Hugo Voting: Let’s Look at the Record (Again)
by George Flynn
[originally published in Apa:WSFS 1995 #4, July 1995; updated in January 1997]
Back in 1988, I wrote an article on Hugo voting statistics for The Mad 3 Party (#30, November 1988). That was long enough ago that it may be worthwhile to do it again. So here’s the new version, with lots more boring numbers. The 1988 version began as follows:
I am often amazed by the kinds of misinformation that people believe about Hugo voting figures. Just an example: One widely distributed report on this year’s nominations said, "There were 122 ballots with best fanzine nominations (29%, a high number)." The 122 was in fact the number of fanzines nominated, not the number of people nominating; but where did the idea come from that 29% would be "a high number"? (As we shall see, 45% would be typical.) This idea that hardly anyone votes on the fan Hugos is perhaps the most widespread of the misconceptions I alluded to.
As Mike Glyer pointed out in a recent File 770, it’s just ten years since full Hugo voting counts began to be routinely released. (Mike and I sponsored the rule that now requires this publication.) The numbers come out every year, but not many people wade through them to see what they all mean. After ten years, though, we should have enough data to do some useful analysis. The numbers that follow mostly come from the annual tabulations in the newszines, with some exceptions that I’ll note.
(The exceptions are from official reports of Hugo administrators, some of them by me.) Well, there’s probably just as much misinformation around today, but we do have an additional eight years of data, including some details not available for earlier years. The rest of this article is organized mostly like the original one; however, I’ve come up with many new ways to tweak the numbers. And we’ll see whether the conclusions turn out the same . . .
Table 1 — Total Vote Counts (1971–96)
|Year||Hugo Ballots||No. of||Site-Selection||Advance||Final Hugo|
(* = overseas Worldcon)
How many people vote on the Hugos? The left side of Table 1 gives the numbers ofnominating and final ballots each year since 1971, along with the numbers of site-selection ballots for comparison. (In this and all the tables, the "year" is that in which the voting takes place, not the year of the works being voted on.) Also included is the number of categories on the final Hugo ballot (including Campbell and Gandalf awards). Note that in the early ’70s the size of the Worldcon was growing rapidly. But since the mid-’70s the "normal" Hugo nominating-ballot count has been fairly stable at around 500. The final-ballot count has fluctuated more, but the "normal" count seems to have declined from around 1200 in the late ’70s to perhaps 900 today. Still, there seems to be a relatively stable population of Hugo voters — more stable, at least, than the Worldcon membership itself.
The most obvious exceptions to this stability can be accounted for by geographical factors: Australia (1975 and 1985) and Holland (1990) had about half the usual vote. (For Winnipeg in 1994, the final vote was similarly low; but since ConFrancisco also distributed nomination ballots, the nominating vote in 1994 was the second-highest ever.) As for other fluctuations, the high final vote in 1980 may derive from the special effort Noreascon II made to get ballots to last-minute joiners; did something similar occur in 1976? Conversely, some relatively low vote counts (e.g., the 1988 and 1991 nomination counts) have been attributed to late distribution of ballots.
In 1996, there were 111 nominating ballots and 605 final ballots for the 1946 "Retro-Hugos"; 558 of the latter voters cast ballots for both 1946 and 1996 Hugos. (Since the voting pattern is likely to be quite different, the Retro-Hugos will not be considered further in this article.)
What fraction of the Worldcon membership votes? Now consider the right side of Table 1. The "Advance Memb." column gives the best numbers I have been able to find for the final pre-convention Worldcon membership; some of these numbers are obviously more accurate than others. And the "Final Hugo Vote Ratio" column gives the number of final Hugo ballots divided by the "Advance Memb." As you can see, this ratio has dropped from around 40% to around 15%. Depressing, yes? One problem with these numbers is that the Worldcon membership tends to increase rapidly at the last minute, and many of these people join too late to vote. So the voting percentage among people who actually get a chance to vote is higher than tabulated here, but probably not by much. (For what it’s worth, in 1980 I estimated that 17% of the eligible-at-the-time members nominated, while 37% cast final ballots. But these figures were probably higher than average, and for most years there are no good data available for the membership-at-the-time.)
Explanation of category tables: Many of the tables that follow give assorted statistics for individual Hugo categories. For these I generally have data only from 1978 through 1996. Since (as described above) some years had votes much lower than usual, I’ve divided the results into "Low" and "Other": "Low" includes Aussiecon II (1985) and ConFiction (1990) for both nominating and final ballots, plus ConAdian (1994) for the final ballot; "Other" covers everything else. Rather than giving the numbers for every year, for both "Low" and "Other" I simply list the range (high and low bounds) and the average. With regard to Hugo categories, note that the Non-Fiction Book category first appeared on the Hugo ballot in 1980; the Original Artwork category was tested in 1990, and appeared regularly from 1992 to 1996 (when it was abolished). What had been the "Fanzine" category in 1978–83 was divided in 1984 into "Semiprozine" and "Fanzine," so separate tabulations for the two periods are given for these categories.
[I defined the"Low" and "Other" sets in the 1994 version of this article. The 1995 final counts were generally somewhat lower than the previous range of "Other" years, but not by enough to justify a complete reorganization.]
