We are pleased that this exchange provides an opportunity to revisit the conclusions of our recent article, ‘The Muddle of the “Westminster Model”’, in this journal (Russell and Serban Reference Russell and Serban2021). Our article was intended to spark a debate about the use of concepts in political science, and of this concept in particular, and the response offered by Matthew Flinders, David Judge, Rod Rhodes and Adrian Vatter (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021, hereinafter ‘the critics’) indicates that it has achieved that goal.
In this ‘response to the response’ we briefly restate our original purpose, methods and findings, before turning to some of the points made by the critics. In all, we conclude that they may have identified some shortcomings in our original piece; but there are many points on which we in fact agree. Their unhappiness seems at least partly due to flaws in the literature, rather than in our exploration of it, meaning that their response risks falling into the trap of ‘shooting the messenger’. Ideally the critics would instead have used the central points of consensus between us as a starting point and explored how our arguments could be developed. In places we also find their critique to be confused and internally contradictory.
We believe that there is broad agreement between ourselves and the critics in terms of the inherent haziness of the ‘Westminster model’ as a concept and hence its unsuitability for use as a determinant of case selection in positivist comparative research. Where the critics may be right is that the term could potentially be rescued in the context of British politics, and of more interpretivist political science. However, in arguing that it may be ‘stretched not snapped’, that case is never clearly or convincingly made. To be convincing, it would require scholars to absorb and accept some of our original points about the unwiseness of the term's use in other contexts.
A summary of our approach and findings
Our study did not seek to justify or advocate for a particular version of the ‘Westminster model’. Instead, guided by the classic piece by Sartori (Reference Sartori1970), it sought to explore how this term has been used as a concept in the recent comparative political science literature. In doing so, it asked three questions: (1) to what extent is the Westminster model explicitly defined by authors? (2) Where the Westminster model is defined, what meaning is it given? And (3) to which political systems is the Westminster model thought to apply?
We set out to research these questions systematically through a study of the recent academic literature. To do so, we searched bibliographic databases for a specific set of terms in works published between 1999 and 2017.Footnote 1 Given that such database searches often overlook books, and that we wanted to explore the common usage of the term in contemporary political science, we conducted further searches to identify core texts recommended on undergraduate and postgraduate reading lists and examined those texts. We read and coded all the items returned by these searches according to whether the term was presented with or without a definition, the extent and nature of any such definition, and any countries with which the ‘model’ was explicitly associated.
We acknowledged from the outset that the idea of the ‘Westminster model’ did not originate within political science, and that it started as ‘a descriptor in British politics’ which ‘developed over time to take on a more widely comparative use’ (Russell and Serban Reference Russell and Serban2021: 745). We also acknowledged that ‘[t]he Westminster model has had different meanings over time, all of which continue to have some vestigial use’ (Russell and Serban Reference Russell and Serban2021: 760). We did not seek in any way to dismiss this heritage of the term, or to deny its continued significance to practitioners. However, similar to other concepts that emerged outside academia but form part of political discourse, the ‘Westminster model’ has been adapted and applied in political science analysis, as its use in the literature demonstrated. We set out to understand this use, and whether it can facilitate the kinds of inferences that scholars deploying the term aim to make. This kind of exercise has been undertaken for other concepts – for example populism (Van Kessel Reference Van Kessel2014), neoliberalism (Venugopal Reference Venugopal2015) or semi-presidentialism (Elgie Reference Elgie2020) – with the same aim of understanding whether a particular concept supports valid and reliable research.
We continued in line with established methodological procedures for this kind of approach: coding by two researchers, with adjudication by a third, and using a mix of deductive and inductive coding to ensure both systematicity and openness to what the collected data could reveal. We did not seek to measure the selected texts against a pre-agreed definition; instead, to reflect current use as closely as possible, we took a definition to be ‘an explicit statement setting out what the author(s) in question considered to be the meaning of the term’ (Russell and Serban Reference Russell and Serban2021: 750). We then reported on our findings, setting out and explaining the variation that we found. In doing so, we did not adopt any normative views on how the term should be defined and were careful not to single out any individual author for using it more or less ‘well’. Our goal was instead to describe its use and to explain observable patterns. Our conclusions were based on careful analysis of an original data set, comprising 239 texts in total.
