Measurement Makes Communicating Political Science Hard
Imagine you are a political practitioner for a social democratic party. Perhaps a member of parliament, or a staffer working on communications, or a local organiser. A question that will be weighing on your mind at this moment is how to deal with the ascendancy of the radical right. It could be Reform in the UK, or the AfD in Germany, or one of countless other ascendent parties of the far right. What might your strategy for winning against this new threat be? In formulating your strategy for dealing with this new political reality, you might consider many sources of inspiration. Pieces of folk political wisdom. Focus groups and opinion polls. Examples of success from abroad. What I suspect many acutal political practitioners wouldn’t consider doing, however, is to consult the academic literature on party strategy.
I The Rise of the Radical Right
Radical right parties have at times enjoyed seemingly sudden success. The underlying social trends behind these successes have however been anything but sudden. The rise of these parties is closely tied with the rise in importance of what political scientists call ‘second dimension’ issues. The implicit ‘first dimension’ in this formulation represents the economic issues which traditionally structured party competition. Simplistically, the left typically favours economic redistribution and larger welfare states, the right typically prefers less taxation and smaller states. Second dimension issues are then those non-economic issues increasingly prominent in our politics now. Crucial for the successes of the radical right is the rise in prominence of cultural issues such as immigration and EU membership1.
The earliest research on second dimension issues suggested that they would create a ‘dilemma’ for social democratic parties. The theory was that the electorates of social democratic parties were made up of a combination of socially conservative working class and socially liberal university graduates. As social and cultural issues rise in their importance, it was expected that this would drive a wedge between these groups of voters. The dilemma social democratic parties were expected to face was a choice between the two halves of their electorate2.
More recent research shows that, considered in terms of strategy, no dilemma exists for social democratic parties. Social democratic parties optimise their electoral performance by taking socially and culturally liberal positions. The most conservative findings suggest only that they do not suffer electorally for taking liberal positions3. If you are a social democratic political practitioner, the extant political scientific evidence strongly plainly suggests that social democratic parties ought to take socially and culturally liberal positions. In the worst case, they do not harm the party. In the best case, they aid it.
The conclusions of this body of work fit well with other research on the radical right and on issue ownership. Radical right parties are typically issue owners on immigration and on Euroscepticism. Where a party is seen to have ownership of an issue (or a particular stance on that issue), there is no reason for voters to choose a different party with the same stance if that issue is the main issue driving their choice. The radical right are issue owners on both immigration and Euroscepticism. Where these issues become more prominent in politics, this will largely serve only to aid them4.
Political folk wisdom typically emphasises moving towards the political centre, or towards one’s opponents, in order to win. But, if we accept the issue ownership model, moving towards the policy positions of the radical right - an ‘accomodative’ strategy - to appease their voters is unlikely to be an effective strategy. This is partly because, in line with the above, adopting their policies and rhetoric simply increases the prominence of their beliefs, of which they will be the beneficiaries. It is also in part because it serves to further normalise and legitimise their beliefs. And there is evidence to support this: work on accomodative strategies shows that they do nothing to harm the performance of far-right parties, and may even harm the party adopting those positions5.
II Why Ignore Political Science?
Imagine you are a political practitioner for a social democratic party. You’ve just read this summary of some of the most relevant evidence for your radical right problem in the academic literature. Perhaps you’ve read many of the papers cited in the footnotes, and many more not cited. What might you change about the strategy you previously had in mind? If you had thought about adopting more stringent immigration policy, or a more Eurosceptic policy, you might now have second thoughts. You know that if you accomodate the policies and rhetoric of the far right, at best you have adopted policies that are likely to go against your own principles and beliefs. At worst, you have helped your enemy, by increasing the prominence of the issues they profit from.
Political practitioners have typically not changed their strategies in this way. Across many countries, far-right success in fact tends to lead to accomodation by mainstream parties of all varieties6. The evidence in the literature argues against accomodative strategies as being potentially harmful, yet political practitioners consistently make the same decisions over and over again. It follows that it is either the case that either they do not read this literature, or they reject its arguments, evidence, and conclusions, or they face some other constraints preventing them from adopting its recommendations7.