The average overall votes (from Table 1) for the subsets defined here are as follows:
Nominating: "Low": range 222–291, average 256
Nominating: "Other": range 352–660, average 515
Final: "Low": range 443–491, average 473
Final: "Other": range 744–1788, average 1137
Table 2 — Minimum Nominating Vote (1978–96)
|Category||Low (85, 90)||Other|
|Original Artwork (90–96)||5||5||7–13||9|
How many nominations does it take to get on the Hugo ballot? Table 2 gives the minimum nominating vote by category, arranged as described in the preceding section. That is to say, the lowest-ranking nominee to actually make the final ballot for Best Novel, say, has gotten as few as 25 nominations (1978 and 1990) and as many as 96 (1983), with an "Other" average of 49. It is depressing to see how few nominations it takes to get on the ballot, with the "Other" average under 25 in six categories. (Interestingly, the low end of the "Other" range in 9 of 11 possible categories is from 1978, although the total nominating vote wasn’t particularly low that year. This is probably because since 1982 there has been a rule requiring nominees to receive at least 5% of the number of nominating ballots in their category in order to appear on the final ballot. We’ll have more to say about this later.)
The variation between categories — and years — is largely a function of the number of likely nominees. For example, there are only a few well-known professional editors, so the vote in that category tends to be concentrated and the minimum-vote-to-nominate consistently high. Similarly, Dramatic Presentation usually has a high minimum, because in most years there are only a few good SF movies (the low of 8 votes in 1978, when Star Wars blew everything else away, may have been a case of too few plausible choices). In the fiction categories, the vote typically gets more and more scattered as you go to shorter lengths: short stories get almost as few nominations as fanzines. The highest minimum ever was 119 for Dramatic Presentation in 1983; while Original Artwork had the lowest minima for both "Low" and "Other" years. (By the way, the Novella average would be 41 if the category shifts hadn’t been made in 1994; the other figures would be unchanged.)
Table 3 — Maximum Nominating Vote (1978–96)
|Category||Low (85, 90)||Other|
|Original Artwork (90–96)||15||15||11–47||30|
How about the high end of the nominating range? Table 2 gave the figures for the lowest-ranking nominees in each category; in Table 3 are the corresponding figures for the highest-ranking nominees. The ranges here are somewhat wider, since the maximum vote is affected more by single blockbuster nominees. As you might expect, the Dramatic Presentation nominee with the most votes nearly always beats out anything in the other categories (the exceptions were 1979, when Varley’s "The Persistence of Vision" outscored Superman; and 1995, when both Pro Editor and Semiprozine had higher-scoring nominees). The single nominee with the most votes (338) is still Star Wars in 1978 — and even that received less than 2/3 of the total nominating ballots. As you might surmise from that datum, the votes cast in individual categories are always appreciably lower than the overall total; the next few tables will give details.
The lowest maximum vote was 11 for Original Artwork in 1995, when the range of nominations was 8–11. Generally, however, the narrowest ranges appear in the Novelette and Short Story categories: the closest things to ties were 43–49 for Novelette in 1983, and 26–30 for Short Story (seven nominees) in 1992, again illustrating the typical flat distribution of votes for short fiction. (By the way, the "Other" maximum and minimum averages in most categories are several votes lower than those I calculated in 1978: further evidence of the decline in Hugo votes. [Three categories had their lowest maxima in 1995, and one in 1996.] Without the category shifts in 1994, the Novelette average would have dropped further to 50.)
Table 4 — Nominating Ballots by Category (1980–96 Partial)
|Category||Low (85, 90)||Other (Partial)|
|Original Artwork (90–96 part.)||75||75||100–216||140|
|Fanzine (80, 83)||—||—||318–364||341|
|Semiprozine (84–96 part.)||109–169||139||176–320||243|
|Fanzine (84–96 part.)||94–152||123||179–299||232|
How many people make nominations in each category? Unfortunately, these numbers haven’t been published every year; the data I have at hand are only for 1980, 1983, 1985–87, 1989–94, and 1996. (For these years the "Other" range of overall nominating votes is still 352–660, but the average is 524.) Table 4 gives the number of nominating ballots cast by category, for those years only. For the "Low" years, the lower end of the range in every case was in 1985. For the "Other" years available, Chicon V (1991) had the lowest vote in 11 of 13 possible categories; ConAdian (1994) was highest in 6 of 14 categories. (Nevertheless, the average "Other" vote has declined significantly since 1988 — by 41 votes in the average category, ranging from 13 for Semiprozine to 78 for Dramatic.) Print chauvinists will be gratified that (for these years at least) the Novel category always gets the most nominating ballots, the highest being 606 in 1983; however, Dramatic Presentation usually takes the lead on the final ballot. At the other extreme, Original Artwork drew the fewest nominating ballots every year it was on the ballot.
Table 5 — Nominating Percentages by Category (1980–96 Partial)
|Category||Low (85, 90)||Other (Partial)||Overall|
|Original Artwork (90–96 part.)||26||26||23–33||28||27|
|Fanzine (80, 83)||—||—||55–56||56||56|
|Semiprozine (84–96 part.)||49–58||54||42–57||49||50|
|Fanzine (84–96 part.)||42–52||47||44–55||46||47|
The raw numbers are interesting, but it may be more useful to consider the numbers of nominating ballots in each category as percentages of the total nominating vote (see Table 1); these percentages are given in Table 5. As one might expect, in this table there is no significant difference between the "Low" and the "Other" years, so I have added a column giving the overall averages. Curiously, Noreascon II (1980) had the highest percentages in 8 of 12 possible categories; I have no idea why (though I ran the Hugos that year). The lowest percentages are scattered over time, but I note that four categories each had their lowest percentages in 1987 and 1996. The "Other" averages for most of the professional categories have decreased by 2–3% since 1988. For what it’s worth, here’s a table of the percentage nominating in the average category over time:
There seems to be a slight downward trend, though not as strong as the trend we will see below for the final-ballot percentages. It seems that Hugo voters (and fans in general?) are tending to have more specialized interests. (Also, we have tended to add relatively low-vote categories.)
