RESEARCH DOSSIER
Comparative Foundations of Economic Explanation
Abstract. This monograph investigates the comparative foundations of economic explanation through a reconstruction of formal, mechanism-based, historical-institutional, empirical causal and predictive approaches. It distinguishes the explanandum, logical premises, mechanisms, evidence, identification and scope of an explanatory claim. Selected primary writings by Smith, Marx, Veblen, Keynes, Hayek and Lucas provide historical and methodological foundations. Case analyses of minimum wage research, historical instruments in comparative development and multi-country randomized interventions examine the relation between theories and evidence. Stylized analytical examples distinguish equilibrium from adjustment, mathematical implication from empirical applicability, prediction from policy invariance and credible local causal estimates from generalizations. The original synthesis establishes the importance of aligning the form of explanation with the economic question and the evidential warrant. The inquiry is a bounded interpretive and methodological investigation of selected scholarship; it reports no newly estimated empirical effects and introduces no new named theory.
Keywords: Economic explanation; economic methodology; economic theory; causal inference; mechanisms; institutions; history of economic thought; economic models; identification; external validity; prediction; policy evaluation
The object of economic explanation
An economic explanation makes an intelligible and assessable connection between an economic phenomenon and the conditions that generate it. The phenomenon may be a price, an allocation, an institutional arrangement, a distribution of income, a sequence of decisions, or a persistent difference among economies. The connection can be expressed through a logical derivation, a causal contrast, a sequence of actions and interactions, or a historically grounded account of how the relevant conditions came into being. Explanation is consequently an achievement relative to a question. The same object may require several explanations because its existence, level, change, and persistence are distinct problems.
Consider unemployment after a decline in industrial demand. A measure of unemployment establishes what has been counted. A forecast estimates its prospective value. A causal analysis asks how employment would have differed without the decline. A mechanism traces changes in orders, cash flows, hours, hiring, and layoffs. An institutional account examines why employers choose dismissal rather than shorter working hours. An evaluation asks whether employment support is desirable, for whom, and at what cost. These inquiries concern one phenomenon, yet their conclusions have different truth conditions. Treating a forecast as a welfare judgment or an aggregate coefficient as a complete account of adjustment loses precisely the distinctions that make the phenomenon scientifically intelligible.
The divisions between economic methodology, rational choice, and normative appraisal identified in Hausman's Philosophy of Economics (2018) establish a useful starting point. Economic explanation shares questions of causal evidence with other sciences while also addressing intentional action, interdependence, institutional rules and collective outcomes. A decision can be intelligible in light of an agent's beliefs even when those beliefs are inaccurate. A market result can follow from the interaction of many independently intelligible decisions while defeating the objectives of individual participants. A legal rule can alter an outcome by changing which decisions are available, even if preferences remain constant.
The central research problem is therefore how explanatory claims with different logical forms can be compared rigorously. A comparison starts by specifying the explanandum: the exact phenomenon to be explained. It then considers the explanatory premises, the steps connecting premises to conclusion, the relevant evidence, and the conditions under which the connection is expected to hold. These ordinary methodological questions permit comparison across schools without assuming that every school addresses the same explanandum.
The present research dossier undertakes a focused comparative reconstruction. Its sources include selected primary texts in political economy, evolutionary economics, macroeconomics and information economics, together with identified methodological and empirical studies. The selection is purposive rather than exhaustive: it is designed to expose disagreements over the meaning and warrant of explanation. The historical texts establish positions within their publication contexts; the empirical cases illustrate competing evidential arguments. Some modern papers were accessible only through abstracts or bibliographic records. Their treatment is confined to propositions documented at that level. No new dataset, original empirical estimate, or claimed priority for a new theory is introduced.
Historical interpretations of explanatory practice
Classical political economy makes visible the connection between production arrangements and the emergence of surplus. In Smith's (1776) discussion of the division of labor, the organization of manufacturing becomes explanatory through the changes it produces in skill, switching time, and techniques. The arrangement of work is itself an economic variable. The pin-manufacturing example is an illustration of these channels; its logical significance exceeds the numerical illustration. The relevant causal comparison would hold some combination of worker qualifications, input quality, and operating circumstances comparable while changing the organization of tasks.
Smith's explanation anticipates a recurring methodological issue. Specialization can increase output per worker while increasing coordination costs, vulnerability to interruption, or dependence on market size. An investigator who observes higher productivity in larger firms must distinguish the productivity effect of specialization from selection into large firms, access to capital, managerial quality, and demand. The original mechanism is scientifically productive because it generates such questions, rather than because it already supplies a modern identification strategy.
The object of Marx's analysis in the opening chapter of Capital, Volume I (1867) is the commodity form within a particular mode of production. The distinction between useful properties and the social determination of value requires an account of how privately undertaken production obtains a social character through exchange. This approach directs attention toward the organization of production and the historically specific categories through which its results are expressed.
Comparing Smith and Marx directly requires identifying the questions being compared. Smith's manufacturing example investigates changes in productive power; Marx's opening analysis investigates the structure of commodity production. A demonstration that specialization raises output would leave the latter conceptual question largely untouched. Conversely, a reconstruction of the social form of value does not estimate productivity gains from task allocation. Their questions overlap when one investigates the distribution and social organization of the results of production. A serious comparison then specifies exactly which proposition is historical, which is conceptual, and which requires independent quantitative evidence.
In Why Is Economics Not an Evolutionary Science? (1898), Veblen treats cumulative economic development as an explanatory problem in its own right. He criticizes approaches organized around a presumed normal or terminal outcome, and also considers the limitations of historical accounts that merely accumulate observations. The important methodological consequence is that causal reconstruction must encompass changing habits, techniques and circumstances, rather than explain a process solely in relation to its presumed endpoint.
An evolutionary account has its own evidential burden. To explain the persistence of a practice, it must identify the channels of reproduction, adaptation, and abandonment. Invoking evolution by itself leaves selection processes underspecified. Survival could result from efficiency, legal privilege, coordination barriers, sunk investments, or the absence of feasible alternatives. These mechanisms can have the same observable implication over a short period and sharply different implications under a change in conditions.