This analysis found that the extent to which the ‘Westminster model’ is defined is highly variable. Its meaning is often assumed to be implicit or understood by readers. Among the 239 texts, 71 (30%) used the term without any form of definition at all. A further 90 (38%) did not contain an explicit definition – that is, a clear statement of meaning – but did mention some attributes with at least a loose implication that the author(s) considered them relevant to the ‘model’. In contrast, just 78 texts (33%) contained an actual definition. Examination of the comparative texts found that many used the term to imply a sampling or case selection strategy, based on the assumption that the countries in question shared a set of common attributes, even if it was never stated what these attributes were.
Next, we took the subset of 168 texts that included either an explicit definition or implicit association with certain attributes, and looked at what these attributes were – in Sartori's (Reference Sartori1970) sense, to explore the ‘intension’ of the term. This allowed us to investigate the extent to which attributes were used recurringly or consistently across the literature. We found that the model is associated with a wide variety of attributes, and we listed the 18 which occurred most frequently. The most commonly mentioned attribute of a ‘Westminster model’ country was use of the first-past-the-post electoral system, which was mentioned in 28% of texts. Other examples included having a neutral civil service (22%), being a unitary state (14%) and possessing either a unicameral (5%) or a bicameral (4%) parliament.
Third, we looked at the term's ‘extension’ – that is, the countries that authors associated with the ‘Westminster model’. Here there was a clear pattern, with the most frequently mentioned countries being the UK (196 texts), Canada (77), Australia and New Zealand (both 67). Other texts referred to India (13), Trinidad and Tobago (12), Jamaica (11), Guyana and Barbados (both 10) and Ireland (9). There were 35 countries cited in total. Most of these countries had formerly been under British rule, so could be seen as having a Westminster ‘heritage’. But we questioned whether they still share enough commonalities today for the ‘Westminster model’ label to be meaningful.
Crucially, all these countries are parliamentary democracies, but beyond that their political systems differ in important ways. As other authors have previously noted (e.g., Dunleavy Reference Dunleavy2010; Paun Reference Paun2011), countries usually associated with the ‘Westminster model’ have individually developed over time, becoming increasingly diverse. For example, Australia, New Zealand and Ireland do not use the first-past-the-post electoral system, and Canada, Australia and India have federal systems.
Going beyond the Sartorian framework of classical hierarchical concepts, we asked whether, using these countries for reference, the Westminster model instead meets the requirements of a ‘family resemblance’ concept. Adopting a widely used definition by Goertz (Reference Goertz2006), this allows cases to share some but not all of a set of attributes and still form part of the same family. But the objectively observable attributes shared by the most commonly noted ‘Westminster model’ countries are now very few, while such countries tend to share at least as many key attributes with others which are not associated with the model. For example, Italy and Japan are both unitary states with parliamentary, bicameral systems, while the USA uses first-past-the-post in congressional elections. On this basis we concluded that the ‘Westminster model’ does not meet even the relatively loose requirements of a family resemblance concept.
In a comparative sense, we concluded that the ‘Westminster model’ is essentially now an ideal-type for which there are few if any empirical referents. This has important implications for political science. In particular, it indicates that membership of the ‘model’ or the so-called ‘Westminster family’ no longer provides a robust basis for a sampling or case selection strategy that would group together countries with similar political institutions. We discussed whether instead there were cultural connotations of the ‘Westminster model’, based on shared heritage, and indicated that this may well be the case. But, if adopting a case selection strategy in comparative research, or drawing inferences from a single case and applying these to other members of the ‘family’, authors should take care to identify that it is these cultural factors which matter, and carefully set out why.
Our response to the critics
Methods and findings
The critics set out strong objections to our methods and findings, without engaging with them in detail. In places, the critics misrepresented what they had been. Their critique centres on the epistemological assumptions of our approach. The main argument is that a ‘positivist’ analysis of the use of the Westminster model cannot reach valid conclusions, given that the term was never designed for operationalization and measurement; examining whether it has been used with precision is hence based on mistaken premises.
