January 17, 2001
G. Bingham Powell ©:
Response to comments of Erik Moberg, Moberg Publications, concerning Elections as Instruments of Democracy.
Estimating Opposition Influence
1. Conceptional underpinnings of opposition influence
2. Influence against majority governments: fragments of evidence
3. Influence against minority governments
4. The empirical range of authorized representation scores: implications
5. Survey of experts: the scale
6. Moberg's suggested approach
Another View ... and yet Another
I appreciate Erik Mobergís thoughtful and careful description of my
book Elections as Instruments of Democracy: Majoritarian and Proportional
Visions (Yale, 2000b.) I am glad to take this opportunity to respond
to his interesting "Comment." First, I would emphasize that we are in agreement
in many points that he makes concerning the nature of the general enterprise
and my efforts to understand how elections may, and sometimes do, connect
citizens and policymakers.
Iíd like to clarify slightly further two points that he does not mention. One is my use of the term "visions" in the title and in the analysis. I introduce this intentionally broad term to include elements of a theoretical model, normative criteria, and empirical structures and processes that are associated in writings about democracy. Thus, in describing the "majoritarian vision," I try to elucidate the conceptual assumptions about citizens and preferences, the normative criteria concerning how those should be taken into account by policymakers, and the empirical theory that predicts that specific institutions in contemporary democracies will create the kind of connection normatively desired. Of course, the normative elements cannot "fail." So, when I speak of the relative successes and failures of the majoritarian (or proportional) vision, I mean that the institutional arrangements linked to that vision do not work to create the kind of citizen-policymaker connection that the empirical theory predicts and which are normatively valued in this vision. The reason that itís important to keep this clear is that it is not always recognized that the normative criteria associated with the two visions in much of the writing about these issues are often somewhat different. This is the sense in which, as Erik Moberg, mentions, I find that to some degree each vision succeeds "on its own terms," in that the institutional arrangements it posits create the kind of connection that it values normatively.
The second point touches exactly on this issue of the vision and institutions
associated with it succeeding on their own terms. In general I do characterize
my findings about the relationship between citizenís votes and the selection
of policymakers in that way. The majoritarian institutions of single member
district election rules and concentrated legislative/executive power usually
elect single-party (or pre-election coalitions) majority governments that
have relatively unchecked policymaking power. The proportional institutions
of low-threshold, (large magnitude), PR election rules and dispersed legislative
power (in committees and elsewhere) usually result in much more dispersed
and representative policymakers. Moberg mentions the most obvious exception:
when the majoritarian institutions give unchecked political power to the
party that does not finish first in the popular vote, a situation that
seems a clear failure of the institutions to perform as normatively and
empirically posited. However, he fails to mention a second point that I
consider important also, but perhaps do not give enough attention: the
failure of "majoritarian" institutions to elect governments that are based
on a majority of voters. It is, in fact, very rare for a majority of voters
to support a single political party, even with single member district election
rules and concentrated power. What I count as a successful winner is usually
the party (or pre-election coalition) that won a PLURALITY of votes, but
not a majority. While this government usually has a majority in the legislature,
it did not win the support of a majority of voters (on average, about 43%,
but as low as 35% in New Zealand in 1993.) From the point of view of the
normative foundations of the majoritarian vision, this is a non-trivial
deficiency, so I wanted to be sure and mention it here.
Estimating Opposition Influence.
Erik Moberg has exactly put his finger on a major concern about the analysis in the book, in raising the question about the magnitude of opposition party influence under various conditions. He argues that I greatly overrate the absolute magnitude of opposition influence in the empirical institutions of the systems classified as proportional. He accepts, for the most part, my evidence that the systems classified as majoritarian have serious deficiencies in fulfilling either their vote-policymaker ideal connection or the desired close congruence between the position of the median voter and the policymakers. But he feels that my overestimation of opposition influence leads me to be too optimistic about performance of the proportional systems: both in terms of the level of authorized representation of the policymakers and their correspondence to the median voter.