Keynes (1936) extends a different dimension of the problem through long-term expectation. Prospective yields and confidence enter investment decisions; the future is consequential even where it is difficult to specify its probability distribution. A theory of investment must therefore distinguish the expected return from the realized return and examine how expectations are formed and revised. It matters whether a fall in investment follows a change in financing costs, weakening expected demand, increasing uncertainty, or an interaction among these conditions.
The historical comparison demonstrates why economic schools cannot be ranked solely by their favored method. They often endow different objects with explanatory significance. Production organization, social forms, cumulative change and uncertain expectations can be parts of one historical economy without constituting one interchangeable set of variables. Their mutual relevance must be argued at the level of the specific economic question.
Formal explanation and the representation of constraints
Formal explanation establishes how conclusions follow from specified assumptions. Economic models describe a domain using variables, feasible sets, relations, and solution concepts. The analytical gain is precision: the investigator can identify what must be true for a result to follow and trace how that result changes as selected assumptions change. The epistemic status of the result is conditional on both the mathematical reasoning and the model's relationship to the intended domain.
Take a consumer with nonnegative quantities x and y, utility u(x,y), income m, and strictly positive prices p and q. The feasible set is defined by p x + q y ≤ m. If preferences are represented by a continuous utility function, then compactness of this budget set for finite nonnegative m provides an existence argument for a utility maximum. If utility is strictly quasi-concave on a convex feasible set, the maximum is unique. Neither property alone establishes that actual households behave in accordance with the model or that they possess full price information.
For the illustrative Cobb-Douglas specification u = x^a y^(1-a), with 0 < a < 1 and positive income, an interior solution allocates the shares a and 1-a of income to x and y, respectively. The first-order condition equates the marginal rate of substitution to the relative price, and the binding budget gives x = a m/p and y = (1-a)m/q. This is an analytical example, not an estimate from household data. If x is rationed below its unconstrained demand, these formulas cease to give the feasible optimum. A satisfactory comparison must consider whether the altered outcome reflects different preferences, constraints, or the absence of the assumed choice mechanism.
The distinction matters for the interpretation of apparent anomalies. A consumer's low demand for a service can be represented by a low valuation, but the inference may be incorrect if the service is unavailable in the consumer's region. The same recorded purchase pattern is compatible with both preferences and access restrictions. Data on available alternatives, prices, eligibility, transport and quality can discriminate between them.
Constraint representation also determines the meaning of policy changes. Removing a legal prohibition expands a lawful feasible set; supplying credit relaxes an intertemporal financing constraint; disseminating price information changes beliefs about opportunities; creating a new transport connection changes physical costs. Each can generate increased consumption or investment, yet each operates on a different mathematical object. Modeling all four changes as a rise in income hides economically consequential mechanisms.
Existence, uniqueness, stability and selection are separate propositions. The existence of an equilibrium establishes the compatibility of a set of conditions. Uniqueness excludes other equilibria under those conditions. Stability describes the behavior of a specified adjustment process following a disturbance. Selection concerns the institution, beliefs, learning, or strategic interaction through which one equilibrium is realized. A proof of existence supplies no adjustment process on its own, and a local stability result cannot automatically be extended to global convergence.
A formal model is particularly useful when a researcher needs to separate logically necessary implications from contingent behavior. An accounting identity, for example, holds by definition under its accounting conventions; a behavioral equation generally requires evidence. An identity between expenditure and income totals cannot by itself decide whether an increase in expenditure caused an increase in output, whether output caused expenditure, or whether a third event moved both.
A comparison of formal explanations proceeds by retaining each relevant model's assumptions and deriving its actual implications. Numerical illustrations serve as explicit counterexamples where universal conclusions do not follow. Mathematical correctness is indispensable to formal explanation, and empirical correspondence establishes its appropriate scope.
Mechanisms, incentives, information and interaction
A mechanism-based explanation identifies the activities and interactions that generate an outcome. It gives substantive content to variables that might otherwise appear only as inputs and outputs in a statistical relation. A mechanism can involve purposeful behavior, physical production, communication, legal enforcement, search, learning, bargaining, or feedback. Its elements must be connected in a way that makes the observed outcome possible and that yields observable implications beyond the outcome itself.
Hayek (1945) distinguishes the calculation of an allocation given comprehensive data from the actual problem of using dispersed economic knowledge. The price system in this analysis can coordinate changes in plans without requiring every individual to know all underlying circumstances. This is a mechanism claim about communication and adjustment. It illuminates how decentralized action can reorganize resource use following a change known initially to only some decision makers.
Consider a supply disruption affecting a specialized industrial input. A rise in its price may induce substitution, reduced use and new production. The proposed channel requires that buyers encounter relevant prices, possess alternative technologies or suppliers, and have enough contractual freedom to adjust. A price rise can also reflect strategic withholding or an increase in demand, and some buyers may respond slowly under fixed contracts. Evidence of the price change alone therefore establishes neither the initial cause nor the efficiency of subsequent allocation. The timing of quantities, inventories, substitution, contract renegotiation and entry matters.
This example illustrates the distinction between a mechanism and an equilibrium description. An equality of quantities demanded and supplied can characterize the final configuration; an explanation of the transition describes the informational and organizational activities connecting the initial and final configurations. Under alternative arrangements, the same final quantity could be reached through rationing, administrative allocation, waiting queues or vertical integration. These adjustment mechanisms may differ significantly in costs and distributional consequences.
Strategic interaction adds another explanatory element. A firm's choice depends upon the anticipated reaction of competitors, and an employee's behavior may depend upon an employer's expectations. In a coordination game with two feasible conventions, the payoffs can support either convention as an equilibrium. A formal equilibrium comparison can show why mutual conformity is sustainable; a historical account of communications, precedents and expectations may explain why a particular convention was selected. Evidence that participants coordinate successfully establishes neither the uniqueness nor the social desirability of the selected rule.
