In October 1998, the Independent Commission on the Voting System (ICVS) recommended AV-plus for electing 659 UK MPs: mostly in around 543 AV constituencies, with 15-20 percent compensatory MPs, in 80 relatively small Top-up areas (electing 4 - 11 total MPs per area, including one or two Additional Members). Compensating parties under- represented by Constituency MPs (AV), d'Hondt allocation of Top-up MPs would render total MPs semi-proportional to Second / Party Votes, with choice of candidate within party (Open List PR).
Is AV-plus a form of PR? Is AV-plus an Additional Member System (AMS)? Indeed, is AV-plus a form of multi-member STV? Answers to all three questions depend on what you mean by PR, AMS and STV, respectively!
"The Jenkins Committee's recommendations have much to recommend them, but there is little more chance of them delivering a proportional result than there is under the present system".
His introduction of a probabilistic element is welcome: here comparing AV-plus with so-called 'First-Past-the-Post' (FPP).
Jenkins estimated that, in the 1997 UK General Election (FPP), AV-plus would have reduced the "DV score" from 21 percent to 13.2 percent. Measuring Deviation from Proportionality, DV = Loosemore-Hanby Index = LHI. LHIs of 4 - 8 percent represent practically "full proportionality"; and for AV-plus, Jenkins claimed only 'broad proportionality'.
Compare other d'Hondt systems. In the May 1999 Scottish Parliamentary Election (FPP-plus: seven Top-up MSPs per Region × 8 = 56 / 129 = 43 percent), the Second / Party Vote LHI was 10.5 percent. Ironically, total MSPs proved more representative of First / FPP Votes (LHI = 5.4 percent)! In the May 1999 Welsh Assembly Election (FPP-plus: four Top-up MWAs per Region × 5 = 20 / 60 = 33 percent), the Party Vote LHI was 11.2 percent (Guardian, 8 May 1999).
In Britain, the June 1999 European Parliamentary Election LHI reached 14.1 percent (Closed List PR: 84 MEPs: 4 - 11 per Region: Guardian, 15 June 1999): 'broad proportionality'. Such pure d'Hondt seat allocation favours larger parties, proving considerably less representative than Largest Remainder (which would have yielded LHI = 6.1 percent).
Over the last 10 Irish general elections (multi-member STV, 1969-97), aggregate First Count LHI averaged 7.0 percent (ranging 3.4 - 12.9 percent between elections: from 'full PR' down to 'broad PR' in 1997). Between three- and five-member STV constituencies (averaging 7.0 and 7.4 percent, respectively), LHIs differed insignificantly. In the June 1998 Northern Ireland Assembly Election (six-member STV), First Preference LHI was 6.6 percent (Irish Times, 29 June 1998).
"the additional member system is, conceptually, a 'closed' list system ... it combines many of the faults of the first-past-the- post system with many of the defects of list systems of proportional representation".
Confusingly, Bogdanor was alluding to "a variant of the German system", recommended by the Hansard Society Commission on Electoral Reform: FPP without separate party voting, topped-up regionally with FPP 'best losers' (25% of all MPs).
At the 1994 German General Election, 328 Constituency MPs were elected by FPP (First Votes); d'Hondt allocating 328 Top-up MPs, in 16 Regions, according to Second Votes (Closed List PR). However, Second Votes may indicate voters' second preference parties 5; as suspected in the 1999 Scottish and Welsh elections (Times, 8 May 1999):
"All electors then had a second vote. This should have been used to indicate their favourite political party. There is widespread confusion on this point and the fear that some people thought that they were being asked for their second preference".
Voting separately for constituency MPs and parties - One Voter Two votes - may well encourage tactical (insincere) voting. Especially in areas safe for the most-favoured party, a Second Vote for that party would elect no Top-up MP (and thus be wasted); and it would be more rational to vote for a less-favoured party, against a least-favoured party 4.
The average area represented by a German MP under FPP-plus in 1994 was over 20 times that of 656 FPP constituencies. In contrast, the mean area covered by each MP under AV-plus, with two Top-up MPs per area, would be only three times that of 659 FPP constituencies - just like three-member STV!
Public outrage forced a constitutional amendment, guaranteeing a bare parliamentary majority to a party exceeding half of all STV First Preferences. At the 1987 Maltese General Election, the Nationalists won the same majority of First Preferences (50.9 percent), and minority of STV seats (47.7 percent); and therefore received four additional seats (totalling 35 / 69 = 50.7 percent of all MPs).