The case against our methodology starts in the following terms: ‘[t]he findings of R&S are a product of their method which focused on “key attributes observed” in selected books, articles and university reading lists’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 5). First, this is inaccurate. Reading lists were used to identify texts, rather than being a direct subject of study. Second, any empirical analysis is by necessity a product of the methods used, and more broadly of its implicit or explicit epistemological assumptions. Any methodology is also, by default, limited in its ability to grasp real-world phenomena; the key question is what it allows us to learn, and what alternative approaches could help to reveal instead. The critics do not suggest alternative methods to survey the literature to understand how the term has been used. If such methods existed, it would have been useful to hear them.
In terms of data and findings, there are also three misleading interpretations in the critics' account of our approach. First, discussing the extent of definitions, they conflate our categories in an inaccurate way. The critics claim that ‘[a]s the basis for an article on the “muddle of the Westminster model”, a finding that 71% of relevant texts did include at least some form of definition, could be seen as a problem’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 3). As summarized above, this is a misrepresentation. Only texts containing an explicit statement of meaning were judged to include a definition – which amounted to only a third in total. We explained that ‘[t]he remaining two-thirds of texts did not provide anything that could be considered a clear statement of meaning’ (Russell and Serban Reference Russell and Serban2021: 750). The critics have chosen to conflate the minority of texts containing a definition with the 38% that lacked such a definition but mentioned the term in any kind of implicit association with certain attributes (examples were provided on our pp. 750–751). Beyond these were the 30% of texts which used the term with no indication, however loose, of even a single attribute associated with the model. For such a significant portion of texts to include not even the slightest indication of the meaning of the term seemed to us both surprising and problematic.
Second, the critics place a very different interpretation on our data concerning the attributes associated with the Westminster model, proposing that ‘[c]ontrary to R&S's core thesis, the main definitional features of the Westminster model appear relatively clear, consistent and understood’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 3). Yet, as indicated in the previous section – and explained on p. 754, and in table 2 of the original article – there is no attribute mentioned by as many as a third of authors in association with the Westminster model, even when limiting the analysis to those texts which give an indication of such attributes. If those texts that failed to identify any attributes at all were included, the most commonly identified attribute – the electoral system – was identified as pertinent by fewer than one in five texts. Altogether, what we found was a collection of relatively disparate attributes, clustered together in highly variable ways. Given the critics' arguments, discussed below, regarding whether the Westminster model is a concept or not, it is doubtful that such precision in definitional attributes should even be expected. But at this point in their article they seem to suggest the reverse – that, notwithstanding our findings, the literature has in fact found conceptual precision regarding a concept that was never meant to be precise.
Third, with reference to the countries associated with the model, the critics suggest that ‘[t]he analysis reveals a relatively tight comparative field that reflects a widely acknowledged colonial heritage. … By focusing on observable, measurable characteristics – on operational variables – R&S underplay both the ideational or cultural relevance of the Westminster model’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 3). Yet, as indicated above, we acknowledged the common ancestry of countries associated with the Westminster model, while noting that their political systems are now dissimilar in important ways. There is no paradox here: countries can clearly share a cultural heritage, based on their origin in British political institutions, and yet deviate over time. On an ideational level, their political elites may also still feel part of a shared ‘Westminster’ tradition, notwithstanding this divergence. But from a comparative research perspective, grouping them together as ‘similar countries’ is no longer justified, unless the study proposes explicitly that ‘shared colonial heritage’ is the key similarity of interest.
One thing that our study did not systematically do, and which could be a useful extension, was to record the frequency of different applications of the term ‘Westminster model’. We set out to explore how the term has been defined in recent political science research by surveying the absence or presence of definitions, as well as their content. As the critics acknowledge at one point, we were ‘mainly seeking to make an argument about the contemporary value of the Westminster model in comparative political science’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 7, original emphasis). But other potential meanings exist, and may be used by different research or policy communities. For example, the term ‘Westminster model’ may be used as a descriptor for the UK political system, or be used by practitioners to signify perceived commonalities between the political systems within which they work. We acknowledged these different uses in our original piece, but did not seek explicitly to classify, quantify, or judge them. Our primary concern – as clearly expressed – was with the frequent use of the term to justify case selection and inferences in positivist comparative research. But there is a worthwhile potential research question here that the critics overlooked.