A mechanism requires discriminating evidence. Suppose customers change providers following a billing reform. A search-cost explanation predicts changes concentrated where switching becomes easier. A liquidity explanation predicts stronger effects among households previously unable to pay upfront. A quality explanation predicts changes associated with service characteristics. The observed increase in switching could arise from any combination of these processes. Measuring the intermediate conditions improves the capacity to distinguish the accounts.
Mechanisms are often nested. A change in a tax rule can affect a firm's balance sheet, shift its financing constraints, and alter investment. Its suppliers may then change prices or capacity, while workers adjust plans in anticipation of new opportunities. Aggregate outcomes depend on those reactions. A mechanism established in one firm can be part of a broader explanation without automatically determining the aggregate sign or magnitude of the effect.
Thus mechanism evidence supplements both formal reasoning and causal estimation. It indicates why an effect should arise in some circumstances, why it might disappear under others, and what other effects should accompany it. Mechanistic plausibility is an evidential contribution with a particular scope. Demonstrating that a mechanism could operate is different from establishing that it operated in the case under study, and estimating its quantitative importance is a further task.
Historical and institutional explanation
Historical explanation reconstructs the formation, transformation, and persistence of economic arrangements. Its unit of analysis can be a legal rule, an organization, a technology, a market convention or an interconnected set of practices. Temporal ordering, documentary provenance and the changing meaning of institutions are central because the object itself may evolve during the period being examined.
North's (1990) introduction to institutional change treats institutions as humanly devised constraints shaping incentives, and develops the problem of how past developments influence present opportunities. A rule can constrain bargaining, allocate rights, determine enforceability, and affect the costs of transactions. The historical origin of a rule and its contemporary consequences are separate explananda. A rule may arise through political conflict and subsequently persist through coordination advantages; it may also survive despite costs because changing it threatens the interests of actors with authority.
Path dependence is a claim about the influence of earlier developments on later feasible transitions. It becomes explanatory when the investigator identifies the relevant inheritance: physical capital, legal entitlements, networks, accumulated knowledge, switching costs, or distributional vetoes. Merely observing that an institution existed for a long time does not establish a particular path-dependent mechanism. A credible account examines why an alternative route was feasible at one decision point and more expensive or politically difficult later.
In an illustrative railway example, an early choice of track gauge affects subsequent compatibility with existing rolling stock and connections. The coordination consequences arise from investments and network complementarities rather than from the mere age of the rule. A replacement may become economical if gains from interoperability outweigh conversion costs, but evaluating that proposition requires information about the network and the distribution of those costs. The example is deliberately stylized and supplies no claim about the origins or efficiency of a particular national railway system.
Historical evidence involves both written rules and operative practices. The date a law is adopted may differ from its effective date, enforcement date or recognition in administrative records. An institutional comparison based on legislation alone can misclassify the treatment experienced by firms and households. In countries where formal arrangements coexist with extensive customary or informal practice, distinguishing de jure rules from de facto enforcement is particularly important.
Chronology can exclude some explanations. An alleged cause arising only after its alleged effect cannot explain that earlier effect through an ordinary forward causal pathway. Anticipation, however, complicates the chronology because economic agents may react before a formal policy date. Investigators should identify the first credible information event and distinguish it from the date of enactment. Historical material also helps detect alternative causal pathways, such as simultaneous legislation, conflict, commodity shocks or migration.
The study of governance illustrates the need to preserve institutional detail rather than assume a simple distinction between state and market. Ostrom's (2010) published lecture on polycentric governance identifies arrangements with multiple centers of decision as a research object. A comparative investigator should distinguish the allocation of authority, the boundaries of resource use, monitoring, and mechanisms for resolving disputes. Their configuration can matter more than a broad label such as public or private. This is a methodological implication of the object's specificity; the lecture's empirical findings are not independently evaluated here.
Historical and institutional explanations are strengthened when their narratives connect to explicit alternatives and independently inspectable records. Cross-case comparison can investigate different paths under partly similar circumstances. Within-case comparison can follow the sequence of implementation and resistance. Documents, measured outcomes and formal reconstructions provide complementary forms of constraint on the interpretation.
Prediction, idealization and policy invariance
Prediction is an inference about an unknown or future outcome in light of available information. Its performance can be evaluated against observations under a specified forecast horizon and loss function. A model may predict well because it captures a stable empirical regularity, even when its variables have no established causal interpretation. The value of prediction is consequently substantial for economic planning, while its relevance to policy intervention depends on additional assumptions.
Friedman's (1953) methodology of positive economics examines the assessment of positive economic hypotheses. His influential predictive emphasis and the ambiguity surrounding unrealistic assumptions are discussed in Hausman (2018). A useful distinction is between simplifications that suppress features irrelevant to the intended prediction, approximations that introduce a known error, and assumptions that determine the direction of the result. The predictive adequacy of a model can justify its limited use without resolving which of these roles a particular assumption plays.
Assume a stylized retailer predicts weekly sales using the relationship Y = 100 + 2X, where X is the observed volume of visits in chosen units. The equation could successfully forecast because visits precede sales, because a third variable such as advertising drives both, or because the measured visits include customers already committed to purchasing. Predictive success alone cannot determine the effect of an intervention that artificially increases visits by people with no interest in buying. The causal question changes the required evidence.
For policy evaluation the problem becomes sharper. Lucas (1976) distinguishes reliable short-horizon forecasts under existing behavior from simulations of alternative policies in his critique of econometric policy evaluation. A change in the policy rule can alter expectations and decision rules, changing the parameters of the observed relationship. The relevant invariance is therefore not simply the repetition of a regression coefficient over adjacent periods. It concerns the economic relationships that remain appropriate under the intervention contemplated.
A simple algebraic illustration clarifies the distinction. Let output deviation y = a(π − e) + ε, where π is actual inflation, e is anticipated inflation, a is a positive sensitivity, and ε is a mean-zero disturbance. If inflation unexpectedly increases by d while e remains fixed, the conditional prediction is a d more output deviation, holding the disturbance constant. If a permanently repeated, fully anticipated change raises both π and e by d, their difference remains the same. The example depends on the stipulated surprise mechanism; it makes no assertion about the empirical magnitude, duration or complete structure of monetary policy effects.