The 1992 General Election required no compensatory seats. Yet at the 1996 General Election - with fine impartiality - the Maltese Labour Party won 50.7 percent of First Preferences, but only 47.7 percent of STV seats! Accordingly, for a bare parliamentary majority, Labour received four additional seats (again totalling 50.7 percent of all MPs).
These few compensatory seats (4 / 69 = six percent) were occupied by STV Final Count 'best losers': runners-up for the party under-represented by STV alone. Thus Additional Members both stood for election and retained their constituency links.
The Maltese AMS (STV-plus) neatly solved an acute political problem. Incidentally, Malta remains a two-party polity, despite the opportunities for party fragmentation afforded by multi-member STV.
In the June 1998 Northern Ireland Assembly Election, the Social Democratic and Labour Party won more STV First Preferences than the Ulster Unionist Party (177,963 / 172,225 votes); but fewer Members (24 / 28 seats). That owed little to vote-transfers (Irish Times, 26 June 1998): even SDLP final 'preferences' exceeded those for the UUP (191,091 / 185,560 votes). The SDLP deserved five Additional Members (29 / 28 total seats proportionating SDLP to UUP votes).
STV-plus could well be generalised to British conditions; and would remedy the corruption of Party Vote Management - a form of tactical voting which disfigures Irish STV. Party Vote Management involves a party's supporters spreading their First Preferences evenly among its candidates: intended to keep them in the STV count for as long as possible (hoovering up stray transfers). In addition, each party nominates one more candidate than it expects seats; avoiding premature elimination through spreading its votes too thinly ('over-nomination').
Proportionating total (Constituency + Compensatory) seats to Party First Preferences, STV-plus could also reconcile the main parties (fearing the spectacle of disunity) to multi-member STV's wider choice of candidate. With each party's candidates competing for the voters' affections, their First Preferences would complement each other in determining parliamentary party strengths under STV-plus. AV-plus could be redeemed likewise.
Likewise, the ICVS stressed "open as opposed to closed lists for Top-up members": Second / Party Votes offering a choice of candidate. However, with three candidates per major party, preferential (rank-ordered, numbered) Second Votes are clearly better than categorical (single choice, X-marked) voting.
In that case, why not simply integrate First / AV with Second / Party votes: semi-proportionating total (AV + Top- up) MPs to AV First Preferences; with AV Final Count 'best losers' as Top-up MPs? Aiming to maximise AV First Preferences (and hence total MPs), each party would become highly motivated to nominate more than one candidate per constituency.
Thus could an improved AV-plus increase voter choice, both within and between parties. With a transferable choice of candidate within party, Party First Preferences are most sincere.
The ICVS argued that separating Constituency from Party votes would liberate voters from unwanted candidates of preferred parties; and that transmuting Constituency 'best losers' into Top-up winners would be hard to explain 4. Valid against FPP-plus, both objections are much attenuated by more than one AV candidate per Constituency Party.
One Voter One Vote could then become far less wasteful than One Voter Two Votes. In both Scottish and Welsh elections, around half of both First and Second votes elected nobody (Guardian, 8 May 1999).
Moreover, the ICVS version of AV-plus (switching between preferential and categorical voting) is even more complicated for voters than multi-member STV. Indeed, it has been argued - rather cruelly - that its very complexity would favour that next step!
AV-plus would be simplified by integrating Constituency with Party voting, each party nominating more than one AV candidate per constituency; rendering total MPs semi-proportional to First Preferences; and exploiting the rich crop of Final Count 'best losers' as Top-up MPs. AV-plus could thus achieve much towards multi-member STV (which may also benefit from some mild topping-up: STV-plus).
It remains unclear why the Scottish Parliament includes more Top-up Members (43%) than the Welsh Assembly (33%): both more than the ICVS-proposed House of Commons (15-20%). With 20-25 percent Top-up MPs, AV-plus would increase Party Representativity ('proportionality').
In the end, parties must nominate parliamentary candidates; while the voter's predicament is paramount. With preferential voting in fairly small Top-up areas, AV-plus essentially places PR on a human scale. Commitment to that principle need not rule out debate on technical improvements (short of multi-member STV) before the Referendum.