The Westminster model as a concept
A central argument of the critics is that the Westminster model was not designed as a concept in the positivist sense, and hence its use cannot be evaluated from a positivist standpoint. One problem with this critique is the lack of proposals about the alternative methods that might have been used to evaluate the term's deployment in the literature – as our own methodological approach, based on a partially quantitative content analysis of texts, could be seen as positivist. But the critics appear to conflate arguments about our methodological approach with concerns about the methods in the texts that we analysed. Had this distinction been drawn more clearly, significant points of agreement might actually have been found.
There is a key point on which we find the critics' position to be unclear. This regards whether the ‘model’ can be considered a concept at all, and the level of precision that should be expected of it. Their initial suggestion is that the Westminster model should be considered an ‘organizing perspective’, pointing out that such a perspective is incompatible with concepts, or with being formed of concepts (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 4). There is then a suggestion that, if it is in fact a concept, it constitutes an ‘essentially contested concept’ of the form identified by Gallie (Reference Gallie1956). This is not a claim that we make directly, but is broadly compatible with our approach. However, the critics also suggest that ‘an essentially contested concept is an idea or organising perspective’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 5), which sits awkwardly with their immediately preceding claim that organizing perspectives cannot be concepts. Later on, the Westminster model is described as an ‘ideal type’ or ‘polar type’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 9), or as a ‘legitimizing tradition’, ‘legitimating tool’, ‘legitimating framework’ or ‘legitimizing mythology’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 7). The reader is left confused, including specifically over whether the Westminster model should be treated as a particular type of concept that needs to be evaluated on specific criteria to do it justice, or is a term that was never intended to be a concept in the first place.
Notwithstanding their contention, cited above, that the attributes associated with the Westminster model in the literature are clear, the critics generally put strong emphasis on how the origin of the model makes it inappropriate for use in positivist political science studies. They suggest that, because the term did not originate in academic research, it ‘was not intended to be a neat, precise, specific idea suitable for scientific analysis’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 4), and ‘was never designed or intended to act as a positivist concept facilitating scientific research’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 14). This is a point on which we completely agree – and hoped that we had made clear.
A core purpose of our study was to demonstrate, through systematic collection of data, that many scholars have understood the term to be usable in this way or have tried to adapt it in order to classify and establish equivalence between countries – when it was never suited to being used as such. We suggest that a key turning point in this debate was the widely referenced work of Arendt Lijphart (initially Reference Lijphart1984, updated in Reference Lijphart1999 and Reference Lijphart2012) which made the ‘Westminster model’ synonymous with ‘majoritarian democracy’ and associated this with 10 observable attributes of the political system. Despite the critics' strong objections to the suggestion that the term should be considered ‘a neat, precise, specific idea suitable for scientific analysis’, they are strangely appreciative of Lijphart's treatment.
Clearly, many concepts that form part of both everyday political vocabulary and of social science terminology do not originate within academia and are not the result of scientific design. This is the case, for example, with long-established concepts cited by the critics, such as ‘“the state”, “democracy”, “citizenship”, etc.’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 5), which are all widely used in political science research. Even for concepts that originate in academia, such as semi-presidentialism, scholars have frequently debated how best to define the phenomenon of interest (Elgie Reference Elgie2020). Such analysis is useful, to draw out what would be a minimalist definition to capture the concept's core attributes and distinguish it from others – in this case from presidential or parliamentary regimes. Many other concepts have undoubtedly faced the problem of ‘over-stretching’, whether originating in academic studies or not. Constructing subtypes, or the ‘democracy with adjectives’ approach (Collier and Levitsky Reference Collier and Levitsky1997), can also be an effective way out of definitional conundrums for such concepts.
This search for conceptual clarity is not just instrumental to particular types of political science studies. Even in everyday use, some aspects need to be agreed upon as core elements of ‘democracy’ before a country can be identified as such. Instead of deploying the Westminster model's origins to conclude a priori, in the absence of evidence, whether the term is ‘useful’ for ‘scientific analysis’, we need instead to pose this as a question and set out to find the answer. That is what our study set out to do, asking whether the Westminster model is a concept that can be used, in the ‘positivist’ sense, as a data container – as a tool that allows us to collect data about equivalent entities – given that many authors in political science have clearly treated it in this way.