A policy critique is itself subject to empirical analysis. An apparent absence of instability in estimated coefficients may reflect a genuinely small change in underlying behavior, insufficient variation in policy, low statistical power, or compensating responses. Conversely, a break could arise from technological change or measurement revisions rather than a policy rule. Forecasting, estimation and structural interpretation should be compared using the nature of the anticipated intervention and evidence about behavioral adaptation.
The relation between abstraction and realism can thus be evaluated more precisely than by asking whether a model resembles reality in every respect. An abstraction is appropriate when the suppressed features are immaterial to the target result within the stated domain. If the suppressed feature is altered by the policy being evaluated, the explanatory adequacy of the abstraction requires renewed examination. In this sense the same model can be adequate for ordinary prediction and inadequate for a different policy question.
Empirical causal explanation and identification
Empirical causal explanation requires a comparison between outcomes under well-defined alternative economic conditions. A causal estimand expresses the intended contrast; an identification strategy describes the assumptions under which that contrast can be recovered from the available observations. Estimation then turns the identified relation into a numerical result, accompanied by uncertainty. These are separate tasks.
Let D denote assignment to an economic intervention, with values 0 and 1, and let Y(1) and Y(0) denote the potential outcomes under the two assignments. The average treatment effect within a specified population is E[Y(1) − Y(0)]. Only one realized potential outcome is observed for each unit under its actual assignment, so the missing comparison requires a research design. Under random assignment, consistency of treatment definition, and an appropriate no-interference assumption, the difference in expected outcomes across assigned groups identifies the assignment effect. Finite-sample uncertainty remains. If assignment and receipt differ, the effect of assignment must be distinguished from the effect of treatment receipt.
Angrist and Pischke (2010) emphasize the development of stronger empirical research designs. Research designs direct attention toward the process that generates comparison groups rather than treating regression adjustment as sufficient by default. A before-and-after comparison, an observational regression and a randomized study may use similar equations while having markedly different identifying assumptions.
Difference-in-differences compares average changes in a treated group and a comparison group. With one pre-intervention and one post-intervention period, its simplest estimator is (Y_T,post − Y_T,pre) − (Y_C,post − Y_C,pre). A causal interpretation requires that, absent treatment, the expected change in the treated group would have matched that of the comparison group, along with assumptions concerning anticipation, stable composition and relevant spillovers. A common baseline level is unnecessary; a common counterfactual trend is central.
For illustration, suppose average weekly employment hours in the treated establishment group change from 100 to 108 and in the comparison group from 95 to 99. The change contrast equals 4 hours. This is the result of arithmetic applied to hypothetical values, not a measurement of actual employment. If the untreated counterfactual trend in the treated group would independently have been 8 hours, the estimated contrast overstates the effect. The identification assumption contributes substantive content that the arithmetic cannot establish.
An instrumental-variable design uses a variable Z that shifts an exposure D while satisfying restrictions needed to interpret exposure-induced changes in Y. Relevance requires a relationship between Z and D. Independence or as-if-random variation relative to the relevant potential outcomes and exclusion of independent pathways to Y are substantive requirements. Under a simple binary instrument and treatment, monotonicity and the other standard assumptions yield a local average treatment effect for compliers. The result may have a different scope from the average effect for all units. A continuous historical instrument with a continuous institutional exposure should not mechanically be interpreted using the binary complier formula.
Regression discontinuity can compare observations near an assignment cutoff under continuity of relevant potential outcomes and suitable restrictions on manipulation or sorting. An effect local to the threshold may answer a narrow policy question very well. Extending it to units far from the threshold requires further support. Observational longitudinal designs, structural estimation and historical comparisons involve different assumptions and can also address causal questions.
Measurement, sampling, and uncertainty determine the quality of estimates after identification has been argued. Standard errors should reflect the relevant assignment level and dependence structure. Missing outcomes, selective participation, and changing composition can alter the intended estimand. Where a treatment is applied at a community level, individual records may yield many observations but relatively few independent assignments. A large number of rows alone cannot establish high precision.
The methodological interpretation of causal evidence depends equally on internal validity and the intended target. A local, credible causal estimate may provide stronger support for a precisely bounded proposition than a broad but poorly identified estimate. The broader question can then be investigated through complementary evidence rather than by enlarging the verbal interpretation of the result.
Competing labor market explanations
The minimum wage offers a transparent setting for comparing formal reasoning, institutional specification and causal evidence. In a simple competitive market, a firm hires labor until the marginal value of its product equals the wage. A binding wage floor can reduce quantity demanded if the demand curve slopes downward and the other conditions of the comparison remain fixed. The result is a conditional comparative-static proposition. It does not account by itself for adjustments in prices, hours, productivity, vacancies, exits, worker composition or the degree of labor market competition.
An alternative illustrative model assumes that a single employer faces an upward-sloping labor supply because workers have limited outside options. Let the inverse labor supply be w(L) = A + B L, with B > 0, and let the marginal revenue product of labor be a constant V greater than A. With a uniform wage, total labor cost is A L + B L²; marginal labor cost is A + 2B L. The unconstrained monopsonist chooses L_m = (V − A)/(2B). In a competitive benchmark with the same supply schedule and constant V, L_c = (V − A)/B. A suitably set minimum wage above the monopsony wage can permit increased employment by flattening marginal labor cost over a range of employment levels. Very high binding floors can reduce employment instead. This example requires a specific uniform-wage and labor supply structure and establishes a logical possibility, not a diagnosis of a particular real labor market.
These models generate different conditional predictions. Discriminating among them demands evidence about worker mobility, recruitment, concentration, quit behavior, wage setting and the relationship between labor demand and the specific policy. An observed employment effect near zero is compatible with multiple mechanisms, including an initially nonbinding floor, offsetting price and productivity changes or market power. The sign of an estimate is not a uniquely identifying test of any one model.