It is not a difficult exercise to begin thinking of some reasons why the literature has been ignored or discounted. One reason is that much of the academic literature is sitting behind paywalls, difficult to access. Another is its sheer volume: unless you know the crucial papers and authors, there is a lot to sift through. It is difficult to read work one doesn’t know of. Yet another is that there is no reason to assume political scientists might be good at political practice. Consider the example of Michael Ignatieff8, who left an academic career in the US to lead the Canadian Liberal Party - one of the main two parties in a two-party system - into third place. Would you place your hopes on someone whose chosen career path incentivises a very different skillset and perspective?
I suspect another reason, based on the literature discussed above, is that it is hard to accept that sometimes the best strategy is to do nothing different. To be strategically optimal is not to guarantee election victory, or to guarantee that the rise of the radical right can be prevented. Sometimes minimising how badly you lose is the best possible outcome. But political practitioners are constantly exposed to criticism, constantly questioned, constantly seeking to defend themselves or those they work for. Losing after changing nothing looks worse and is treated worse than losing after doing something. I sometimes worry that social democratic practitioners fall into a fallacy of informal reasoning: “we must do something, this is something, therefore we must do it”.
These other explanations as to why the evidence literature is not followed notwithstanding, I will propose another I have not come across elsewhere. This explanation is that, as social science, poltical science is dependent on measurements of core concepts which are pragmatic in their construction and highly abstracted in their nature. This simple fact places profound limits on the extent to which much of the literature I’ve discussed above is directly actionable.
III Measurement In Political Science
Measurement can be understood as a process of going from observed data to quantities representing concepts of interest9. It is useful to think in terms of two measurement paradigms, which conceptualise the way in which this is done in different ways.
In the representational paradigm, the concept of interest exists in the real world and has a causal relationship with the observed data. The thing being measured causes its measure. Consider measurements of length. If you measure something in cm, then the number you get out is a function of the length-attribute of thing you are measuring in the real world. Importantly, representational measures need to be isomorphic: they must preserve the relationships of attributes in the real world. Something that is 2cm long should be twice as a long as something that is 1cm long. Quantification in this paradigm preserves information10. This is the representation in representational measurement.
In the pragmatic paradigm, the concept of interest is defined or even created by its measurement. Measurements in this paradigm are refined by merit of how useful they are. This is a much more common approach in the social sciences. Consider psychological concepts, such as pain. I can find out whether you are in pain by asking you. But different people’s perception of pain can vary wildly, resulting in fairly noisy, or even biased, measures. As a result substantial amount of work goes into guaranteeing the quality of pragmatic measures. It isn’t that there is not correspondence with the real world per se, but isomorphism is not the goal. Usefully capturing human concepts is.
Political science is no different to any other social science. The discipline abounds with concepts such as democracy, ideology, or nation. These concepts in many ways reflect things in the real world. But they are properly understood as human concepts which useful describe and simplify the real world in ways that are useful. Is liberal democracy a binary, where states either are or aren’t this regime type? Or is it a continuum, where states vary in the extent to which they are liberal democracies? Different measures are based on different assumptions and conceptualisations, and can be useful for different pieces of work.
When working with party strategy, we are often thinking in terms of the ‘spatial model’ of vote choice. If you’ve ever uttered a sentence to the effect of ‘I vote for the party closest to my views’, you’ve already grasped the intuition of this model. Imagine ideology as a continuous spectrum, going from left to right. Parties and voters are imagined to sit on this spectrum, or ‘dimension’, and have some sort of distance from one another. Voters prefer parties ‘closer’ to themselves on this dimension. We can introduce other dimensions, while retaining a fairly low-dimensional representation of ideology and ideological distance.
This is a useful model for generating knowledge about politics. It works well in part because there’s good evidence to suggest that ideology in the mass public can be accurately represented in a low-dimensional space11. What this means is that we can usefully abstract away from specific policy attitudes (if voters hold these at all) to higher, abstract dimensions like the left-right and cultural issues discussed earlier.
A typical way of measuring public ideology might be this. Years of research have defined the issues that we think belong to say, economic left-right or to the EU. We might ask a single question, where a respondent places themselves on a 0-10 scale. Or we might ask a series of agree-disagree questions, each relating to an aspect of the economic left-right dimension, then aggregate them. In either case we will probably go through several rounds of refinement to identify and remove questions capturing different forms of survey response bias. It is impossible to get around the pragmatic nature of this: to a non-trivial extent, the researcher writing the survey defines to some extent the concept being captured.