Let me be clear that I agree that this is an empirical issue, which could be only resolved through detailed analyses of opposition influence under various conditions in various systems. I think, in fact, that we probably agree that it is a complex problem to study comparatively and systematically, even with great expenditure of time and resources. But I am more optimistic than Moberg that we can get a good sense of the relative answer to this, in comparing authorized representation and citizen-policymaker congruence in majoritarian and proportional systems.
To save time and space Iíll focus this reply on Mobergís analysis of my authorized representation measure, especially the coefficient assigned to weight the influence of members of the legislature not sharing cabinet positions in the government (or giving it formal "outside" support.) As Moberg accurately notes, I suggested that opposition representatives be given a minimum weight of 10% of their proportion of the legislative if facing an unchecked majority government, greater weight if facing other types of government, especially a minority government, and also greater weight if enjoying a more favorable committee system (and associated legislative rules.) Moberg objects to the absence of a coherent theoretical story about those weights, which are primarily suggested on the basis of a rough reading of descriptive literature, although also buttressed by expert attributions of opposition influence from the Laver-Hunt comparative study. He proposes as an alternative a specific index, based on the probability that opposition parties would provide the decisive vote to a minority government, if there were eight such parties of equal size.
I appreciate the desire for a coherent, theoretically based, story and am sorry that I canít provide one. But I can do several relevant things: sketch my conceptual underpinnings, which help explain why I donít think Mobergís approach is the right one; mention some diverse empirical evidence about oppositions facing majority and minority governments; indicate the range within which the coefficients must lie and why that supports my basic conclusion of the relative superiority of the representational systems; and raise some specific objections to Mobergís coefficient, even given his concept.
1. Conceptual underpinnings of opposition influence. This goes back to the point, reported by Moberg in his summary, that the proportional vision favors all citizens having some influence, being taken into account to some degree, in policy-making. I think that having to be a decisive vote in passing legislation is much too strong a criterion to apply here (although it might be more relevant to the majoritarian, "citizen control" oriented, vision.) This is one reason why I think that even if the government has a majority, and can count on relatively strong party cohesion, the legislative opposition is not without some influence.
2. Influence against majority governments: fragments of evidence. Admittedly, this is a good example of the "fundamental problem of causal influence:" itís not easy to show that the government behaves differently when faced with a parliamentary opposition, able to challenge its proposals and threatening to mobilize the electorate against them in the next election or in the streets at the moment. But I think that in fact this is the case, that the legislative opposition by the fact of its existence and potential resistance causes the government to propose legislation taking some account of the oppositionís positions. We do know that even majority governments in parliamentary systems do not get all their legislation passed. British majority governments get 97% passed; Germany majority coalitions only get 85% passed, according to official statistics, and much of what is passed is substantially modified in committee. Baumgartnerís study of French education legislation showed the government got about 30 bills passed, but opposition and street protest caused three to be withdrawn (Baumgartner 1989.) So even by these criteria majority governments are not all-powerful. None of these examples take account of the degree to which governments shape their legislative proposals in anticipation of opposition challenges. Admittedly, my assigning a 10% coefficient to the opposition representatives is based neither on a hard model, nor on detailed case analysis, nor on comparative statistics. Itís just based on a general impression of the rough magnitudes. But the combination of the conceptual meaning of opposition influence in the proportional vision and what we know about policymaking in majoritarian parliamentary systems suggests strongly to me that a zero-influence coefficient is much too small.
3. Influence against minority governments. I think there is quite a bit of evidence that oppositions have more influence against minority governments. British data suggest that minority governments are less successful in passing their proposals. A study by Lanny Martin (1999) shows that in four parliamentary systems opposition party positions on issues have more impact on the timing of legislation under minority governments. How large a coefficient to assign is certainly more debatable. I suggested that oppositions against minority governments have an influence coefficient of .5. (I also added a substantial analysis (in the appendix) about the alternative measure assuming that more distant parties, unless needed to form a majority, had less influence than closer ones.) However, as I admit in the book, I have no hard evidence based on case studies or comparative policymaking that support the .5 coefficient. The only support is based on the Laver-Hunt survey and the idea of minimum/maximum range, which is, however, indicative.