We uncovered that the term's use in the literature is currently muddled, while the critics wish to recover it. They cite approvingly debates over the precise definition of terms such as consociationalism, but stop short of suggesting that the Westminster model might be subject to such treatment, let alone seeking to join such an exercise – and it remains unclear whether they would consider this useful or not. We would hope that our exposure of the ‘muddle’ in the literature might invite the opening of exactly this kind of debate among scholars keen to preserve the term.
Political science concepts undoubtedly come in different forms. The critics incorrectly claim that we rely exclusively on Sartori's approach, and that we fail to acknowledge other perspectives. We are criticized for ‘dismissing the more expansive and flexible “family resemblance” approach that assumes valid cases will share a set of properties while not all cases share them all’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 9), and told that we ‘shy away from the loose notion of family’. In fact, we devote a substantial part of the article to discussing exactly that topic. While our starting point is Sartori's focus on concepts formed of a hierarchy of attributes – what are referred to as ‘classical’ concepts (Russell and Serban Reference Russell and Serban2021: 748), we do not at any point promote this as ‘the’ approach. Instead we discuss it in parallel with family resemblance (also introduced on p. 748), explicitly suggesting that this might explain the evolution of the Westminster model better than the classical approach.
We discuss the family resemblance approach at greater length in connection to our findings on page 760. Using the definition of family resemblance from Goertz (Reference Goertz2006), we demonstrate (as summarized above) that the term cannot be classified as this type of concept. Goertz himself is cited as an authority by the critics (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 5), but his definition is based on observable attributes and deploys the same ‘formalism and deductive logic’, which they consider so inappropriate in our hands. This inconsistency of treatment is puzzling.
Ultimately, the critics suggest that ‘[t]he scholarly community will … “self-regulate” the Westminster model as a valuable concept [sic.] in a broad analytical toolkit’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 11). No evidence is provided to underpin this apparent certainty, nor examples given of such ‘self-regulation’ having occurred in other cases without deliberate action. It seems to us well established that it is through deliberation and debate that the scholarly community can expose the weaker points of the conceptual tools that are used in empirical research. Elgie's (Reference Elgie2020) piece on semi-presidentialism cited above is just one recent example. Revisiting problematic concepts is not just useful for political science research; it serves the broader purpose of helping us to understand and distinguish between different political phenomena. We therefore sought to move the conversation in that direction – but the critics did not acknowledge the merits of such an approach.
The Westminster model in British and comparative politics
Our article did not explore in a systematic way the various purposes for which the Westminster model is employed in the literature, focusing instead on its varying definitions. We did note (see Table 1) that 67 of our 239 texts were single-country studies focused on the UK, while 82 were single-country studies focused elsewhere, and 90 were more explicitly comparative. However, this is not the same as noting whether the term was used as a descriptor for the UK system, rather than presented as having more comparative reach (notably, some UK-focused texts proposed the latter). This omission could be seen as a shortcoming of our study. However, it is clear from these figures that the great majority of our texts had focus beyond the UK.
Another concern of the critics is that we dismissed the use of the Westminster model as a descriptor for the UK. They somewhat inexplicably devote three pages (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 5–8) to criticizing our treatment of the Westminster model in British politics, considering it ‘something of a surprise to find that the UK, seemingly by default, should be subject to R&S's (Reference Russell and Serban2021: 761) blanket verdict that “it is time for the ‘Westminster model’ to be retired”’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 6). While we do propose the term's retirement, this conclusion was rooted in the clear evidence that we uncovered about the model's comparative use. From the outset this was our central focus. The critics’ statement that the Westminster model's use in British politics is subject to our critique ‘seemingly by default’, and their later suggestion that the muddle is ‘automatically assumed to have afflicted the concept's core territory as well’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 6) are difficult to justify. We never, in fact, discussed the application of the term in UK politics. Only after this three-page discussion do the critics concede that ‘It is important to acknowledge that R&S were mainly seeking to make an argument about the contemporary value of the Westminster model in comparative political science’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 7, original emphasis).