While several categories fairly consistently fall below 50%, in these years only Original Artwork went below 25% (in 1992 and 1996); the lowest in other categories is 26% for Fan Artist in 1986. In particular, the numbers for Fanzine demonstrate the point I made in the 1988 introduction: nearly as many people nominate fanzines as nominate in the short-fiction categories. (As we’ll see below, even the low-vote categories get somewhat higher voting percentages on the final ballots.)
Table 6 — Number of Items Nominated (1980–96 Partial)
|Category||Low (85, 90)||Other (Partial)|
|Original Artwork (90–96 part.)||101||101||113–219||166|
|Fanzine (80, 83)||—||—||142–190||168|
|Semiprozine (84–96 part.)||26–48||37||38–60||50|
|Fanzine (84–96 part.)||75–102||88||100–176||138|
(*: "Other" numbers for these categories do not include 1980)
How many different items receive nominations? Here’s something I didn’t include in 1988. We have data on this question from 1980–85, 1988–91, 1994, and 1996. (As you may have gathered, any two Worldcons seldom report quite the same sets of data.) These numbers appear in Table 6, in the usual format. (In 1980 I lumped the three short-fiction categories together on the first pass; the best I can say is that 468 items were nominated in one or more of these categories.) The "Low" years still have smaller numbers, but not by as much as in Tables 2–4. Curiously, 1996 had the lowest "Other" numbers in seven categories. There certainly are a remarkable number of things that someone considers worthy of a Hugo nomination. Nevertheless, the relative paucity of candidates in the Novella and Semiprozine categories is particularly striking.
Table 7 — Total Number of Nominating Votes Cast, and Nominations per Ballot (1982–96 Partial)
|Total Votes Cast||Noms. per Ballot|
|1992||82–83, 89, 93–94, 96||1992||83, 89, 93–94, 96|
|Original Artwork (90–96 part.)||208||270–500||366||1.8||2.3–2.7||2.5|
|Semiprozine (89–96 part.)||429||448–687||560||1.8||2.1–2.4||2.2|
|Fanzine (89–96 part.)||270||436–678||547||1.2||2.3–2.5||2.4|
How completely do people fill out the nomination ballots? For 1982–83, 1989, 1992–94, and 1996 (all "Other" years) I have the total numbers of votes cast by category on all the nomination ballots. Since for all of these years except 1982 we also know the number of ballots cast in each category, we can divide and obtain the average number of nominations per ballot in each category. As you can see, I have separated 1992 from the other years, since in almost every case the 1992 numbers are significantly lower than all the others. (I suspect an error in the 1992 report; for example, analysis of my 1989 data shows that the averages would typically be lowered about 0.7 if the "Total Votes Cast" actually omitted nominees with fewer than 5 votes.) In any case, it is clear that the average voter makes only about half as many nominations as are allowed. (This is not necessarily a bad thing: "Bullet-voting" for one or two nominees is certainly a rational strategy if one’s paramount aim is to get those particular nominees onto the ballot. It’s on the final ballot that failing to vote for all the nominees wastes one’s voting power.) Excluding 1992, 1996 had the lowest numbers of votes in eight categories, and the lowest ratios in four.
By the way, while I have not made a full calculation, it appears that the fraction of all votes cast for those nominees that actually make it to the final ballot varies from around 20% for Novelette and Short Story up to around 70% for Semiprozine. (The reasons are left as an exercise for the reader.) While at the other extreme, a substantial number of people pick nominees that receive only a couple of votes; just as an illustration, here’s a breakdown of the novel nominations in 1989 (for a total of 283 items):
What would be the effect of changing the minimum vote to get on the final ballot? At present the only minimum is the "5% rule" (in Sec. 2.6), which requires that nominees receive votes on at least 5% of the nomination ballots cast in the category (or be among the top three vote-getters). This rule went into effect in 1982, but it has actually been invoked only in the Original Artwork category in 1994 and 1995. (However, it would have applied to the Campbell Award in 1989, to Short Story in 1994, and to Music in 1995; in each case other factors supervened . . . And in 1991 it was initially applied erroneously in eight categories.) Two motions to change this rule were defeated in 1994: one would have reduced the 5% criterion to 2% (under which nothing would ever be eliminated), the other would have replaced it with a 10-vote minimum. Other criteria have also been suggested, so it seems worthwhile to survey the data.
Here is a tabulation of the number of nominees in some recent years that would have been made ineligible by several possible criteria, in the absence of an at-least-three-nominees clause. (For 1994, I am giving the numbers as if the short-fiction category shifts had not been made, since that’s more likely to correspond to future practice.)