Card and Krueger (1994) examine the New Jersey minimum wage change. On April 1, 1992, the statutory hourly minimum increased from $4.25 to $5.05 in New Jersey while the neighboring Pennsylvania minimum remained unchanged. Their published abstract describes a survey of 410 fast-food restaurants and reports no indication of lower employment following the increase. The observational design compares changes in neighboring areas as well as restaurants with different initial wage exposure.
Neumark and Wascher (1995) subsequently used payroll-based employment records from 230 restaurants and reported a contrasting estimated employment response in their working-paper abstract. The researchers' dispute continued through a published comment (Neumark and Wascher, 2000) and a published reply (Card and Krueger, 2000). The latter records establish the documented debate; the complete published tables and reply estimates have not been reproduced in the present dossier.
The comparison exposes several layers of an explanatory claim. A statutory change is the treatment, but its effective intensity depends on prior wages, compliance and timing. Employment can refer to headcount, full-time equivalents, payroll hours or the number of establishments. Data collected by telephone and administrative payroll data may carry different measurement errors; samples may have different coverage or composition. Even an accurately computed estimator can produce different quantities when the outcomes or samples differ.
A careful comparison would align the geographic coverage, observation windows, establishment population, outcome definitions and treatment exposure before interpreting the difference in estimates. It would investigate establishments that exited or entered, missing observations, changes in demand across the comparison border, and sensitivity to the functional form. Because this dossier inspected accessible abstracts rather than original establishment records or all full-text tables, it does not adjudicate the final magnitude or general sign of the employment response.
The analytical implication is independent of that verdict. Conditional theoretical results provide a basis for comparing proposed mechanisms; credible empirical designs constrain the plausible responses under a particular intervention; measurement and institutional evidence determine which treatment and outcome are represented. A complete explanation of the labor market response requires their connection at the same geographical, economic and temporal level.
Comparative development and historical instruments
Differences in income per person across countries can be approached through productivity, capital accumulation, education, geography, market integration, state formation, and political institutions. These explanations need not be independent. A legal institution can affect investment, which changes capital stocks and productive opportunities; geographic circumstances can affect transport and disease exposure while also shaping settlement. A causal study of development must therefore distinguish the channel it seeks to isolate from the background conditions that may affect the same outcomes.
Acemoglu, Johnson and Robinson (2001) use historical colonial mortality as an instrument for present institutions. Their study proposes that European settlement conditions influenced colonial institutional arrangements, which persisted and affected subsequent economic performance. The article reports substantial estimated effects of institutions on income per capita. The argument is historically structured: the instrument is relevant through a proposed settlement and institutional channel, while exclusion requires that the relevant mortality variation not affect the economic outcome through independent pathways after the specified conditioning.
The distinction between relevance and exclusion is important in this setting. A statistical association between colonial mortality and institutional measures can support the first-stage relationship without demonstrating that the mortality measure captures an otherwise isolated historical source of variation. Disease exposure, geography, migration patterns, colonial investment and political violence are examples of potential pathways that require substantive examination. Whether they violate exclusion depends upon the defined exposure, conditioning variables and causal structure. It cannot be decided by enumerating possible pathways alone.
A historical instrument also inherits the measurement history of the records used to construct it. Records of soldiers, settlers, officials and laborers may refer to different populations, periods, activities and conditions. Comparability among countries becomes an evidential question, not a clerical detail. An assigned value for a country based on a neighboring area must be justified through the relevant geography, administrative context and source provenance.
Albouy's (2012) critique of mortality data challenges the comparability and assignment of observations and reports reduced robustness of the instrumental-variable results. Acemoglu, Johnson and Robinson's (2012) response to the critique disputes important exclusions and coding decisions and reports robustness under alternative handling of influential observations. These are attributed findings from opposing analyses. The present investigation has checked the published abstracts and bibliographic identities; it has not replicated the datasets or resolved the disputed historical coding.
The substantive lesson concerns the relation between historical and econometric explanation. A numerical coefficient derived from earlier records depends on the identity and meaning of those records. If the process generating the instrumental variation is misunderstood, the causal interpretation can fail even when the computation is correct. Conversely, a persuasive historical narrative does not by itself establish a quantitative causal effect or its uncertainty. Detailed source reconstruction and statistical analysis address different indispensable parts of the research problem.
Many international development debates also involve measurement differences in the dependent variable. Income per capita can be expressed in current domestic currency, constant domestic currency, converted market exchange rates or purchasing-power-parity units. These choices affect the magnitudes and comparisons. In a long-run historical study, territorial boundaries and changing statistical capacity add further difficulties. The present dossier makes no quantitative cross-country income comparison, avoiding the need to supply a new series and its vintage.
The development case demonstrates why generalization is substantive. An institutional mechanism identified through one historical process may not establish the impact of changing present law through an entirely different political process. The margins of variation differ, and changes in enforcement may matter more than changes in statutory language. Rigorous comparison keeps the historical channel, estimated parameter and contemporary policy interpretation distinct.
Randomized evidence across countries and institutional contexts
Randomization changes the evidential status of a comparison by creating a known assignment mechanism. In economic development research, experiments can investigate policies and organizational arrangements that vary across communities or households. Their results contribute to scientific explanation when the treatment, outcome, population, implementation and comparison are described with sufficient precision.
The six-country synthesis of microcredit evaluations by Banerjee, Karlan and Zinman (2015) finds modest rather than transformative effects in broad terms. The review concerns causal evaluations of expanded access to microcredit across diverse locations and institutional arrangements. Its conclusion is restricted to the programs and outcomes investigated; it does not identify a universal effect of all lending technologies, interest rates or borrower populations.
A different multi-country investigation by Banerjee and coauthors (2015) evaluates a multifaceted Graduation program involving productive assets and complementary forms of support. The study reports improvements in consumption and other measured outcomes that continued for at least one year after the intervention ended. The six settings include Ethiopia, Ghana, Honduras, India, Pakistan and Peru. The breadth of this program of research provides evidence about implementation across distinct settings, while the findings remain tied to the evaluated treatment bundle, populations and follow-up period.