There may be a valid question regarding whether the term ‘Westminster model’ could usefully be retained to describe the British case. (It should be noted here that straightforward literal references to ‘Westminster’ as the British parliament, or similar, were excluded from our sample.) One difficulty, as acknowledged on our page 755, is that the UK system has developed in ways that arguably move it away from the essentials of the ‘model’. This is not solely our point, but one on which we cite claims by previous authors (Hazell Reference Hazell and Hazell2008; Norris Reference Norris2001), including Flinders (Reference Flinders2005, Reference Flinders2009). It is curious then to be accused by the critics of excessive ‘precision’, and of ignoring ‘partial institutional modifications and deviations’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 8), given that we specifically set these out. These words appear alongside a homage to the ‘flexibility’ of Lijphart's association of the model with 10 observable attributes (on the basis that these included three slightly different iterations between editions). Yet Lijphart (Reference Lijphart1984: 19) explicitly acknowledged that Britain did not fit the ‘model’ as neatly as New Zealand, as we pointed out. This provides a very uncertain basis to continue pinning the moniker to the UK.
Turning to our actual focus, of comparative political studies, the critics seem to see our analysis as equally wrongheaded. They claim that, ‘Writing from a comparative perspective, the sheer prevalence of recent analyses that utilise the concept of the “Westminster model” suggests that not all scholars find the term problematic’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 8). This is of course true, but while presented as a critique, it is not in any way in conflict with our position. The opening line of our piece notes that the term ‘appears frequently in both the academic and practitioner literatures’ – indeed, had that not been the case, our study would have been unnecessary, and potentially impossible to conduct. While numerous scholars may use the term uncritically, we sought to explore whether this approach was justified, as prior anecdotal evidence suggested that it was frequently used without adequate care. The fact that an unexamined concept is widely used clearly does not automatically make it unproblematic, as demonstrated well by the already-discussed examinations of various other political science terms.
Our findings indicate that the term ‘Westminster model’ is deployed by different authors in comparative politics to mean different things. Based on careful reading of our 239 texts, we concluded that:
For some authors it is (erroneously) a term interchangeable with parliamentarism – as distinct from what is seen as its polar opposite, US presidentialism. For others it is a special case of parliamentarism, founded on a majoritarian (normally first-past-the-post) electoral system – or indeed shorthand for that electoral system itself. For many public administration scholars, the term instead denotes something about civil service culture, while some others associate it with centralization (notwithstanding the fact that several of the most commonly cited exemplars are federal states). (Russell and Serban Reference Russell and Serban2021: 760)
These varying attributes may not be mutually incompatible (though the association by some authors with bicameralism and by others with unicameralism clearly is), but as our analysis of the term's ‘extension’ shows, countries associated with the model often lack these very attributes.
Importantly, it is often unclear which attribute (or attributes) authors citing the ‘model’ in empirical comparative studies are seeking to suggest is relevant. At worst, it is commonplace for authors to use the term ‘Westminster model’ without any form of definition at all. As discussed at some length in our article, this creates significant problems if used to imply that countries are comparable. As we suggest, ‘comparative politics scholars seem frequently to deploy the term either as a convenient cloak to imply that single-country studies have comparative application, or to suggest that a rational sampling or case selection strategy was used in small-n comparative studies’ (Russell and Serban Reference Russell and Serban2021: 761). This is problematic, potentially leading to false generalizations and invalid inferences.
The critics take issue with our claim about single-country studies, suggesting that this ‘risks painting an inaccurate picture of the existing research base’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 10). We did not of course imply that this was what all studies do (and indeed explicitly referred to other uses), but we did find the practice to be commonplace. The critics seek to refute this while offering no empirical evidence. We have been careful not to single out particular scholars for criticism – but our piece cites several examples of this kind of use, and additional examples are cited in the critics' response.
The critics likewise claim that ‘In major medium-/large-N studies, the “Westminster model” is not used as a sampling strategy on its own’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 10). They again do not offer any empirical basis for this claim. We certainly encountered such studies in our reading of the literature. Indeed, an additional piece cited by the critics in their response (So Reference So2018) further bears this out.