³ 10% Novel 2, Novelette 4 (of 6), Short Story 3, Orig. Artwork 2, Fan Writer 2, Campbell 2 ³ 10 votes Orig. Artwork 2, Campbell 1 ³ 20 votes Novelette 3 (of 6), Short Story 2, N-F Book 2, Orig. Artwork 3, Fanzine 2 (of 6), Fan Writer 3, Campbell 3 ³ 25 votes Novelette 4 (of 6), Short Story 4, N-F Book 3, Prof. Artist 2, Orig. Artwork 5, Fanzine 2 (of 6), Fan Writer 4, Fan Artist 2, Campbell 3
³ 5% Orig. Artwork 2, Music 3 (I can’t calculate any other percentages for 1995, since the numbers of nomination ballots per category weren’t released.) ³ 10 votes Orig. Artwork 3, Music 4 ³ 20 votes Short Story 2 (of 6), N-F Book 3, Orig. Artwork 5, Fan Writer 2, Campbell 4, Music 4 ³ 25 votes Novelette 2 (of 6), Short Story 3 (of 6), N-F Book 3, Orig. Artwork 5, Fan Writer 3, Fan Artist 2, Campbell 5, Music 5
³ 5% Short Story 3 (of 6), Orig. Artwork 2 ³ 10% Novel 2, Novelette 2, Short Story 4 (of 6), N-F Book 1, Dramatic 1, Orig. Artwork 4, Fan Writer 1, Campbell 2 ³ 10 votes Orig. Artwork 1 ³ 15 votes Orig. Artwork 3 ³ 20 votes Short Story 3 (of 6), Orig. Artwork 4 ³ 25 votes Short Story 3 (of 6), N-F Book 2, Orig. Artwork 4, Fan Writer 1, Fan Artist 3 (of 6), Campbell 2
³ 10% Novelette 1, N-F Book 3 (of 6), Fan Writer 3, Campbell 3 (of 6) ³ 10 votes Orig. Artwork 2 ³ 15 votes N-F Book 3 (of 6), Orig. Artwork 4, Fan Writer 3, Campbell 3 (of 6) ³ 20 votes Novelette 1, N-F Book 4 (of 6), Orig. Artwork 4, Fan Writer 4, Fan Artist 2 (of 6), Campbell 4 (of 6) ³ 25 votes Novelette 3, N-F Book 5 (of 6), Orig. Artwork 4, Semiprozine 1, Fanzine 2, Fan Writer 5, Fan Artist 2 (of 6), Campbell 4 (of 6)
(Previous versions of this article included data back to 1990, but this table was getting too damn long. See also Table 8 below for more information about low-vote nominees.)
One thing to bear in mind is that sometimes nominees with only a few nominating votes go on to win. The most striking case was in 1991, when Julia Ecklar got only 9 nominations for the Campbell Award, but on the final ballot led with 197 first-place votes, and won. And there are ten other cases since 1989 of nominees receiving over ten times as many first-place final votes as they had had nominations. (But nine of those were in Original Artwork, and only two of them won.) Since the electorate for the final ballot is so much larger (partly because many people haven’t had time to survey the field by the nomination deadline), it is perhaps unwise to limit their choices on the basis of the nominators’ apathy.
Table 8 — Numbers of Nominees Matching Various Criteria (1989–96 Partial)
|Category||No. of Nominees (89–90, 92–94, 96) Receiving||No. of Nominations (89, 93–96) for|
|³ 10%||³ 5%||10th place||15th place|
(*: not including 1989 or 1995)
For how many nominees should nomination numbers be published? This is of course the subject of a Constitutional amendment we ratified in 1995. It requires publishing the numbers for all nominees down to 15th place or 5% of the ballots cast in the category, whichever is less. Might this be too low (so as to threaten the privacy of individual voters), as David Bratman suggested? For reference, Table 8 gives a summary of the data David and I supplied to Apa:WSFS in 1994, plus some data for 1995 (numbers but not percentages) and 1996. Draw your own conclusions. (But note that this table is available only because the 1989 and 1993–94 administrators did release the numbers down to 5% and 15th place. For 1990 — a "Low" year — the 15th-place numbers were generally not reported, but the 10th-place ones, in the same order as above, were: 18, 21, 11, 9, 8, 12, 16, 14, 3, <9, 11, <8, 8, 6.) For Original Artwork, the ³5% figures for the six available years (including 1995) were 9, 8, 12, 3, 3, 15, with an average of 9, and the two consecutive 3’s appeared ominous; also, in 1995 there were only 5 nominations for 5th place, and no data for 10th or 15th. (1996 was somewhat better, but not a whole lot, and the category has now been eliminated.)
That’s about all the conclusions I can derive from the nominating-ballot data. (Note that the last four sections were all added since the 1988 version.) Let’s move on to the final ballots.
How many people vote in each category on the final ballot? Here we once again have numbers for every year since 1978. Table 9 gives the final-ballot vote by category (corresponding to Table 4 for the nominating ballots), and Table 10 the final-ballot percentages (corresponding to Table 5 for the nominating ballots).
Table 9 — Final Ballots by Category (1978–96)
|Category||Low (85, 90, 94)||Other||Recent Other (91–93, 95–96)|
|Original Artwork (90–95)||319–345||331||488–633||566||488–633||566|
For the absolute numbers in Table 9, the "Low" years clearly form a tight cluster in every category. Beyond being significantly higher, the numbers for the "Other" years are not so obviously a single population, because they are distorted by the long-term downward trend that we have already noted (cf. Table 1). This is the biggest change since 1988, when I found "no clear-cut evidence of any trend over time." Some details that make this trend especially obvious: 1980 had the highest vote counts in 10 of the 12 categories that then existed (1984 was higher for Fan Writer and Fan Artist, as well as the then-new Semiprozine). Conversely, 1995 had the lowest "Other" counts in every category except Fanzine and Fan Writer (for which 1994 was lowest); before that, 1993 was lowest in every category; and before that, 1989 was lowest in 10 categories, 1991 in the other 3. Indeed, no individual-year count since 1988 has been anywhere near as high as the corresponding category average I calculated then. The category averages have themselves fallen by an average of 152 votes since 1988 (including 17 from 1995 to 1996). (By the way, this same trend makes the 1984–96 Fanzine average appear abnormally low relative to the other fan categories, which include the higher 1978–83 data: for 1984–96, the "Other" average is 557 for Fan Writer and 561 for Fan Artist.)
Given all this evidence of a downward trend, this time around I have added a "Recent Other" tabulation, covering the last five "Other" years. The drop from earlier years is striking, but the downward trend is noticeable even within this period. Let us look at the percentages in Table 10 to see if we can eliminate the effect of the decline in the overall vote.
