i.e. it claims that the explanans is necessary
and sufficient for the explanandum (Hempel & Oppenheim, 1948). Despite the
possibility of nature, and economic phenomena, being deterministic, it is currently an overbearing criterion to suppose that economists should be able to provide an
explanatory account of economic phenomena according to a regularity theory of
causation. For any regularity theory would be falsified by a single instance where it fails to hold, and given the structural complexity of economic systems it is likely that all current non-tautological, economic theory with empirical content could be rejected. Economic events do not seemingly occur in a determined, regular fashion (Hausman, Explanation and Diagnosis in Economics, 2001). If one does not hold a deterministic view of natural phenomena, then the applicability of the D-N model to economics is already non-credible. At any rate, some economic phenomena have the appearance of being non-determined whilst also being deemed to have been explained (Hausman, 2001, p. 314). The credibility of applying the D-N model to economics is, therefore, quite weak.
The Probabilistic-Causal model has many desirable attributes that make it a stronger candidate to be applied to economics than the Deductive-Nomological model, such as allowing for imperfect regularities, and identifying causal irrelevancies by means of statistical relevance, and requiring a factor to have an effect on the probability of a phenomenon’s occurrence. Where it fails is in the ability of the model to identify causal direction. By the definition of conditional probability, on which this model of scientific explanation is based, if the probability of a phenomenon occurring is different to the probability of it occurring given another phenomenon it is deemed an
explanation. However, if these phenomena are switched, the probabilities are also different, and also explanatory
4 Criteria For Good Explanation in Economics
What has been covered so far is a discussion of some differing models of scientific explanation, some (common) problems that they encounter in providing an account of explanation, and a subsequent discussion regarding economics and the nature of economic phenomena and an evaluation of applying these models to economics. This section of the paper shall address the central aim of providing criteria for good explanations in economics.
Daniel Hausman writes: ‘Since philosophers of science have no gospel for scientific practice, economic methodologists have no prepared sermons.’ (Hausman, 1989, p. 123). Whilst it is evident that a model of scientific explanation that accounts for causation, links cause and effect seamlessly, and is applicable to all fields of scientific enquiry has not yet been given, it is possible to learn from the various apparent flaws of accounts of explanation. Humbly then, we can apply what has been learnt, according to the evidence cast up by counter-examples used and our seemingly intuitive prior notions of causation and explanation in refuting the aforementioned models of explanation, to provide basic criteria that an explanation in science should meet in order to be considered satisfactory.
Firstly, any explanation must contain an account of the causes of the phenomena it wishes to explain. In identifying the causes of the phenomena, the issues of causal and explanatory relevance must be addressed. This is common to the Probabilistic-Causal and Causal-Mechanistic models of scientific explanation. Without providing an account of a phenomenon’s causes, and their causal relevance to that phenomenon, it may be the case that a mere regularity exists between the one event and the other. However, does an economic explanation necessarily need to provide a full set of the causes of an economic phenomenon? Or, does an explanation need to provide a sufficient number of causes for the phenomenon to have occurred? This can only be answered by personal preference. Nancy Cartwright holds the view that a complete account should be
provided (Cartwright, 1983, p. 29). This is extraordinarily difficult to achieve, especially in economics, given the complex structures of its understudied phenomena. Therefore, instead, an explanation in economics needs to account for the causes of an economic phenomenon to the minimum extent to which had any one of these causes been absent, the event of the phenomenon would not have occurred. These causes are then
necessary and sufficient for the occurrence of an event. More causal factors may be included if desired, and this can only serve to strengthen an explanation.
Secondly, in identifying an economic cause, a cause must be statistically relevant to the phenomenon to be explained at some point in space and time. This criteria also provides an empirical quality to an explanation upon which it can be falsified or not. However, an economic theory may contain a causal factor that is difficult to observe, or has not been observed. What then should be made of these theories that also satisfy the first criterion of providing a minimal number of sufficient and necessary causal factors?
If it is believed that these theories cannot be currently tested, but there exists a
reasonable hope for the possibility of testing in the future, it has the potential to become a meaningful theory. If there is no possibility of testing the theory now or in the future, then it is of no worth in economic science.
Thirdly, an explanation of economic phenomena should contain an account of the link, or causal mechanism, between the causal phenomena and the effect.
There are further desirable criteria, although not necessary, that will improve the usefulness of an economic explanation. For example, how well an explanatory account in economics provides foundation for prediction of phenomena in the same class. This is obviously a benefit as it aids in achieving the goal of prediction in economics. Finally, if an explanation is consistent with the current body of economic knowledge, then it is a desirable feature as it may corroborate what is already known or deemed to have been explained.
A Final Comment On Methodological Prescription; Mäki’s Commentary on Blaug
In his commentary paper ‘Mark Blaug’s unrealistic campaign for realistic economics’, Uskali Mäki attempts to sway the reader to view Mark Blaug’s desire for a realisation of a culture of falsificationism in economics as a ‘normatively unrealistic’ prescription (Mäki, 2013, pp. 89-90). Mäki believes that the practicability of falsificationism in economics is dubious, or ‘descriptively unrealistic’, and to subscribe to Blaug’s insistency of the implementation of falsificationism in economics is to subscribe to a ‘normatively unrealistic’ goal (Mäki, 2013, pp. 91-92). The current applicability of falsificationism to economics may, admittedly, be difficult to defend given the nature of the phenomena under study and the current method used to explain them. In my opinion, this does not translate into a flaw of prescription, which I believe to be implied earlier in the paper.