Taking issue with our approach, the critics suggest that, ‘Depending on the (institutional) trait that is of interest, the Westminster model is … not an arbitrary, but indeed a valuable “purposive” sampling strategy’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 10). An example, they suggest, comes from their own work, based on the common cultural heritage of ‘Westminster’ countries: ‘It is precisely that common ancestry that Rhodes et al. (Reference Rhodes, Wanna and Weller2009: 230) take as their rationale for comparing Westminster’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 12). This is entirely consistent with our recommendations, and there is no criticism – real or implied – of this work in our piece. Our conclusion states:
The strongest argument for common bonds between ‘Westminster model’ countries is one based on heritage, and hence possibly on culture. Yet continued emphasis on this ‘family’ risks trapping certain countries, whose systems now radically differ, in a colonial past. Scholars wanting to compare systems with a shared British heritage should make this attribute explicit, and justify why it is an appropriate one to include within their sampling or case selection strategy. (Russell and Serban Reference Russell and Serban2021: 761)
The circumstances in which such colonial heritage will be relevant in comparative research are likely to be few. The problem that we identified concerns the greater number of comparative studies where there is an implication of shared observable attributes, without these being either made explicit by the author or necessarily even being applicable to the countries concerned. The critics concede the need for authors to specify their attribute of interest, commenting elsewhere that in small-n comparative studies, ‘cases may be selected on joint colonial heritage’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 10); yet of the five recent examples that they cite in support of their argument, only one (O'Brien Reference O'Brien2019) explicitly provides any such justification for case selection.
A final concern here is semantic. At various points, the notion of ‘sampling’ is used by the critics to denote the selection of countries for comparison in small-, medium- and large-N studies. This risks confusion. The logic of case selection is different from that of sampling. Case selection in social science research involves answering the question ‘what is this a case of?’ and, more specifically, ‘how can we classify a unit as part of a broader set of units that share similar features?’, thereby establishing equivalence as a basis for comparison. ‘Cases’ do not pre-exist in the social world ready to be ‘sampled’. The process of ‘casing’ (Ragin Reference Ragin, Byrne and Ragin2009) requires the researcher to use concepts to carve out parts of the social world – for example to classify an instance of people casting ballots on a particular day as either a ‘case of’ an election or a case of a referendum, depending on other particularities.
This is where our argument comes in – badly specified case selection leads to faulty inferences. It does not allow us to extrapolate correctly from the cases selected to a universe of cases. Definition and conceptual clarity are key for such research, where the goal is to draw conclusions about a broader set of similar units (Gerring Reference Gerring2004). This is often the root of difficulty if assuming in comparative research conducted in a positivist framework that there is a common understanding of the ‘Westminster model’.
Interpretive and real-world perspectives
In a further denunciation of what they identify as a ‘narrow’ positivist approach, the critics imply that we were dismissive, or perhaps even unaware, of interpretivist methodology. Yet our article explicitly acknowledged the important contribution of the interpretivist strand of research on the Westminster model, focusing on its cultural and historical significance (Russell and Serban Reference Russell and Serban2021: 747). While our study may not have systematically explored how the Westminster model is used in political science (rather than how it is defined), some portion – albeit undoubtedly a minority – of our sampled texts took an interpretivist approach.
The critics emphasize that ‘[v]ariables, models and generalization are not the only game in town’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 13). This is an accepted fact in the political science community, just as it is recognized that positivist and interpretivist methodologies have different aims and rest on different assumptions about the social world. Some research questions may be better suited to study through one approach rather than the other; but pitting them against each other is unhelpful and restricts scholarly debate and collaboration.
Small-n comparative research, in particular, has recently been enriched by the development of an interpretivist case study methodology. There is a growing literature on interpretivist or contextualized comparisons – comparative case study research that is context- and meaning-focused, rather than variable-focused (cf. Della Porta Reference Della Porta, Della Porta and Keating2008; and see Abramson and Gong Reference Abramson and Gong2020; Burawoy Reference Burawoy2009; Simmons Reference Simmons2016; Simmons et al. Reference Simmons, Smith and Schwartz2018). This recent literature has gone to great lengths to think outside the traditional Mill's methods of comparison. Such an approach offers valuable insights into meaning-making processes and helps to uncover and explore bottom-up, everyday conceptualizations, practices and understandings. The interpretivist approach can therefore be useful in exploring how the Westminster model is understood.