One way of reducing this dependence on the fact we only observe answers to a very small subset of the possible questions which could be meaningfully asked is the use of scaling methods. Not all political scientists use these, but I am personally a fan of them. When I aggregate agree-disagree items, I am essentially deciding to place the same weight on all of them - to treat each survey item as equally important to and indicative of left-right. In a scaling model, different items are weighted to ‘reveal’ a ‘latent’ dimension, which is then interpreted by the researcher. The dependence does not go away: but it allows the dimension to be defined in part by the pattern of responses in the data.
I will use an example from my own work12. The plot below is from a paper I published last year, on whether the Labour party could have improved its performance inthe 2019 UK general election by changing its EU position:
To construct this plot, I used a scaling method which scales survey respondent self-placements against their placements of political parties on the same scale13. I like this method because it provides estimates of the ideological positions of both voters and parties at the same time, on the same scale. This allows for the creation of meaningful distance measures.
I used a scaled measure of positions on the EU, and created a distance between each voter and party. I then estimated a statistical model of vote choice. This model allows for a simple14 simulation, where by changing the Labour party’s EU position and re-computing its distances from voters, the model can be used to generate new predictions and thus new estimates of how the parties would have done.
The x-axis shows the simulated Labour party position with negative values representing pro-EU and positive values representing anti-EU attitudes15, while the y-axis shows estimated party vote share. The curved, somewhat horizontal lines show the estimated vote shares of different parties, and are coloured according to the party colours (except the Brexit party, which is coloured purple to better differentiate from the Conservatives). The black vertical line shows the ‘true’ Labour party position from the election, while the red vertical line shows the vote-maximising position and the blue vertical line shows the position which minimises the gap between Labour and the Conservatives’ vote shares.
What the plot shows, and what my paper as a whole suggested, is that the Labour Party’s (pro-EU) position was broadly its best in terms of maximising votes. I think this is a good example of a pragmatic measure being used to reach a useful conclusion.
IV Back To The Problem
Imagine you are a political practitioner for a social democratic party. Imagine I have just presented this plot to you, with some time to go before the upcoming election. Imagine that the evidence suggests that Labour’s best EU position is actually at about -1, rather than the current position of about -0.4. What sort of questions might you ask? I think it’s probable that the question on your mind would be what you should be doing to make a -0.6 move along the x-axis.
The answer I would have to give you is that I have no idea. The measure I’ve used is a pragmatic measure of a concept at a high level of abstraction. There are ways to try and make some sense of it. Given the method I’ve used, I could look at the estimated positions of other parties, such as the Liberal Democrats or Greens, to see where they sat. In the paper, this was roughly at -1, so I suppose I could in this case say that Labour ought to be as pro-EU as these parties. But imagine that the optimal point had been at -0.8, or -0.2, where no party sits.
I could try and discretely divide up the space. Say, decide that points between -0.2 and 0.2 are ‘neutral’. But there’s a lot of loss of detail, and Brexit was a fairly binary issue: -0.2 in different in important ways to 0.2. And it is difficult to relate this pragmatic, abstract measure to specific policy decisions. Soft Brexit or hard? Stay in the single market? In the customs union? How should this be framed in relation to the referendum? The scale does not give answers because it is not designed to do so. It is not isomorphic to combinations of policy decisions. It captures these things, and usefully so - just like psychological measures of personality like the Big 5 are pragmatic yet have a relationship with the real world. But like these psychological measures, it does so in a pragmatic, not representational way. This is not because I have scaled the data: everything I have just said would be true of the raw survey data used to produce the scaled data.
Imagine you are a political practitioner for a social democratic party. Every single day, your party is being attacked as too liberal, too woke, too whatever else by a print media predominantly owned by rich right-wing moguls all too willing to amplify the talking points of the radical right. You have tens of decisions to make every day: where to invest campaign money, what messaging to use, who should be giving the message, the framing to use. You need to decide on policies based on a combination of evidence as to their impact, your own ideology, constraints on public spending, constraints such as potential political backlash, and constraints from public opinion. Every single day, the candidate you are trying to get elected is being attacked. You are probably working overtime, on weekends, on holidays, and not being paid for it, all with the upcoming election at the top of your mind. You are probably very stressed.
