Regression model for the coming election
Posted by Possum Comitatus on May 16, 2007
After gathering together the newspoll data from 1985 to the present and aggregating those into a monthly series, I thought I’d build a quick election model regression.The series used are a mix of newspoll series and economic series sourced from the RBA.These were:
GOVPRIMARY: The primary vote for the government according to the newspoll.
OPPRIMARY: The primary vote for the Opposition according to newspoll
PMDISAT:The dissatisfaction rating for the Prime Minister
PMDISAT(-1):The dissatisfaction rating for the Prime Minister lagged by one period.
OPSAT: The satisfaction rating of the Leader of the Opposition
GST: A dummy variable for the GST where it equals 1 for the period it has been operating and zero for the periods before it was introduced.
CAMP: A campaign dummy variable which equals 1 the month before the election and zero otherwise.
INT: The standard bank variable home loan interest rate.
AR(1):A first order serial correlation termm
MA(1):A first order moving average error component.
The model became:
GOVPRIMARY = 64.8207675 + 1.374899313*INT – 0.565567558*OPPRIMARY – 0.128756676*PMDISAT + 0.05746403116*OPSAT – 0.1212621617*PMDISAT(-1) – 1.720345869*GST + 2.518176146*CAMP + [AR(1)=0.8682231246,MA(1)=-0.6409529542,BACKCAST=1996:03]
Using the quaint little Eviews, the more important technical bits were:
| Dependent Variable: GOVPRIMARY | ||||
| Method: Least Squares | ||||
| Date: 05/17/07 Time: 00:40 | ||||
| Sample: 1996:03 2007:05 | ||||
| Included observations: 135 | ||||
| Convergence achieved after 36 iterations | ||||
| Backcast: 1996:02 | ||||
| Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| C | 64.82077 | 3.467005 | 18.69647 | 0.0000 |
| INT | 1.374899 | 0.470391 | 2.922884 | 0.0041 |
| OPPRIMARY | -0.565568 | 0.075075 | -7.533329 | 0.0000 |
| PMDISAT | -0.128757 | 0.028567 | -4.507237 | 0.0000 |
| OPSAT | 0.057464 | 0.024617 | 2.334312 | 0.0212 |
| PMDISAT(-1) | -0.121262 | 0.027496 | -4.410223 | 0.0000 |
| GST | -1.720346 | 0.743948 | -2.312453 | 0.0224 |
| CAMP | 2.518176 | 0.735535 | 3.423599 | 0.0008 |
| AR(1) | 0.868223 | 0.067859 | 12.79458 | 0.0000 |
| MA(1) | -0.640953 | 0.115439 | -5.552327 | 0.0000 |
| R-squared | 0.818619 | Mean dependent var | 43.11037 | |
| Adjusted R-squared | 0.805559 | S.D. dependent var | 3.360798 | |
| S.E. of regression | 1.481958 | Akaike info criterion | 3.695793 | |
| Sum squared resid | 274.5250 | Schwarz criterion | 3.910998 | |
| Log likelihood | -239.4660 | F-statistic | 62.68402 | |
| Durbin-Watson stat | 1.847315 | Prob(F-statistic) | 0.000000 | |
I spent alot of time checking for all sorts of relationships, but what was interesting was the CAMP result.The Coalition get a 2.5% boost to their primary vote in the month before the election and the GST hurt there primary vote and it has not yet recovered.
Another interesting phenomena that I came across was that the Coalition appears to get a boost in their primary vote when economic bad news is occurring, and a decline in their primary vote when the economic sunshine comes around.When the Coalition is in government:
- a 1% increase in unemployment has walked hand in hand with a 0.96% increase in their primary vote.Likewise a 1% decrease in unemployment has walked hand in hand with a 0.96% decrease in their primary vote.
- a 1% increase in the interest rate level has walked hand in hand with 1.3% increase in their primary vote.Likewise a 1% decrease in the interest rate level has walked hand in hand with 1.3% decrease in their primary vote.
When the Coalition is in opposition:
- a 1% increase in unemployment has walked hand in hand with a 1.6% increase in their primary vote.Likewise a 1% decrease in unemployment has walked hand in hand with a 1.6% decrease in their primary vote.
- a 1% increase in the interest rate level has walked hand in hand with 0.4% increase in their primary vote.Likewise a 1% decrease in the interest rate level has walked hand in hand with 0.4% decrease in their primary vote.
When times are tough, the people run to the Coalition and economic sunshine kills them.See for yourself:
| Dependent Variable: COALITION | ||||
| Method: Least Squares | ||||
| Date: 05/17/07 Time: 01:03 | ||||
| Sample(adjusted): 1986:01 2007:05 | ||||
| Included observations: 257 after adjusting endpoints | ||||
| Convergence achieved after 8 iterations | ||||
| Variable | Coefficient | Std. Error | t-Statistic | Prob. |
| C | 27.46739 | 3.446556 | 7.969518 | 0.0000 |
| DGOVLIB*UNEMP | 0.960052 | 0.419546 | 2.288308 | 0.0230 |
| DGOVLIB*INT | 1.325942 | 0.437369 | 3.031630 | 0.0027 |
| DGOVALP*UNEMP | 1.562745 | 0.296463 | 5.271298 | 0.0000 |
| DGOVALP*INT | 0.385121 | 0.166081 | 2.318868 | 0.0212 |
| AR(1) | 0.671602 | 0.047794 | 14.05187 | 0.0000 |
| R-squared | 0.600228 | Mean dependent var | 44.33366 | |
| Adjusted R-squared | 0.592265 | S.D. dependent var | 3.318514 | |
| S.E. of regression | 2.119009 | Akaike info criterion | 4.362844 | |
| Sum squared resid | 1127.040 | Schwarz criterion | 4.445702 | |
| Log likelihood | -554.6254 | F-statistic | 75.37167 | |
| Durbin-Watson stat | 2.145189 | Prob(F-statistic) | 0.000000 | |
I bet the government is praying for a recession.
Steve said
Does the government do these kinds of analysis or do you think they just look at raw poll data and think “Oh F**K”?
possumcomitatus said
I’m sure the governments pollsters like Crosby-Textor do analysis like that all the time, probably a lot more sophisticated to boot.As for the “Oh Fuck” – I think that’s happening no matter which type of polling analysis the government looks at
netvegetable said
Are you saying that a tory government is less likely to win an election when the economy is doing well?
possumcomitatus said
That’s exactly right netveggie! But its not only Coalition governments, coalition oppositions also suffer from the same allergy to economic sunshine.
The evidence seems to fly in the face of alot of political mythology to the contrary.
Albert F said
Thanks Possum – great work.
I think that govts can use bad economic news to their advantage. I’m not sure its purely a Coalition thing. Keating won the ‘93 election and had the first million plus unemployment rate in history in (I think) the final week of the campain. It seemed to help him as voters were scared of Hewson’s radical reform agenda.
Possum Comitatus said
Hi Albert,
I agree.Historically at least, interest rate levels have walked hand in hand with the Coalitions primary vote with a high degree of statistical significance, although only contributing a small amount of explanatory power in the Coalitions primary vote (between 5-10% depending how you measure it).Historically, bad economic news is good political news for the Coalition particularly.Whether this time that effect will be overpowered by the “Keeping interest rates low” fallout is something that time will tell.
Ash said
“Tory” is a description or label I *&^%ing hate. I am a rusted on Coalition voter and there is nothing Tory about me. I have worked battler jobs, I have put myself through two degrees. I send my kids to public schools. I could say a lot more. Tory is a word that belongs in the class warfare of Britain prior to the 1950s and not in Australia in the 21st century.