4. The empirical range of authorized representation scores: implications. Moberg and I agree that given the concept of citizen influence in the proportional influence systems, voters for government party should be assumed to have an influence coefficient of 1.0. This means that the minimum authorized representation score will be the percent of voters who supported the government. Empirically, this was 43% in the average government in a majoritarian system and 50% in a proportional influence system. So, even if we were to assume that the government voters should get an influence coefficient of 1.0 and opposition voters should get a coefficient of 0, the advantage in authorized representation would go to the proportional systems. (By a difference of 7.) At the other extreme, if we assume, as disproportionality measures do implicitly (see Gallagher 1991) that government voters and opposition voters both get authorized representation scores of 1.0, so just getting into the legislature is all that counts, we find that in the majoritarian systems voters get 83% representation in the legislature, while in the proportional systems, they get 94% representation, a difference of 11. Now, there is a large difference in absolute authorized representation levels reflected in the two different assumptionsĖin this sense it makes a lot of difference whether we think opposition votersí legislative representation should be multiplied by 0 or 1. But in both cases the proportional systems do notably better. The interesting thing is that if we make various assumptions about differential weighting affected by things like minority governments, committee systems in which the opposition shares chairmanships, and so forth, all these things serve to upweight the opposition coefficient more in the proportional systems than in the majoritarian ones. And that increases fairly sharply the proportional system advantage (to around 20, where the oppositions against minority governments and also enjoying the most favorable committee systems get a coefficient of .75.) While we can argue just where in the differential between 7 and 20 the proportional systems lie, and I would stand by the bookís general case that itís closer to the latter than the former, there is little doubt in my mind of the relative advantage of the proportional systems in generating significantly higher levels of effective authorized representation. (A more extensive discussion, with further simulations, can be found in my recent APSA paper, Powell 2000a.)
5. Survey of experts: the scale. Return to the one empirical study that directly and systematically addresses this issue: the Laver and Hunt survey of experts, in which they are asked to assess the impact of the opposition on government policy (Laver and Hunt 1992.) As Moberg quite correctly points out, the problem is in translating the 1-9 scale ("no impact" to "high impact") into coefficients of opposition influence on policymaking. There doesnít seem to be much dispute about the low endĖ"No impact" is clear enough. In my view it is important to note that even in the highly majoritarian systems, such as Britain and Greece, the average scores assigned by the experts were NOT 1, but about 2 (in New Zealand, 3.4; in Canada 3.5), so experts see some opposition influence even in the majoritarian systems. At the high end, it clearly is more disputable; I assumed that "high impact" meant "the same as government parties," and Moberg properly challenges this assumption. He may be right; I may be right. We canít tell without detailed empirical studies. But the experts do see the opposition as having a lot more impact against minority governments. Regression analysis suggests that a minority government is worth about half the distance on the 1-9 scale. Countries like Denmark and Norway are getting scores around 6.5 on the 9 point scale. To me this seems good evidence that policy making is taking some account of the representatives of opposition voters in these contexts, which is what the proportional influence vision demands. So, again, we could disagree about the exact amounts, but we shall come to the same conclusion about relative influence in the two system types. And, in my view, the rough estimates used in the book are certainly not inconsistent with the available (admittedly scanty) empirical evidence about absolute influence.
6. Mobergís suggested approach. I was intrigued by Mobergís approach suggested in his comment. He considers a minority government example, which assumes a minority government with 44 seats in a 100 seat legislature, faced by 8 opposition parties of equal size (7 seats each.) The support of one party is enough for the government to get a majority of 51 seats. If there are no differences between opposition parties, this means each has a 1/8 chance of being decisive in a given vote. He argues that this 1/8 should be the influence coefficient (.125), which is, of course, far smaller than my .50. He then generalizes this to a formula for the coefficient of influence: (.51-G)/(1-G), where G is the share of the legislature controlled by the government.