The two bodies of research do not provide a head-to-head randomized comparison between credit access and the Graduation approach. They examine different interventions with different costs, operational requirements, selection processes, and outcome distributions. Comparing their broad findings can motivate questions about complementarities among resources, support and constraints. Establishing that a particular component is responsible for a difference would require more specific evidence.
An economic mechanism may involve a combination of constraints. A household could possess the skills to use a productive asset but lack savings sufficient to purchase it. Another household could receive an asset while facing high consumption risk that induces its sale. A third could lack access to customers or local services needed to operate an enterprise. An intervention combining an asset with training and temporary support addresses several possibilities simultaneously. Its effectiveness, when observed, does not identify which component is necessary or sufficient.
A component-level causal question would require a design that distinguishes alternatives, such as variation in assets, training, or consumption support. Interactions are important: the combined effect can exceed or fall short of the sum of the separate effects. In a simple factorial design, the interaction is the contrast between the effect of one component in the presence and absence of the other. Such an illustrative comparison differs from the effect of the complete bundle against ordinary conditions.
Deaton and Cartwright (2018) examine misunderstandings of randomized trials, including the distinction between random assignment, statistical precision and the applicability of estimated effects to other populations. Their analysis supports evaluating experiments within a cumulative program of scientific inquiry. In a single trial, randomization balances covariates in expectation across hypothetical assignments, not necessarily exactly in the realized groups. Transport to a new context requires a justified connection between the characteristics of the study and target populations.
This issue becomes economic when a policy is scaled. A small intervention that helps selected firms may raise their market shares through displacement of nonparticipants. A nationwide intervention changes the conditions under which firms compete and may alter prices, wages, or input availability. The effect estimated in the smaller experiment then answers a specified local question; the aggregate effect requires further theory and measurement of those feedbacks.
Cross-country experiments advance generalization by testing effects over a wider range of institutional realizations. A scientific synthesis should compare treatment definitions, targeting, costs, implementation quality, outcome measures, price levels, follow-up horizons and possible spillovers. Repeated findings strengthen evidence for a class of phenomena, and variation across settings can be especially informative about the circumstances and mechanisms determining the result.
Comparing explanatory strength across schools
Rigorous comparison begins when two claims concern the same defined outcome and the same counterfactual or historical question. It is possible for explanatory traditions to contribute to one inquiry without competing directly. In other cases, they make incompatible predictions given the same conditions and must be assessed through the evidence. The task is to establish the relation between their claims before judging which has greater explanatory strength.
A formal derivation can establish impossibility under a defined set of assumptions. A mechanism account can establish a coherent process connecting a cause and consequence. A historical account can establish formation and persistence through identifiable events and institutions. An empirical design can establish a causal contrast conditional on its identification assumptions. The following comparison describes the characteristic contributions and limitations of those approaches; the boundaries are methodological distinctions rather than rigid classifications of researchers.
| Explanatory approach | Characteristic question | Strongest form of warrant | Typical unresolved issue |
|---|---|---|---|
| Formal | What follows from the stipulated conditions? | Valid derivation and explicit assumptions | Applicability and choice of model |
| Mechanism-based | How is the outcome generated? | Observed or otherwise evidenced intermediate processes | Relative importance and hidden alternatives |
| Historical-institutional | How did the current arrangement arise and persist? | Traceable chronology, documentary comparison, process evidence | Counterfactual attribution |
| Empirical causal | How much does a defined change affect an outcome? | Credible identification, measurement, and uncertainty | Transfer to new regimes and populations |
| Predictive | What outcome should be expected under stated information? | Verified performance against suitable benchmarks | Causal or intervention interpretation |
The table identifies dimensions relevant to particular research tasks. It does not assign exclusive ownership of any method to a school. Formal analysis can represent evolving institutions, historical evidence can support causal identification, and experiments can investigate mechanisms. The same research project may consequently use several approaches without changing its object.
Comparative judgment should distinguish the status of assumptions. A definition fixes the meaning of an object. An idealization isolates features considered important. A behavioral hypothesis makes a claim about economic action. An identifying restriction connects observable quantities to a causal parameter. A normative premise specifies a criterion of evaluation. A model can contain all of these at once, but the evidence relevant to each differs. Definitions require coherence and appropriate use; behavioral and identifying claims require contact with the phenomenon; normative premises require an articulated evaluative justification.
A useful counterexample is the correlation between insurance coverage and losses. A positive association might reflect adverse selection, a change in risk-taking after coverage, a pricing arrangement that directs high-risk customers toward particular products, or variation in reporting. A model of selection provides a mechanism conditional on information asymmetry; a model of moral hazard describes behavioral responses; a randomized coverage change can address certain causal questions; historical contractual records can establish how coverage was actually assigned. Income or age adjustment alone will not generally distinguish all these interpretations.
When two accounts have the same predicted outcome, discriminating evidence may lie in additional outcomes or timing. If one explanation predicts a rise in entry and another predicts more intensive use of existing capacity, the total output can coincide while the composition differs. The investigator can seek information on establishments, hours, capacity, and productivity. When such information is unavailable, the scientifically warranted result may be a set of compatible explanations rather than a unique verdict.
Comparative evaluation therefore turns on relevance, logical consistency, specific evidence, discriminating implications and scope. A broad account with many adjustable elements can accommodate numerous observations, but accommodation has less evidential value where parameters were selected after observing the outcomes. A narrower account can be more informative when it produces precise, independently assessable predictions. Scope and specificity must be jointly considered.
Original analytical synthesis
The preceding comparisons establish that economic explanation becomes strongest when the form of its claim is aligned with the question, the evidence and the institutional setting. This is a synthesis of the examined traditions and cases. It does not introduce a new named theoretical system.
First, the existence of a logical implication is a different scientific achievement from establishing the prevalence of its premises in the world. The formal labor-market illustrations show why a competitive model and a monopsony model can imply different responses to a binding wage floor. The result of the comparison is an exact statement of the conditions under which a sign follows. An empirical result then restricts possible accounts in the actual setting without making the premises of a single model true merely by agreement of signs.