Table 10 — Final-Ballot Percentages by Category (1978–96)
|Category||Low (85, 90, 94)||Other (1978–96)||1991–96 (All)|
|Original Artwork (90–95)||66–70||68||55–74||66||55–74||67|
As in Table 5, there is no significant difference between the "Low" and the "Other" years. However, the surprising thing is that the decline over time is still present! The highest percentages were in the period 1978–82 for every category that then existed (9 of 11 possible in 1978–79), and the lowest percentages for every current category were in 1991–96 (7 of 14 categories in 1996, and 4 more in 1995). And as one might expect, the "Other" averages have typically declined by several percent since 1988. To display this trend more clearly, here are the year-by-year average percentages of all the categories — that is, the percentage of final ballots that include votes in the average category (which is equivalent to the percentage of categories in which the average voter votes):
Given the downward trend, the averages for the entire period are of questionable significance; so I have included in Table 10 an additional two columns summarizing the data for the last six years only, to give some idea of what the percentages are today.
For yet another way of seeing the change over time in individual categories, here is a comparison of the 1978–88 averages (based on my 1988 article; numbers in parentheses are for shorter periods) with the 1991–96 averages:
Dramatic Presentation appears to be the only category whose voting percentage has not dropped significantly over this period (and even it has dropped some). It also nearly always has the highest vote — practically everyone votes for it — with Novel usually coming in second (the reverse of the order on the nominating ballots; the only exception was in 1987, when Novel got 914 votes and Dramatic 901). The Dramatic voting percentage has fallen below 90% only twice (88% in 1991, 84% in 1995). All this makes it clear what is really going on here: Not only is the overall number of Hugo voters decreasing, but of those who do vote, fewer and fewer are voting for the categories other than Dramatic. Do you find this as disturbing as I do?
Among the other categories, it’s surprising that Original Artwork, which on the nominating ballots was abysmally low by almost every criterion, on the final ballots was right up in the middle of the pack. (But I can’t help wondering how many of these people were voting on the basis of the artists’ names rather than the particular pieces nominated.) And the three fan categories currently average just below 50% participation, not much higher than on the nominating ballots, with the Campbell Award doing not much better.
It should be noted that there is a rule (in Sec. 2.5) that "no award shall be given whenever the total number of valid ballots cast for a specific category (excluding those cast for ‘No Award’ in first place) is less than 25% of the total number of final award ballots received." Deducting the first-place "No Award" votes will of course give percentages lower than those in Table 10. Here are all the 1980–96 instances in which the voting percentage defined in this rule has fallen below 45%:
|1980:||Fan Writer 42.8%|
|1986:||Fanzine ("No Award" campaign) 44.7%, Fan Writer 44.8%|
|1987:||Fan Artist 42.3%|
|1989:||Fanzine 39.6%, Fan Writer 40.1%|
|1991:||Fan Writer 40.3%, Fan Artist 41.9%|
|1992:||Fanzine 41.5%, Fan Writer 41.8%, Fan Artist 41.8%|
|1993:||Fan Artist 37.8%, Fanzine 39.4%, Fan Writer 41.5%, Campbell Award 44.9%|
|1995:||Fan Artist 39.0%|
|1996:||Fan Writer 36.1%, Fanzine 38.7%, Fan Artist 39.8%|
(In recent years the fan categories often, and the Campbell Award sometimes, have fallen below 50%; Non-Fiction Book has done so once [49.8% in 1992], no other category ever since 1980.) It seems unlikely that the 25% rule will be invoked in the foreseeable future, and we could just as well do without it.
What about special categories? Note that the Hugo rules (Sec. 2.5) say a category can be cancelled "if the lack of nominations or final votes . . . shows a marked lack of interest in that category on the part of the voters"; however, there’s no definition of what this means quantitatively. The last time one of the permanent categories was dropped for lack of interest was for Dramatic Presentation in 1966 — by no coincidence, the last year before Star Trek. The special categories added by Worldcon committees are another matter. True, two such categories did well enough to be converted into permanent categories, and thus appear in the tables above: Non-Fiction Book in 1980–81 and Original Artwork in 1990 (when it wasn’t technically a Hugo). And the Other Forms category in 1988 also did fairly well (nomination range 18–118; 182 items nominated; 965 votes [82%] on the final ballot), though no other Worldcon has picked it up. (Also, in 1978–80 the Gandalf Award was still on the Hugo ballot: the Gandalf for Grand Master of Fantasy received about as many final-ballot votes as Best Novel, while the Best Fantasy Novel in 1978–79 got about 10% less.)
But other experimental categories have flopped badly, and didn’t even make it to the final ballot. Contribution to the Field (1975) was dropped because the nominations were "too diffuse" (Locus reported "141 nominations on 158 ballots," probably referring to the number of people named). Best Non–English-Language Novel (1990) was said to have an insufficient number of nominations, though no numbers were reported. Best Translator (1993) drew only 40 nominating ballots (10%), with the top 5 nominees ranging from 2 to 14 votes. And Best Music last year seems to have done even worse, with only one nominee receiving more than 7 votes. In general, it seems unproductive to have a Hugo category with which the great majority of the voters are unfamiliar, or for which the field of potential nominees is very diffuse. (These principles might also be usefully applied to some of the shakier current categories.)