Although I was intrigued by this approach, and appreciate the effort to create a consistent, theory based, way to estimate the coefficient, I think it is not adequate and, in fact, misleading. It is just a way to a minimum coalition of legislative seats (51%). There are two problems here. First, as I have already argued above, it assumes that the sheer voting power is sufficient to estimate opposition influence, but experience in majoritarian systems suggests that legislative opposition seats are worth something more than zero even against a government majority. Second, it assumes that there will be an opposition divided into enough seats exactly to get to 51%. Moberg comments that "Even so, however, and since parliamentary parties usually behave as indivisible lumps, the authorized representation may be somewhat higher in real situations." There is a lot concealed in that admission. To take an example at the other extreme, suppose we have a government with 44 seats and only two, equal sized, opposition parties. Now each opposition has 28 seats, and with the same assumption we get a coefficient of .5 and authorized representation scores of 72. Perhaps the opposition coefficients in a more general version of Mobergís model should be 1/Opp where Opp is the number of opposition parties. As the average effective number of parties in European legislatures in this period is about 4, it can be seen that if the government has about half of these, the number of opposition parties will be in the range of 2 or 3, yielding coefficients much closer to my .5 than Mobergís .125 (because the number of opposition parties is closer to two than to eight equally sized parties.) I do not argue that we should use this method, but if we want to explore it, we have to do something more serious about these lumpy legislative parties. As Riker and his followers discovered long ago, if we interpret minimum winning coalitions to be 51% of the legislature, we very seldom find it.
7. Congruence. The basic finding about congruence is that governments are further from the citizen median in the majoritarian systems than in the proportional systems, quite the contrary of what we might expect from the Duverger (1954) plus Downs (1957) models of representation in such systems. (See the elegant formulation by Cox, 1997, Ch. 12.) As reported in the article published by Georg Vanberg and myself in the BJPS last summer, (Powell and Vanberg 2000,) the legislative medians are also (even) further from the citizen medians in majoritarian systems than in PR systems. The reasons why this is so are discussed in the book. As Moberg fundamentally accepts these, I wonít go into them here, but I think this is a major contribution of the work. In the case of congruence the correspondence of medians is a normative standard implied by majoritarian theory, so here we do not have the situation where each system works well according to its own standard.
Still, the discussion of about policymakers, as distinct from governments and legislatures, and how to calculate their position is also an important part of the bookís analysis. As the contrast of government congruence is even sharper than of vote support, we do have a strong base line for the relative superiority of the proportional systems. If we assumed, with Moberg, that government positions completely dominated policymaking most of the time in all systems, this would be the end of the story. This comparison already includes, let me be clear, the minority governments without any adjustments. Their somewhat greater distance from the median citizen (itself a somewhat surprising finding, by the way, given theoretical expectations about the dominance of the median legislator,) and greater frequency in proportional systems diminishes the difference between majoritarian and proportional system results, but that difference is nonetheless statistically significant. (I argue that it is also substantively significant, a point that I wonít go into here.)
The analysis of congruence of policymakers and citizens is somewhat
more complicated than authorized representation, because it isnít necessarily
true that assuming more influence to the opposition parties leads to better
congruence as it does (by definition) to better authorized representation.
Indeed, majoritarians would expect just the opposite and in a few cases
we do find the government on the median and the opposition pulling it away.
But only in a handful of the 70 cases. What basically drives the results
in the policymaker analysis is that governments and oppositions are nearly
always on opposite sides of the median citizen, and both are some distance
from that median. (Governments are over 1 ½ points in majoritarian
systems and about 1 point in proportional ones; the net opposition distances
are a bit further away, but on the other side of the median.) Under that
configuration, giving the opposition substantial weight, but not more than
the government, will work to pull the "policymakers" (weighted combination
of government and opposition) towards the median. This helps the congruence
in both majoritarian and proportional systems. Indeed, if in both systems
we found that oppositions had as much influence as governments (which,
of course, they almost never do,) it would help the (more distant) majoritarian
governments even more than the proportional ones. But in practice for quite
a wide range of opposition influence coefficients the effect of increasing
those coefficients Ėbecause of minority governments and opposition-friendly
committee systemsĖ is to pull the proportional system governments towards
the median faster than the weak opposition effects are pulling the governments
in the majoritarian systems. Some of these simulations are shown in my
APSA paper (Powell, 2000a), although I donít have any very general theoretical
results. (Iím trying to work on this problem.) The opposition influence
coefficients used in the book cut the policymaker distance in half in the
average proportional system, while reducing it only about 15% in the majoritarian
systems. The specific margins are open to debate, but the general pattern
is pretty robust. Thus, Mobergís dismissal of the congruence findings for
policy makers because of uncertainty about the exact coefficients seems
Another View.....and yet Another.