Second, historical evidence contributes directly to causal identification when past records define the source of contemporary variation. The colonial mortality controversy exemplifies this relationship. The provenance and comparability of a historical instrument are part of the conceptual interpretation of an estimate. A material source defect cannot be repaired merely by a more sophisticated estimator. In the opposite direction, historical continuity in the documentary record does not itself identify the quantitative magnitude of the effect of an institution.
Third, experimental variation contributes most directly to the comparison for which the treatment and assignment are designed. The development cases distinguish the effect of expanding access to a particular financial opportunity from the effect of an integrated package. Their contrast makes complementarities a meaningful subject of investigation, while preserving the difference between that question and evidence identifying a specific component. A general economic explanation is built by investigating the mechanisms that account for the observed effects and the contexts where they change.
Fourth, a reliable prediction can coexist with ambiguity about causation. The Lucas critique illustrates the additional importance of the policy regime when an intervention alters expectations. A forecasting model's success for a particular horizon supplies evidence about that forecasting task. A policy claim must establish the applicability of the model under the changed rule, and a structural model must in turn be tested against observations that are informative about its purported invariance.
These results lead to an ordinary practical sequence for economic research. Define the economic object at the level where it is observed; specify the comparison implicit in the question; derive the implications of the relevant theories; identify the institutions and mechanisms linking conditions to outcomes; evaluate the design and provenance of the evidence; distinguish the remaining alternatives; and delimit the interpretation. The sequence describes the research work performed here and follows from established methodological distinctions, without claiming originality as a general method.
A comparative explanation of a hypothetical technological change in a port illustrates their integration. A formal production model predicts that lower handling costs can expand throughput conditional on demand and capacity. A mechanism account examines shorter unloading time, reduced vessel delays, and responses by carriers. Historical and institutional analysis identifies concession rights, labor arrangements and the sequence of infrastructure adoption. An empirical study might compare changes in ports with different adoption dates, taking account of selection and preexisting trends. A predictive model could estimate future congestion. Welfare analysis would additionally assess costs, wages, local environmental effects and the distribution of benefits.
Each approach answers a scientifically meaningful question. A strong integrated account specifies which claims have actually been established, how their conclusions are connected, and how feedback across trading networks affects the result. It can remain incomplete in one dimension while supplying strong evidence in another. This is preferable to collapsing a complex event into one explanatory label.
Normative implications likewise require an explicit bridge between findings and objectives. A policy could increase aggregate production while reducing income for a particular population. Whether it is desirable depends on distributional weights, rights, transition costs, fiscal consequences and the feasible alternative policies. Estimating the production effect is an important input; it is not a complete evaluation. A comparative science should preserve both the empirical results and the evaluative premises so that the reader can determine precisely where disagreements originate.
The authorial contribution is the cross-case analysis of these differences in warrants and the demonstration, through stylized derivations and counterexamples, that explanatory success in one dimension does not automatically establish success in another. No empirical estimates are attributed to the author. No priority claim is made for the synthesis or its underlying methodological distinctions.
Research boundaries and open scientific questions
The present dossier compares selected research traditions and empirical controversies rather than reconstructing the entire history of economic methodology. Its primary-text readings focus on identifiable passages relevant to the explanation of economic phenomena. The treatment of modern papers for which only abstracts or metadata were accessible is limited to documented claims. No original data collection, systematic review, meta-analysis, or replication of source estimates was undertaken. The illustrative models and calculations are intended to test logical distinctions; their parameters are stipulated.
A continuing substantive question is how to distinguish mechanisms operating concurrently within complex economies. When several changes affect a market simultaneously, empirical designs may identify a total effect more convincingly than the component effects. Research can improve this situation through measurements of intermediate outcomes, independent sources of variation, transparent structural assumptions, and designs that compare the effects of components. Each strategy has limitations when intermediate variables are themselves endogenous or measured after treatment.
A second question concerns the stability of explanations under institutional transition. The same observed relationship may have different meanings as accounting practices, technologies, legal rights and policy expectations change. Comparisons across decades require careful reconstruction of the measurement regime. Future research can investigate when parameter changes are economically meaningful and when they arise because the measured variables no longer represent comparable objects.
External validity is a third persistent issue. The six-country examples indicate the value of investigation across distinct settings. Yet breadth of geographic coverage alone does not establish transferability to a particular new population. A research program can investigate which treatment mechanisms depend on individual characteristics, implementation institutions, market structure, or macroeconomic conditions. Theory, local institutional evidence and multi-site data can thereby complement one another.
Finally, comparative history of economic thought can examine how explanatory problems were reformulated as new forms of evidence became available. Historical texts should retain their original conceptual and institutional settings. Their propositions may motivate present research questions without being treated as ready-made modern causal estimates. Distinguishing historical exegesis from contemporary empirical application preserves the value of both tasks.
The dossier's examples are drawn principally from European intellectual history, United States labor-market research, comparative colonial development, and multi-country studies of development interventions. They support the methodological comparison and demonstrate multiple international settings. They do not establish exhaustive geographic representation of economic research. Further work may extend the comparative analysis to specific Asian, African, Latin American, Middle Eastern and other economic traditions through their original intellectual and institutional sources.
Conclusion
Economic explanation establishes an assessable connection between an economic outcome and the conditions that generate or sustain it. Different explanations achieve different forms of scientific understanding. Formal reasoning identifies what follows under stated assumptions; mechanism analysis investigates the processes of generation; historical and institutional research reconstructs formation and persistence; empirical causal designs estimate specified contrasts under identifying conditions; predictive analysis tests performance for defined information and targets.
The comparative approach developed in this dossier preserves the distinct evidential status of each achievement while examining their connection. Its historical readings and empirical cases show that a disagreement may concern the object being explained, the assumptions supplying a result, the validity of a measurement, the causal comparison, or the institutional context in which the relation is expected to hold. Identifying that specific disagreement is essential to assessing explanatory strength.