How completely do people fill out the final ballots? We have data on this from 1980 and 1989. Table 11 gives the percentages of people voting in each category who filled in 1, 2, 3, . . . lines on their ballots. (In most categories there were 6 lines — 5 nominees plus "No Award" — but in a few cases there were 6 or even 7 nominees.) For both years I calculate that the average ballot had about 3.5 lines filled in in the average category. Typically, about 20% of the voters in the fiction categories, 30% in the fan categories, cast "bullet" votes for only one nominee, thus throwing away their vote if that nominee fails to win. (How much of this is due to failure to understand the preferential voting system?) At the other extreme, typically 30–45% fill out their ballots completely: voting for either 5 or 6 out of 6 nominees is equivalent, since leaving only one line blank is the same as voting it last. (By the way, when one line is left blank, about 90% of the time it’s "No Award.") The percentages for both extremes increased slightly from 1980 to 1989, but this may not be significant.
Table 11 — Percent of Ballots Voted for n Places (1980, 1989
As for the variations between categories, one surmises that bullet voting is more likely in those categories where fewer nominees are well known to the electorate as a whole. In both years I made separate calculations of the numbers of ballots on which each nominee was voted in any place. I won’t reproduce the details here, but as one might expect, these numbers correlated well with the order of finish — though even the winners (except for Roger Rabbit in 1989) received votes on less than 80% of the ballots voted in their categories; for the Campbell Award in 1989, the winner was voted on only 51% of the category ballots. Whether out of ignorance or apathy, it is clear that many voters fail to have as much effect on the Hugo results as they are entitled to.
How many people use the "No Award" option? There are those who are quite fond of using "No Award"; indeed, I know some people who vote it not only above things they dislike, but above anything they’re unfamiliar with. But these people are a minority, and on the whole "No Award" gets very little use. In fact, in the entire history of the Hugos "No Award" has won only five times (Dramatic Presentation in 1959/63/71/77, New Author in 1959). Time was when you could at least count on Dramatic Presentation getting a high "No Award" vote, but that’s largely vanished in the era of blockbuster SF films since Star Wars. Most years, "No Award" comes in last in every category: In the years 1980 through 1996 there were a total of 225 categories on the final Hugo ballots; in only 15 of these did "No Award" do better than last place, coming in ahead of a total of 28 nominees (out of about 1150). The only instances of "No Award" doing better than fourth place during this period were second place for Fanzine in 1986 (when there was an organized campaign urging people to vote for it), and third place for both Fanzine and Campbell Award in 1987.
Table 12 — Percentage of Ballots with "No Award" Voted Above Last Place (1980, 1989)
Anyway, for the two years for which I have data, Table 12 gives the percentage of voters in each category who voted "No Award" ahead of anything else. The highest figure is for Dramatic Presentation in 1980, which as it happens was the last year when "No Award" came in higher than last place in that category (it beat The Black Hole). But usually only about a quarter of the voters in a typical category use "No Award" at all. (As with bullet-voting in Table 11, the higher numbers tend to be for the fan categories and the Campbell Award — plus Dramatic Presentation in this case.)
The Business Meeting attendees tend to be among those who want to see "No Award" win more often, so they keep coming up with gimmicks to increase its chances. Most recently, they have presented us with a rule (Sec. 2.9.3) mandating a one-on-one playoff between "No Award" and whatever wins by the conventional calculation. This has been in effect for only the last four years (1993–96), but I also have the equivalent numbers for 1989–90. Table 13 gives the results for all six years. As you can see, "No Award" never came close to beating the winners. This should hardly be surprising, since "No Award" comes in last nearly every time, usually losing next-to-last place by a substantial margin. And if it can’t even beat the least popular of the nominees, why would one expect it to do better against the winner?
Table 13 — One-on-One Preference Votes Between Winners and "No Award"
|Novel||548–129||282–45||523–41 & 475–37 (tie)||265–51||386–62||469–43|
|Fan Artist||299–111 & 298–84 (tie)||178–34||215–64||162–33||194–67||262–59|
To see why this happens, we need to look at the dynamics of Hugo voting. In fact, a fair-sized minority of voters often do vote for "No Award" in first place. But (as we’ve seen in Table 12), typically three-quarters of the voters vote it last or not at all, and — the key fact — hardly anyone votes for it in second or third place. Thus, even when "No Award" is in the running for first place, it picks up very few votes as other nominees are eliminated, and practically every other surviving nominee gains on it. So it has no chance of winning unless it starts with a substantial lead in first-place votes, and even then the odds are against it. To illustrate this, let’s look at how the voting went in two of the very few cases when "No Award" was in the running to the end (names omitted to protect the losers):
|Fan Writer, 1986|
(In the Fan Writer vote, although "No Award" had the most first-place votes and was the last nominee eliminated for first, it ultimately came in last, losing 320–244 to nominee D for sixth place!)
So the "No Award"–playoff rule is unlikely to have any effect on the results. Its principal effect, as with most of these epicycles added to the rules, is to create more work for the Hugo administrators — and thus to increase the probability that some inexperienced administrator will misinterpret the rules, something that has happened too often already. (Indeed, in 1991 the Hugo results were reported using a different proposed version of this rule, which at that time hadn’t even been adopted. By the way, most of this section is adapted from a manifesto that I distributed — to no avail — at that year’s Business Meeting.)
Does it really hurt an author to have two stories on the ballot in the same category? Many writers firmly believe that their chances of winning a Hugo are reduced if they have two stories on the ballot in the same category. This doesn’t seem very logical a priori, since on a preferential ballot the presence of one story shouldn’t affect a voter’s ranking of the other story. On the other hand, it’s true that in the history of the Hugos such an author has won only once (Nancy Kress in 1992). The last time I encountered an argument about this on-line, I decided to research the facts; the results follow.
This situation has come up in the fiction categories 12 /2 times in the 36 years since 1959 (before which there was no nomination ballot, and thus no "nominees"). Before 1980 the vote counts usually weren’t published, so all we have are the lists of nominees, sometimes with a partial order of finish:
For more recent years, the vote counts are available. So as not to bury you in numbers, I’ll give only the counts for the two same-author stories and the eventual winner, immediately before and immediately after the elimination of one of the two; in this way we can see where the votes for the "B" nominees went.