A political scientist presents to you a paper that might look like mine, and presents evidence a very high level about abstract concepts. Having assured you that, actually, the current shitshow is the best of all worlds, they cannot give you a single piece of concrete advice. They can’t tell you what policies to adopt, or what messages to use, or how to frame their upcoming social media campaign. They can’t tell you how to handle the print media, and make your candidate more likable. Perhaps worst of all, they might even suggest that the current shitshow is the best case scenario for you. They tell you not to change course, not to try and tackle the far right by appealing to the voters who support them.
Imagine you are a political practitioner for a social democratic party. How would you react?
V Coda: A Solution?
I want to end by reflecting on how I think findings like mine and others cited could, indeed in my view should inform political practice16. It’s worth starting from with a research method which has been heavily developed on by political scientists, yet is widely used by political practitioners. I have in mind the humble survey experiment. Survey experiments have been widely used in political campaigning in order to select the optimal messaging for persuading voters17.
The reason I think this method has been so widespread in its use, beyond providing clear causal estimates, is you can focus on quantities directly relatable to the real world. The outcome variable will be similar to mine - an estimate of which party voters will choose. But your independent variable will be something tangible, like discrete combinations of messages, or framings, or speakers, or whatever else you’re trying to work out. Practitioners can design experiments to directly use the messages they might use, and so learn which messages are actually likely to achieve their desired outcomes. And the evidence in the paper cited above is clear: this sort of thing is highly context-dependent.
Imagine you are a political practitioner for a social democratic party. What messages would you consider spending money on in an experiment, as you start charting a course to beating the far right? Resources are not infinite: you can only consider so many options, and so many variations of the same options. You might do well to start with the political scientific literature on the far right and on social democratic party strategy.
My closing claim is that it is precisely the sort of abstract, general research based on pragmatic measures I’ve discussed above that could inform the experiment you design. Instead of investing resources in testing anti-immigration messages, which you now know will at best not help you and at worst backfire, you can begin searching for messaging that reduce the salience of immigration to voters. You could, for instance, test whether voters are less likely to list immigration as their most important issue by trying to make different combinations of your economic policies prominent on an ad or leaflet.
For practitioners, this is where the value of the political science literature is to be found. Abstract, pragmatic measures cannot be related directly to your decisions because they are not isomorphic to things like policy decisions and messaging strategies. But they can inform your overall strategy by enabling you to eliminate strategic options you might otherwise have followed, or have invested resources into considering. Abstraction and theory are useful things, and I think this is true for practitioners too.
For a general overview of the ‘second dimension(s)’ of European politics:
Bornischer, S. (2010) The New Cultural Divide and the Two-Dimensional Political Space in Western Europe, West European Politics 33 (3). DOI: https://doi.org/10.1080/01402381003654387
Ford, R. and Jennings, W. (2020) The changing cleavage politics of Western Europe, Annual Review of Political Science 23. DOI: https://doi.org/10.1146/annurev-polisci-052217-104957
Kriesi, H. (2010) Restructuration of Partisan Politics and the Emergence of a New Cleavage Based on Values, West European Politics 33 (3). DOI: https://doi.org/10.1080/01402381003654726
See also Kitschelt (1994) below in footnote 2, or much of the work of Inglehart from the 1970s onwards on ‘materialist vs postmaterialist value priorities’ for earlier versions of this.
On the relationship of the rise new ‘political dimensions’ with the rise of the far right (some more directly related than others):
Bornischer, S. (2010) - as above
Green-Pederson (2019) The Reshaping of West European Party Politics: Agenda-Setting and Party Competition in Comparative Perspective. DOI: https://doi.org/10.1093/oso/9780198842897.001.0001
Hobolt, S. and de Vries, C.E. (2015) Issue Entrepreneurship and Multiparty Competition, Comparative Political Studies 48 (9). DOI: https://doi.org/10.1177/0010414015575030
Hooghe, L., Marks, G. and Wilson, C.J. (2002) Does Left/Right Structure Party Positions on European Integration?, Comparative Political Studies 35 (8). DOI: https://doi.org/10.1177/001041402236310
See Kitshelt (1994) in the list below for the (to my knowledge) original statement of a ‘dilemma’ facing social democratic parties. Kitschelt and Rehm (2014) discuss working class social conservatism and the decline of the blue-collar working class, while Gingrich and Häusermann (2015) argue that the decline of the working-class vote has left social democratic parties dependent on highly educated voters.
