Wednesday, January 23, 2008

Why the Major Dailies Make Lousy Line Dancers and Things Opinion Polls Tell Us about the Fukuda Administration

I compared the latest public opinion polls from Asahi, Mainichi, Yomiuri, and Sankei-FNN.

Polling Dates
Asahi (January 11-12)
Mainichi (January 19-20)
Yomiuri (January 12-13)
Sankei-FNN (January 13-14)

Nothing spectacular happened on the political scene between January 11 and 20, so I assume that the numbers are roughly comparable. Now look:

Fukuda administration
Asahi : support 34%, not support 45%
Mainichi : support 33%, not support 45%
Yomiuri : support 45.6%, not support 41.6%
Sankei-FNN: support 36.6%, not support 47.3%

Support political party
Asahi : LDP 26%, DPJ 25%
Mainichi : LDP ??%, DPJ ??%
Yomiuri : LDP 35.5%, DPJ 16.9%
Sankei-FNN : LDP 32.1%, DPJ 25.0%

There’s the Who do you want to win, which yields better results for the DPJ. But as I always say, watch the trends, not the numbers. So take a look at this:

2007: July 24-26, Abe administration
2008: January 12-13, Fukuda administration

2007: support 36.5%, not support 51.8%
2008: support 45.6%, not support 41.6%

Support political party
2007: LDP 32.2%, DPJ 25.6%
2008: LDP 35.5%, DPJ 16.9%

The 2007 July poll was the last one Yomiuri took before the July 29 Upper House general election that resulted in a landslide victory for the opposition.

I am somewhat skeptical of the conventional wisdom that the Japanese electorate uses Upper House elections to punish/reward the LDP while it plays it safe in Lower House elections. Still, it is reasonable to infer from the latest numbers that the DPJ has a ways to go before they can be favored to gain a plurality in the Lower House election. It’s not enough to have the LDP look bad. On the other hand, it’s highly likely that the LDP-New Kōmeitō coalition would be substantially diminished if Prime Minister Fukuda called a snap election now. So where’s the hurry for the LDP?


Janne Morén said...

The exact way the questions are phrased, alternatives presented and answers tallied and analyzed makes a major difference. Unless two newspapers go out of their way to make sure they are both polling in the same way, it's not going to be comparable. For instance, the questions "Which party do you support", "which party will you vote for in the 2009 election" and "which party would you vote for if an election was held today" will give you three different results.

Looking at trends from the same source is better, but only works if you actually confirm that the two polls where conducted in the same way; the above stands for that case as well.

And I really, _really_ wished these dailies would give us the confidence intervals (or equivalently, the margin of error); presenting results like "51.8%", as if the data really was significant to a tenth of a percent, is frankly more than a little misleading.

Jun Okumura said...


As a general proposition, what you say is correct. But I've watched the polls long enough to be confident that each one is quite consistent within the general trend line and consonant with major political events.

I agree that it would be marginally helpful if they published the confidence intervals (maybe they do; remind me to dig up the January 14 hardcopy Yomiuri that I still have lying around somewhere. My guess is, they settle for 95%. But garbage in, garbage out.