1981 Novella (1015 votes cast):
|Dickson, "Lost Dorsai" (1st)||267||293|
|Martin & Tuttle, "One Wing" (2nd)||175||235|
|Martin, "Nightflyers" (3rd)||154|
(49 of the 154 votes went to other nominees; 19 had no next choice and were thus discarded. By the way, this was the "half" case I mentioned above.)
1981 Short Story (873 votes cast):
|Simak, "Grotto of the Dancing Deer" (1st)||294||303|
|Duntemann, "Cold Hands" (4th)||85||105|
|Duntemann, "Guardian" (5th)||61|
(25 votes went to other nominees; 7 were discarded.)
1988 Novella (838 votes cast):
|Card, "Eye for Eye" (1st)||220||230|
|Robinson, "The Blind Geometer" (3rd)||170||206|
|Robinson, "Mother Goddess of the World" (4th)||91|
(27 votes went to other nominees; 18 were discarded.)
1992 Novella (614 votes cast):
|Kress, "Beggars in Spain" (1st)||166||199|
|Kress, "And Wild for to Hold" (5th)||66|
(27 votes went to other nominees; 6 were discarded.)
1992 Short Story (642 votes cast):
|Landis, "A Walk in the Sun" (1st)||148||175|
|Resnick, "One Perfect Morning, with Jackals" (2nd)||123||143|
|Resnick, "Winter Solstice" (4th)||94|
(37 votes went to other nominees; 10 were discarded. The final vote for 1st place was 267–228.)
1995 Novelette (496 votes cast):
|Gerrold, "The Martian Child" (1st)||101||105|
|Le Guin, "Solitude" (4th)||60||86|
|Le Guin, "The Matter of Seggri" (6th)||56|
(23 votes went to other nominees; 3 were discarded.)
1996 Novella (597 votes cast):
|Steele, "The Death of Capt. Future" (1st)||161||170|
|Le Guin, "A Woman’s Liberation" (2nd)||138||168|
|Le Guin, "A Man of the People" (4th)||68|
(16 votes went to other nominees; 13 were discarded. The final vote for 1st place was 255–240.)
1996 Novelette (619 votes cast):
|Kelly, "Think Like a Dinosaur" (1st)||150||163|
|Egan, "Luminous" (5th)||74||92|
(11 votes went to other nominees; 2 were discarded.)
Conclusions: (1) Most voters appear to vote for individual stories, not for the author, since over half the transferred votes usually go elsewhere. (2) Typically over 10% of the voters for the eliminated story "bullet-vote" to the extent of having no follow-up choice. (3) To say that an author’s stories hurt each other’s chances, we must assume that those who didn’t vote for story A after story B was eliminated would have done so if B hadn’t been on the ballot in the first place (and with another author’s story there instead, as additional competition). This seems improbable to me. (4) And in my opinion most of these stories would have lost even if the authors had withdrawn their other stories. But you can judge that for yourselves . . .
Does the nominee with the most first-place votes always win? I think just about all of you realize that the answer to this question is "No," but you might be surprised at how often it doesn’t happen. Here’s a tabulation of the number of categories each year in which the first-place leader didn’t win:
(In 1989 one leader ended up in a tie for the win.) Your opinion of these results will probably depend on what you think of the Hugo voting system in general.
By the way, it is much more common for the nominee with the most nominations not to win. Here is the corresponding table for the nomination leaders. (In some years the counts for individual nominees were not announced.)
People vote quite differently in an open field than when they have a list of nominees set before them. (One factor here may be that the nomination leader often benefits from a bloc vote.)
So how well does the Hugo voting system work? Well, it depends on what you want a voting system to accomplish. I’m not going to write a treatise on electoral systems (which could easily be longer than the rest of this article), but will merely note a few significant points. For any election with more than two candidates, it can be shown that no electoral system will give an optimal result (as defined by the satisfaction of the voters) all the time. Nevertheless, some systems do work better than others . . .
The straight single-vote, plurality-win system used in most U.S. real-world elections has the virtue of simplicity: everyone understands how it works. But when no one has a majority, the candidate with the largest minority wins, even if that candidate is despised by the majority of the electorate. (An effective two-party system helps to overcome this defect, but has defects of its own; usually the minority-winner problem is merely transferred to the primaries.) And the results are distorted, because voters must make judgments like "Do I vote for A, who is my real preference; or for B, who has the best chance of beating the candidate I hate?"
The Hugo preferential-vote system, in contrast, asks for more information about the voters’ preferences, and thus is more likely to yield a satisfying result. Voters can vote according to their real preferences, without fear that this will help to produce a result they dislike. And the elimination system guarantees a choice by majority vote in the final round. But the system still isn’t perfect. In particular, candidates still need a significant number of first-place votes to be in the running: If nominee B is everyone’s second choice, it might be the most satisfactory winner; but in the Hugo system it can’t win unless it has enough first-place votes to avoid early elimination. Furthermore, the system is sufficiently complicated that a fair number of voters just don’t understand it — which of course makes it hard for them to feel comfortable about the results.
Now there are systems (which I won’t attempt to describe) that take account of even more information about the voters’ preferences, and thus are even more likely than the Hugo system to give an optimal result. Unfortunately, they are also more complicated than the Hugo system. This means that such a system would be very hard to explain to the voters, not to mention the difficulty of getting it through the Business Meeting in the first place. (There is an inherent problem in translating mathematical formulas into legalese: the two languages have different grammars.)