Mäki rightly comments that when there exists a ‘discrepancy between theory and the facts’ it is possible to resolve this discrepancy iff ‘one either revises the theory or modifies the facts (or both…)’ (Mäki, 2013, p. 91). According to falsificationism, if there is discrepancy between economic theory and the facts, then one should reject or revise the theory (whilst maintaining its falsifiability). But, Mäki tacitly claims that this
should hold for methodological theory as well, possibly proposing that one should reject falsificationism if the current validity of applying falsificationism to economics is
negated. This viewpoint would, however, take prescriptive methodology and make it subject to the descriptive and to the current abilities of human scientific endeavour. The current realisticness of the methodological viewpoint is too demanding a criterion to reject it as a methodologically prescriptive goal. Instead, if agreed upon, the
prescription should be worked towards. The tools available to the economist and econometrician will only grow in number and quality with the progression of
mathematics and studies in the field of psychology and sociology, and so too will the ability of the field of economics to conform with its methodological prescription improve. If the reader is unconvinced, take Newton and/or Leibniz and their independent work on developing calculus as inspiration for what progress can be achieved in the various fields of science by progress in mathematics.
The role of prescriptive methodology is to posit how a science should be
practiced and the current applicability of a methodological prescription to a field is an admirable quality, but not necessary. Should it be the case that, in the present state of a field of science, that a methodological prescription is not applicable, it should only be rejected if it can never be possible to adhere to such a prescriptive standard. Otherwise, steps should be taken to progress the capabilities of the scientist to be able conform to these standards.
5 Conclusion
The central aim of this paper was to provide clarification on the matter of what is considered to be good explanation in the positive science of economics. In doing so, literature concerning models of scientific explanation have been reviewed and the various qualities and flaws of each model have been discussed. It was found that in order for an account of a phenomenon to be considered explanatory, it must provide information on the causes of the phenomenon. Furthermore, there must be some empirical content in the explanation and there must be a link between the causes and the phenomenon. These models of scientific explanation often encountered difficulties in either attempting to correctly identify causal factors that seem to be causally
relevant, or by failing to eliminate those that are seemingly of non-causal relevance. A further undesirable aspect is that these models of explanation had difficulty in
providing a case of causal direction, and were therefore susceptible to the problem of asymmetry.
The credibility of applying these models of explanation was then evaluated in the third section of the paper. Given that economic phenomena are different to those in the natural sciences, and given the difficulties in observing them independent from other phenomena, certain aspects of the models of scientific explanation were not desirable for economics.
The important idea in this paper is that it is not necessary for economic methodologists to have some ‘gospel’ according to which all economic explanations should adhere to. Rather, it is possible to learn from what we intuitively consider to be explanations, as cast up by refutations and counterexamples to models of scientific explanation, and form criteria that economic explanations should meet in order to be considered good or satisfactory.
Finally, a short comment regarding the current applicability of methodological prescription has been provided so as to ascertain whether the acceptability of
methodological prescriptive goal should be subject to the current state of affairs in economic research.
Bibliography
Boumans, M. J., & Davis, J. B. (2010). Economic Methodology: Understanding Economics as a
Science. Basingstoke: Palgrave Macmillan.
Bromberger, S. (1966). Why-Questions. In Mind and Cosmos (pp. 86-111). Pittsburgh: University of Pittsburgh Press.
Cartwright, N. (1983). How the Laws of Physics Lie. Oxford: Clarendon Press. Friedman, M. (1953). The Methodology of Positive Economics. In Essays in Positive
Economics . Chicago: University of Chicago Press.
Haavelmo, T. (1944). The Probability Approach in Econometrics. Econometrica, 12, 1-115. Hausman, D. M. (1989). Economic Methodology in a Nutshell. The Journal of Economic
Perspectives, 3(2), 115-127.
Hausman, D. M. (2001). Explanation and Diagnosis in Economics. Revue Internationale De
Philosophie, 55, 311-326.
Hempel, C. G., & Oppenheim, P. (1948). Studies in the Logic of Explanation. Philosophy of
Science, 135-175.
Hitchcock, C. R. (1995). Salmon on Explanatory Relevance. Philosophy of Science, 62(2), 304-320.
Mäki, U. (2013). Mark Blaug's Unrealistic Crusade for Realistic Economics. Erasmus Journal
for Philosophy and Economics, 6(3), 97-103.
Reichenbach. (1956). The Direction of Time. Berkeley and Los Angeles: University of California Press.
Reiss, Julian. "explanation." The New Palgrave Dictionary of Economics. Second Edition. Eds. Steven N. Durlauf and Lawrence E. Blume. Palgrave Macmillan, 2008. The New Palgrave Dictionary of Economics Online. Palgrave Macmillan. 16 November 2009 <http://www.dictionaryofeconomics.com/article?id=pde2008_E000290>
doi:10.1057/9780230226203.0534(available via http://dx.doi.org/)
Salmon, W. C. (1971). Statistical Explanation. In Statistical Explanation and Statistical
Relevance (pp. 29-87). Pittsburgh: University of Pittsburgh Press.
Salmon, W. C. (1989). Four Decades of Scientific Explanation. Minneapolis: University of Minnesota Press.
Woodward, J. (1989). The Causal/Mechanical Model of Explanation. In W. C. Salmon, & P. Kitcher, Scientific Explanation. Minnesota Studies in the Philosophy of Science (pp. 357-383). Minneapolis: University of Minnesota Press.
