30.04.2015 Views

Hela boken - SOM-institutet - Göteborgs universitet

Hela boken - SOM-institutet - Göteborgs universitet

Hela boken - SOM-institutet - Göteborgs universitet

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

A reaffirmation of public opinion<br />

Table 1<br />

Correlations Between Aggregate Opinion Measures Across Time<br />

Lag 1 Lag 1 Lag of Two Lag of Three<br />

(Unweighted) (Weighted) Years Years<br />

Pct Pro .96 96 .91 .83<br />

Pct Con .95 .95 .91 .82<br />

Balance Measure .96 .96 .91 .83<br />

Issue Intensity .92 .93 .86 .80<br />

Pro Intensity .92 .94 .89 .84<br />

Con Intensity .90 .90 .86 .83<br />

Mean .95 .96 .91 .83<br />

Percent Don’t Know .65 .65 .44 .51<br />

Percent Opinionation .87 .87 .84 .78<br />

Number of Cases 245 73 190 150<br />

Comment: Entries are positive product moment correlations between facets of public opinion<br />

measured at different times (autocorrelations).<br />

While the correlations in Table 1 are based on a large number of cases (245), some<br />

of the item pairs appear in several surveys, while others may have appeared only<br />

once in successive years. Only the proposal to “reduce the size of the public sector”<br />

appeared in all 15 surveys. So, it could be argued, these correlations are as large as<br />

they are because of certain stable items being asked in many years and thus such<br />

items may have a disproportionate effect. Therefore, the analysis was repeated<br />

so that each issue was weighted the same. For example, If an issue was asked in<br />

successive years only once, it was weighted at 1.0. If an issue appeared in four sets of<br />

successive years, each instance of that issue was weighted at 0.25. When the analysis<br />

was repeated, with the weighting procedure in effect, the results were essentially<br />

unchanged (see column 2 in Table 1).<br />

Columns 3 and 4 show what happens when the lag is increased from one year to<br />

two and three years. If aggregate public opinion is stable and slowly evolving, one<br />

would expect the two year lag to yield a slightly lower correlation than the one-year<br />

lag, and for the three year lag to produce a slightly lower correlation than the two<br />

year lag. That is exactly what happens. For the balance measure, the correlations<br />

were +.96 for the one year lag, +.91 for the two year lag, and +.83 for the three year<br />

lag. The comparable correlations for issue intensity were +.92, +.86, and +.80 for<br />

the one, two, and three year lags, respectively. Time, as such, cannot cause anything<br />

or any change to occur, but it does give greater allowance for other things to occur<br />

in a dynamic system. The picture then is of public opinion as a stable but evolving<br />

force. The only exception to this type of trend was in the percentage leaving the<br />

item blank where the three year lag yields a somewhat higher correlation than<br />

the two year lag.<br />

429

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!