Connectedly, the critics suggest that ‘the Westminster model “remains the dominant narrative” among central actors in Whitehall and Westminster (Rhodes Reference Rhodes2011: 306)’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 7) and that we failed to acknowledge the importance of the Westminster model to practitioners, for whom it encapsulates the features and conventions of the political systems in which they operate. Yet we start our article by observing that ‘[t]he term “Westminster model” appears frequently in both the academic and practitioner literatures’ (Russell and Serban Reference Russell and Serban2021: 744) and express a concern that ‘[p]ractitioners … may assume that a term frequently used by political academics has a solid basis’ (Russell and Serban Reference Russell and Serban2021: 761). If practitioners share an understanding of what the Westminster model means, and rely on that in their decision-making, it is of course important to explore their experiences, as well as the potential consequences of those shared understandings. But that is not what our study set out to do.
Instead, the key point emphasized in our article was that political scientists should be mindful not to perpetuate the idea that ‘Westminster model’ countries share similarities beyond their common political origin. While practitioners may be an important subject of study in understanding how framing via a ‘Westminster model’ can shape real-world beliefs and actions, they are also recipients and users of research. In this context, political scientists conducting comparative studies have a responsibility to define their terms clearly and not to lead practitioners into false conclusions about how findings in one system may or may not play out in their own.
Conclusion
In all, we find the critics' response to our original article disappointing. While valuing the ‘Westminster model’, and acknowledging its multiple and complex meanings, they essentially dismiss our exploration of its use, which was based on systematic and detailed study of over 200 texts. In places their piece misrepresents our methods and findings, while offering no alternative suggestions for how questions about the multiple meanings of the ‘Westminster model’ could better be explored. The negative tone of their piece is in stark contrast to the approach that we adopted, which pointed out the contradictions in the literature, while taking care not to criticize or deride any individual author. In particular our piece included no criticism – real or implied – of the critics' prior work.
The truth unacknowledged by the critics is that there are important points of agreement between us. Their own representation of the Westminster model displays significant inconsistencies – is it a concept or an ‘organizing perspective’? Can it appropriately be characterized by a set of observable attributes or not? But a central concern of the critics seems to be that it is not a term suited to precise treatment in positivist empirical studies – which was exactly our central point. A recognition of this fundamental element of agreement would allow for more constructive engagement with our findings and with how these could be refined or improved. In some respects, the critics seem to have conflated the methods that we adopted in our study with the findings that we uncovered. Their instinctive hostility to the treatment of the Westminster model as ‘a positivist concept facilitating scientific research’ (Flinders et al. Reference Flinders, Judge, Rhodes and Vatter2021: 14) has been thrown at us as an accusation, when actually this was the very problem that we sought to highlight.
The critics' analysis does give pause for thought on three specific points – none of which they made directly. First, an analysis of how the Westminster model has been used in the literature, not simply how it has been defined, could be a useful extension of our work. Second, and connectedly, while the appropriateness of the Westminster model's continued applicability in contemporary British politics has already been explored (e.g. Flinders Reference Flinders2005; Hazell Reference Hazell and Hazell2008; Norris Reference Norris2001), there may be a valid question about whether it could and should be resuscitated for this purpose, even if unsuited to comparative empirical research. Third, carefully conducted exploration of the continuing influence of the ‘Westminster model’ in an interpretive framework remains potentially useful. Our call for its complete ‘retirement’ could therefore validly be argued to be over-expansive. Nonetheless, on both of these latter points we have concerns – as it is doubtful whether the use of a term can be ‘ring-fenced’ or policed in this way, and indeed the critics never explicitly suggest that it should be. The history of the Westminster model shows that it has leached from relatively limited and perfectly appropriate applications to a broader and more problematic range of uses.
It is hard to conclude, on the basis of the critics' response, that there is not ‘muddle’ about the meaning of the Westminster model. Indeed, by arguing on multiple fronts that this is not the case, they seem inadvertently to have demonstrated our point. We agree with their analysis that this term, with its long historical roots in British politics and export to former British colonies along with Westminster-inspired political institutions, is ill-defined, constantly changing, and cultural as much as it is political. We agree that this is why it is completely unsuited to use as a variable or a case-selection strategy in positivist comparative research and are more concerned than the critics about the – no doubt unintended – consequences of the term's use by Lijphart (Reference Lijphart1984, Reference Lijphart1999, Reference Lijphart2012) in encouraging this trend. It is a shame that the critics preferred to seek conflict, rather than recognizing that on these central points we were actually in agreement.