Gingrich, J. and Häusermann, S. (2015) The decline of the working-class vote, the reconfiguration of the welfare support coalition and consequences for the welfare state, Journal Of European Social Policy 25 (1). DOI: https://doi.org/10.1177/0958928714556970
Kitschelt, H. (1994) The Transformation of European Social Democracy. DOI: https://doi.org/10.1017/CBO9780511622014
Kitchselt, H. and Rehm, P. (2014) Occupations as a Site of Political Preference Formation, Comparative Political Studies 47 (2). DOI: https://doi.org/10.1177/0010414013516066
Abou-Chadi and Wagner (2020) find no improvement for social democratic parties that take anti-EU positions (and that, if anything, these positions are harmful to these parties’ electoral success). Abou-Chadi and Wagner (2019) meanwhile find evidence to suggest that social democratic parties do best when adopting a combination of pro-investment economic stances and culturally liberal stances. My own work shows that in the 2019 UK General Election (where EU membership/Brexit was the main issue), the Labour Party maximised its votes by being pro-EU. Evidence on seat maximisation was somewhat mixed and so I avoid conclusions, but the balance of evidence was towards being pro-EU there too. I include Wagner (2021) because it’s worth noting that the same does not apply to far-left parties, as the evidence suggests they best optimise their performance by taking Eurosceptic positions.
Abou-Chadi, T. and Wagner, M. (2019) The Electoral Appeal of Party Strategies in Postindustrial Societies: When Can the Mainstream Left Succeed?, The Journal of Politics 81 (4). DOI: https://doi.org/10.1086/704436
Abou-Chadi, T. and Wagner, M. (2020) Electoral fortunes of social democratic parties: do second dimension positions matter?, Journal of European Public Policy 27 (2). DOI: https://doi.org/10.1080/13501763.2019.1701532
Swatton, P. (2023) Social democratic party positions on the EU: The case of Brexit, Party Politics 30 (5). DOI: https://doi.org/10.1177/13540688231177557
Wagner, S. (2021) Euroscepticism as a radical left party strategy for success, Party Politics 28 (6). DOI: https://doi.org/10.1177/13540688211038917
Meguid (2008) is the classic study of niche party success via issue ownership and issue salience. Another classic Hobolt and de Vries (2015) in footnote 1 on issue entrepreneuership as another important study on this point. In line with models of issue ownership, Vasilopoulou and Zur (2024) use simulations to show that far-right parties do best in Western Europe when the salience of EU issues is high.
Meguid, B. (2008) Party Competition between Unequals. DOI: https://doi.org/10.1017/CBO9780511510298
Vasilopoulou, S. and Zur, R. (2024) Electoral Competition, the EU Issue and Far-right Success in Western Europe, Political Behaviour 46. DOI: https://doi.org/10.1007/s11109-022-09841-y
Krause et al (2022) show that accomodative strategies (moving towards the far right parties’ core positions) do nothing to reduce their support, and may even be harmful to other parties. Abou-Chadi et al (2022) shows the same in the specific case of centre-right parties.
Abou-Chadi, T., Cohen, D. and Wagner, M. (2022) The centre-right versus the radical right: the role of migration issues and economic grievances, Journal and Ethnic Migration Studies 48 (2). DOI: https://doi.org/10.1080/1369183X.2020.1853903
Krause, W., Cohen, D. and Abou-Chadi, T. (2022) Does accommodation work? Mainstream party strategies and the success of radical right parties, Political Science Research and Methods 11 (1). DOI: https://doi.org/10.1017/psrm.2022.8
Abou-Chadi, T. (2016) Niche Party Success and Mainstream Party Policy Shifts – How Green and Radical Right Parties Differ in Their Impact, British Journal of Political Science 46 (2). DOI: https://doi.org/10.1017/S0007123414000155
Abou-Chadi, T. and Kreuse, W. (2018) The Causal Effect of Radical Right Success on Mainstream Parties’ Policy Positions: A Regression Discontinuity Approach, British Journal of Political Science 50 (3). DOI: https://doi.org/10.1017/S0007123418000029
Decisions going against the direction academic literature suggests is best is perhaps true of many other domains. Consider any number of policy-making areas with rich bodies of academic literature. But in politics sometimes seemingly sub-optimal policy will be made, typically because optimising for one desired outcome can entail trading off against other desired outcomes.