So the better a system is in the abstract, the harder it is for people to understand it. If I knew a way around this dilemma, I’d be engaging in real-world politics instead of this stuff . . .
Oh, yes: And when do the ballots come in? The best data on this are from 1980, when I picked up the mail every day and kept a running tally. The nomination ballots went out at the end of November (resulting in some people nominating before the year was over), with a postmark deadline of March 15. Here are the week-by-week counts of ballots received:
|Dec 30–Jan 5||12|
|Jan 27–Feb 2||15|
|Feb 24–Mar 1||34|
(11 ballots postmarked after the deadline were rejected; one valid ballot spent three weeks in the mail and arrived too late to be counted.) The final ballots were mailed on May 1, with a postmark deadline of July 15:
|Jun 29–Jul 5||89|
|Jul 27–Aug 2||7|
(32 late-postmarked ballots were rejected; the Aug 11 ballot came by sea mail from England, and was just in time to break a tie.) The striking thing about both tables is that over half the ballots came in within two weeks of the deadline;
the final ballots had a secondary peak of people who sent them in right after they arrived. While the absolute numbers weren’t as high in other years, everything I’ve heard suggests that this distribution over time is typical.
I think I’ve about run out of things to say. (By the way, the last four sections were added since 1988.) I think that voting counts of any kind are fascinating, in that they’re about as close as one can come to direct measurement of how people think. I hope that I have managed to interest you too, or at least not totally bore you.
Appendix — Site-Selection Voting
While I was working on the 1988 version of this article, Bruce Pelz asked me if I could compile a tabulation of site-selection voting figures. The two subjects seemed to go together well, so I decided to publish them together; this still seems appropriate. The total vote counts have already been included in Table 1, but these depend a lot on how many (and which) candidates are running. Of greater interest are the actual voting counts, to which this appendix is devoted. (It would of course be nice to have a compilation of the Hugo voting counts too, but that would take a fairly large volume.)
The present site-selection system, with both mail and at-convention balloting, has been in effect since 1972; before that the voting was done only at the Business Meetings (with a couple of exceptions for overseas cons). The last serious contest under the old system was in 1969 (at St. Louis), when Boston defeated Baltimore 169–119 for the 1971 Worldcon; as it happened, that was my first Business Meeting. (Note that the at-con voting continued to take place at the Business Meeting until the mid-’70s.)
As in the Hugo discussion, the years listed below indicate when the voting took place (with the voting sites in parentheses). Note that the 1971 through 1988 Worldcons were chosen two years in advance, and those since 1989 three years in advance. During the ’70s the numbers released were somewhat spotty, so I’ll start by discussing each year separately.
1972 (Los Angeles):
109 mail, 147 at-con ballots; total 256. At the last minute New York withdrew, and Washington was declared the unanimous winner; no voting counts were released.
Melbourne 330, Los Angeles 45; total 375. (If any separate mail and at-con figures were announced, I can’t find them. Here and for the next few years, I don’t have any numbers for "No Preference," write-ins, etc., so the "totals" are probably low.)
|1st ballot||2nd ballot|
This was one of the only two times in the mail-ballot era when a second ballot has been required to determine the winner (1990 was the other one). I assume the extra 32 votes on the second ballot came from what I call "other" votes below.
Orlando 298, New York 111, Washington 91, Philadelphia 28; total 528. I don’t have the mail/at-con breakdown, except that the great majority of votes were by mail.
1976 (Kansas City):
1977 (Miami Beach):
884 votes cast. Brighton overwhelmingly defeated New Orleans; vote made unanimous without revealing vote counts.
Boston defeated Baltimore, but the breakdown wasn’t released; the "No Preference" votes were announced as for the hoax bid of Flushing. Here’s what was released:
|Boston or Baltimore||379||584||963|
(The invalid ballots mostly lacked the voting fee — probably because the ballot neglected to mention that one was required. As far as I know, there haven’t been anywhere near as many invalid ballots in any later year.)
From 1979 on, complete voting counts have been released, and since 1980 the WSFS Constitution has required their publication. In the tables that follow, the totals for "No Preference," "None of the Above," write-ins, and (sometimes) void ballots have been combined into a single "other" line to save space; they’re usually negligible except when there’s an unopposed bid (or one with no effective opposition). No further commentary should be necessary.
1984 (Los Angeles):
1986 (Atlanta), for 1988:
1986 (Atlanta), for 1989:
1988 (New Orleans):
1990 (The Hague):
(The "other" column comprised problem ballots that were counted separately; I presume they included both mail and at-con ballots.)
1993 (San Francisco):
1996 (Los Angeles):
I note that every site-selection vote since 1981 (even in Australia) has had more at-con than mail ballots, and that the typical ratio seems to be getting more disproportionate in recent years. Is this due to an increasing distrust of the mails?
This suggests that my 1980 counts by date of receipt are no longer very relevant, but I might as well include the data anyway; the ballots went out on May 1, with a nominal mail-in deadline of Aug 15:
|Jun 29–Jul 5||39|
|Jul 27–Aug 2||31|
(1 mail ballot arrived after the con.) The major peak was of course at the Hugo/preregistration deadline.
For the at-con distribution of votes, we do have data from some recent years:
The first three averaged out pretty steadily to about 15% on Thursday, 36% on Friday, and 50% on Saturday. But in 1995 the breakdown was 22%, 36%, 43% (earlier voting because of the more central location?); while in 1996 (with a less central location) there was a shift in the opposite direction, giving 11%, 27%, 62%.
[You know, the 1988 version of this thing had only 4 1/4 pages; this version had 14 pages as printed in 1994, and in this update has gone over 15 .]