Let me conclude with comments on Mobergís point about "another view" of politics in proportional systems, and suggest yet another beyond that, though perhaps not dissimilar in spirit. In his last substantive section Erik Moberg suggests a problem that is among many not considered in my book, the question of when politics focuses on "special interests rather than on ideological issues of common or general relevance." He discusses this problem in terms of spatial modeling versus zero-sum game theory. I donít profess to have the technical competence to discuss these as abstract conceptions, but I do take his point, or what I conceive to be his point, about special interests. I agree that this is an interesting problem for general analysis. As I havenít analyzed it, I canít say whether theoretically we would expect this to be more of a problem in (empirically) majoritarian or proportional systems. I once tried to take a bit of a look at the relationship between these two kinds of systems and various measures of political corruption, and I donít find any systematic relationship in models specified thus far. Living in a two-party system, (but one in which the parties are, admittedly, less cohesive than in the average parliamentary democracy) Iím not especially impressed with the resistance of two-party competition to special interest demands. But, again, this is an important question that it would be extremely interesting to investigate empirically.
Another problem concerns me more, however, and I would welcome thoughts on it from Moberg and his readers. This is the problem of deadlock and the status quo (revision point) in the proportional systems. As Moberg suggests, my analysis assumes that increasing the influence of the opposition will systematically pull government policies towards the opposition positions. On one hand, this is consistent with the proportional vision that suggests all citizens should be taken into account in policymaking. But on the other hand, in practice being pulled towards the opposition may mean being blocked from undertaking innovative policies that are not acceptable to the opposition. This would seem to be more desirable to the various "haves" in a society, and to preference groups of all kinds who put defending their interests ahead of introducing change, than to the "have-nots" who favor change. Obviously, this is not a new idea, but it worries me that I have found no way to handle it in my current analysis. Iím not sure whether this is a theoretical problem built into the proportional vision or an empirical problem that requires an as yet undiscovered mode of analysis.
Again, I want to thank Erik Moberg for his careful, constructive and
thought-provoking analysis. While I disagree with some of his suggestions,
his desire for theoretical underpinnings and his attention to details of
operationalization should encourage us all.
Baumgartner, Frank. 1989. Conflict and Rhetoric in French Policymaking. Pittsburgh: University of Pittsburgh Press.
Cox, Gary W. 1997. Making Votes Count: Strategic Coordination in the World's Electoral Systems. Cambridge: Cambridge University Press.
Doering, Herbert. 1995. Ed. Parliaments and Majority Rule in Western Europe. New York: St. Martinís.
Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.
Duverger, Maurice. 1954. Political Parties: Their Organization and Activity in the Modern State. (B. North and R. North, trans.) New York: John Wiley.
Gallagher, Michael. 1991. "Proportionality, Disproportionality, and Electoral Systems." Electoral Studies 10:33-51.
Laver, Michael and W. Ben Hunt. 1992. Policy and Party Competition. London: Routledge.
Martin, Lanny. 1999. "The Strategic Government Agenda in Parliamentary Democracies." Rochester, N.Y., University of Rochester, typescript.
Powell, G. Bingham. 2000a. "Election Laws and Representation: Beyond Votes and Seats." Delivered at the 2000 Annual Meetings of the American Political Science Association, Washington, D.C.
.....2000b. Elections as Instruments of Democracy: Majoritarian and Proportional Visions. New Haven: Yale University Press.
Powell, G. Bingham and Georg Vanberg. 2000. "Election Laws, Disproportionality and Median Correspondence." British Journal of Political Science 30: 383-411.