For instance: these days it is easy enough to come across a substantial amount of evidence in favour of changing the UK’s planning regime so as to enable making house-building easier. But if sufficient amounts of house building will reduce house prices, this also goes against the interests of homeowners: both those paying their mortgages (whose houses are now worth less than their debt and so can no longer move) and those who’ve paid off their mortgages (so as to, for instance, maximise the value of the inheritance they can bequeath to their loved ones). If politicians do not follow the advice of economists arguing for house building, it is not necessarily evidence that they have not read or listened to that advice, but rather because their preferences lead to a particular trade-off being favoured.
Part of my assumption here is evidence on electoral strategy and outcomes is, at least a lot of the time, different in nature: if an academic paper claims to show something to the effect of “taking action x will result in your party losing, and/or your opponent(s) winning/being in a stronger position”, then taking action x is necessarily a sign that either the paper was not seen, or that its argument or methodology was regarded as incorrect, because there was no trade-off evidenced in strategic terms.
One way in which my assumption will be incorrect is the trade-off between a politician’s preferences and optimal strategy. Another is where other strategic constraints prevent an optimal policy being pursued, such as a party where party members have greater control over policy and wish to pursue a non-optimal direction. A third situation is where winning was not the goal of a political party, which at times is not true, such as when a party is attempting to normalise their policy positions for future elections. For this last point, see:
Izzo, F. (2023) Ideology for the Future, American Political Science Review 117 (3). DOI: https://doi.org/10.1017/S0003055422000843
But it is difficult to imagine the first and third of these applying. I do not suspect that very many social democratic politicians willingly adopt far-right positions and rhetoric on immigration, but rather do so for strategic reasons. The bitter irony of this is that on the current evidence the decision likely undermines their own goals. And it is hard to believe that social democratic parties adopting anti-immigration or Eurosceptic stances do so with the goal of normalising those stances for the future, rather than winning in the here and now. The second possible violation of my assumption - constraints on decision-making - strikes me as being much more plausible.
Though I am not sure if he would accept the label ‘political scientist’, the example still broadly illustrates that expertise about politics isn’t expertise in doing politics.
For introductions to the theory of measurement, I would recommend one or both of Hand (1996) and Hand (2016). The latter is the more appropriate treatment for general readers, the former more appropriate for statistical practitioners (though these should probably read both). I also strongly recommend for social scientists Lauderdale’s (n.d.) work in progress textbook, which was used in a class I was a teaching assistant for and has really shaped my own thinking on measurement, and from which I have taken the causal diagrams for both measurement approaches.
Hand, D.J. (1996) Statistics and the Theory of Measurement, Journal of the Royal Statistical Society. DOI: https://doi.org/10.2307/2983326
Hand, D.J. (2016) Measurement: A Very Short Introduction. ISBN: 9780198779568. URL: https://global.oup.com/academic/product/measurement-9780198779568
Lauderdale, B. (n.d.) Pragmatic Social Measurement, DRAFT edition 2022-07-18. URL: https://benjaminlauderdale.net/files/papers/pragmatic-social-measurement.pdf
Stevens, S.S. (1946) On the Theory of Scales of Measurement, Science 103.
Note that different levels of measurement preserve different qualities. In general:
Nominal measures only allow for equality checks (x is the same as y). They preserve neither order nor distance. An example might be labels for colour: red, blue, yellow, etc.
Ordinal measures preserve order, and so allow for greater than (x > y) or less than (x < y) checks in addition to equality. They do not however preserve any distance relationships. An example might be size categories: small, medium, large.