The labor-market case illustrates the difference between a conditional model prediction and an intervention estimate. Comparative development shows the substantive importance of the historical provenance of causal evidence. Experiments across countries show how evidence concerning particular interventions can contribute to a broader investigation of mechanisms and boundary conditions. The distinction between forecasting and policy evaluation reveals the consequences of changing expectations and regimes.
The resulting conclusion is that economic research gains explanatory precision through correspondence among its questions, assumptions, mechanisms, historical evidence and methods of identification. The most suitable combination depends on the economic phenomenon and the claim at issue. This comparative discipline enables explanations from different traditions to be assessed within a common scholarly inquiry while retaining their substantive specificity and evidential limits.
References
- Acemoglu, Daron, Simon Johnson, and James A. Robinson (2001). The Colonial Origins of Comparative Development: An Empirical Investigation. American Economic Review 91(5): 1369–1401. DOI: 10.1257/aer.91.5.1369. Accessible record.
- Acemoglu, Daron, Simon Johnson, and James A. Robinson (2012). The Colonial Origins of Comparative Development: An Empirical Investigation: Reply. American Economic Review 102(6): 3077–3110. DOI: 10.1257/aer.102.6.3077. Accessible record.
- Albouy, David Y. (2012). The Colonial Origins of Comparative Development: An Empirical Investigation: Comment. American Economic Review 102(6): 3059–3076. DOI: 10.1257/aer.102.6.3059. Accessible record.
- Angrist, Joshua D., and Jörn-Steffen Pischke (2010). The Credibility Revolution in Empirical Economics: How Better Research Design Is Taking the Con out of Econometrics. Journal of Economic Perspectives 24(2): 3–30. DOI: 10.1257/jep.24.2.3. Accessible record.
- Banerjee, Abhijit, Dean Karlan, and Jonathan Zinman (2015). Six Randomized Evaluations of Microcredit: Introduction and Further Steps. American Economic Journal: Applied Economics 7(1): 1–21. DOI: 10.1257/app.20140287. Accessible record.
- Banerjee, Abhijit, Esther Duflo, Nathanael Goldberg, Dean Karlan, Robert Osei, William Parienté, Jeremy Shapiro, Bram Thuysbaert, and Christopher Udry (2015). A Multifaceted Program Causes Lasting Progress for the Very Poor: Evidence from Six Countries. Science 348(6236): 1260799. DOI: 10.1126/science.1260799. Innovations for Poverty Action publication record. Accessible record.
- Card, David, and Alan B. Krueger (1994). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. American Economic Review 84(4): 772–793. RePEc:aea:aecrev:v:84:y:1994:i:4:p:772-93. Accessible record.
- Card, David, and Alan B. Krueger (2000). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Reply. American Economic Review 90(5): 1397–1420. DOI: 10.1257/aer.90.5.1397. Accessible record.
- Deaton, Angus, and Nancy Cartwright (2018). Understanding and Misunderstanding Randomized Controlled Trials. Social Science & Medicine 210: 2–21. DOI: 10.1016/j.socscimed.2017.12.005. Accessible record.
- Friedman, Milton (1953). The Methodology of Positive Economics. In Essays in Positive Economics, pp. 3–43. Chicago: University of Chicago Press. Reprinted in Uskali Mäki (ed.), The Methodology of Positive Economics: Reflections on the Milton Friedman Legacy (2009), pp. 3–44. Cambridge University Press. Reprint DOI: 10.1017/CBO9780511581427.002. Accessible record.
- Hausman, Daniel M. (2018). Philosophy of Economics. The Stanford Encyclopedia of Philosophy. Originally published September 12, 2003; substantive revision September 4, 2018. Stanford University. Accessible record.
- Hayek, Friedrich A. (1945). The Use of Knowledge in Society. American Economic Review 35(4): 519–530. Authorized online reprint, Library of Economics and Liberty. Accessible record.
- Keynes, John Maynard (1936). The General Theory of Employment, Interest and Money. Chapter 12, The State of Long-Term Expectation. London: Macmillan. Online English transcription, Marxists Internet Archive. Accessible record.
- Lucas, Robert E., Jr. (1976). Econometric Policy Evaluation: A Critique. Carnegie-Rochester Conference Series on Public Policy 1: 19–46. DOI: 10.1016/S0167-2231(76)80003-6. Accessible HTML transcription of original article. Accessible record.
- Marx, Karl (1867). Capital: A Critique of Political Economy. Volume I, chapter 1, Commodities. Hamburg: Otto Meissner (German original); English online transcription, Marxists Internet Archive. Accessible record.
- Neumark, David, and William Wascher (1995). The Effect of New Jersey's Minimum Wage Increase on Fast-Food Employment: A Re-Evaluation Using Payroll Records. NBER Working Paper 5224. DOI: 10.3386/w5224. RePEc:nbr:nberwo:5224. Accessible record.
- Neumark, David, and William Wascher (2000). Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment. American Economic Review 90(5): 1362–1396. DOI: 10.1257/aer.90.5.1362. Accessible record.
- North, Douglass C. (1990). Institutions, Institutional Change and Economic Performance. Chapter 1, An Introduction to Institutions and Institutional Change, pp. 3–10. Cambridge University Press. DOI: 10.1017/CBO9780511808678.003. Accessible record.
- Ostrom, Elinor (2010). Beyond Markets and States: Polycentric Governance of Complex Economic Systems. American Economic Review 100(3): 641–672. DOI: 10.1257/aer.100.3.641. Accessible record.
- Smith, Adam (1776). An Inquiry into the Nature and Causes of the Wealth of Nations. Book I, chapter I, Of the Division of Labor. London: W. Strahan and T. Cadell. Online edition edited by Edwin Cannan, Library of Economics and Liberty. Accessible record.
- Veblen, Thorstein (1898). Why Is Economics Not an Evolutionary Science? Quarterly Journal of Economics 12(4): 373–397. DOI: 10.2307/1882952. Reprinted in The Place of Science in Modern Civilisation and Other Essays (1919), pp. 56–81; Project Gutenberg eBook 39949. Accessible record.