Interval measures preserve both order and distance. This means that addition and subtraction become meaningful operations (e.g. x-y = z, z is the difference between x and y). An example might be degrees Celsius as a temperature measurement (see more on this on the next point).
Ratio measures preserve both order and distance, and additionally have a meaningful 0 point. This means that multiplication and division become meaningful operations. An example might be Kelvins as a temperature measurement.
To understand the difference between interval and ratio measures, consider that an increase of 1 Kelvin is the same change in temperature as an increase of 1 degree Celsius. However, Kelvins begin at absolute 0 - you cannot have less Kelvins. Degrees Celsius by contrast begin at -273.15°. But this is arbitary - the number still measures absolute 0, which means a complete absence of temperature. The shift was made largely to make Celsius a more useful measure for day-to-day life.
If I go from 1 Kevlin to 2 Kelvin, I have a 2/1 = times 2 increase in temperature, which is isomorphic to the increase in the energy state of the matter in question. If I go from 1° Celsius to 2° Celsius, I do not have a times 2 increase, but a 275.15/274.15 ≈ times 1.0036 increase in temperature. Note the need to convert back to Kelvins to compute this: I cannot apply the same division to degrees Celsius while capturing the proportional increase in temperature in an isomorphic manner. This is what differentiates ratio from interval measures.
The classic on this is Stevens (1946), in the footnote above. Lauderdale (n.d.) gives a nice introduction to levels of measurement at the beginning of chapter 3. And if I recall correctly, the temperature example is also in Hand (2016).
Downs (1957) is the classical text for the spatial model of ideology. Be careful: there is both a paper and a book from him with very similar names, making very similar points. Converse (1964) is the classic on the idea that ideology is ‘constrained’, which is to say certain things tend to ‘go together’ - though he questioned the proportion of the mass public for which this was true. Ordeshook (1976) coined the term ‘basic space’ which is my favoured one for the low-dimensional hypothesis, while Peffley (1985) developed a nice theoretical model of hiearchcal ideology, from more general to more specific. Hare (2024) provides to my mind one of the best tests of the low-dimensional hypothesis, albeit in the US setting. I tend to take the emphasis on ‘second dimension’ ideology in the European setting as evidence for the hypothesis holding in Europe.
Converse (1964) The Nature of Belief Systems in Mass Publics.
Downs (1957) An Economic Theory of Political Action in a Democracy, Journal of Political Economy 65 (2). DOI: https://doi.org/10.1086/257897
Hare, C., Highton, B. and Jones, B. (2024) Assessing the Structure of Policy Preferences: A Hard Test of the Low-Dimensionality Hypothesis, The Journal of Politics 86 (2). DOI: https://doi.org/10.1086/726961
Ordeshook, P.C. (1976) The spatial theory of elections: A review and a critique, in Budge, I., Crewe, I. and Farlie, D.J. (eds) Party Identification and Beyond: Representations of Voting and Party Competition.
Peffley, M.A. (1985) A Hierarchical Model of Attitude Constraint, American Journal of Political Science 29 (4). DOI: https://doi.org/10.2307/2111185
See footnote 3 for the reference.
The method is Bayesian Aldrich-McKelvey scaling. See:
Hare, C. et al (2014) Using Bayesian Aldrich-McKelvey Scaling to Study Citizens' Ideological Preferences and Perceptions, American Journal of Political Science 59 (3). DOI: https://doi.org/10.1111/ajps.12151
Simple in the sense that it does not consider how other parties might change their own positions in response to Labour changing its position. In reality, multidimensional party positioning is rarely this simple - but it is my view that the model is still useful. YMMV.
I realise how weird a choice this might seem on writing it out. An informal convention in political science is for left/progressive/liberal-coded positions to be negative/small, and right/conservative/authoritarian-coded positions to be positive/big. This is so that ‘left’ stuff goes on the left of the plot and ‘right’ stuff goes on the right of the plot.
The others should, anyway, in part because they’ve won the provisional consensus of the political science community.
Hewitt, L. et al (2024) How Experiments Help Campaigns Persuade Voters: Evidence from a Large Archive of Campaigns’ Own Experiments, American Political Science Review 118 (4). DOI: https://doi.org/10.1017/S0003055423001387