Election 2016 Social Desirability Bias / Trump Shaming: Obama Approval Rating
Social Desirability Bias influences politics in many ways. Though ‘Trump Shaming’ has received the most attention, this issue as it relates to politics is much larger. It directly impacts support for the different political parties, policies, and politicians to a significant extent. In fact, it is fairly surprising that it has not been covered before now.
Here, we look at Obama’s approval rating. Using a very similar method as used in other posts, we take the net approval rating for Obama in live polls during a month and subtract the same as calculated from anonymous polls. In a perfect world, there would not be any significant difference between live and anonymous polls. If the live polls tend to be higher, we can infer there is an element of Social Desirability Bias occurring in favor of Obama. If the net live – anonymous poll data tends to trend around important events and news coverage, we can further infer that there could be an element of causality.
In Obama’s case, there appears to have been two major trends in this indicator implying the existence of Social Desirability Bias in his favor. The first occurred around his major foreign policy engagements of opening up relations with Cuba and in negotiating a nuclear deal with Iran. The result was relatively positive media coverage for Obama and his policies which were not matched by the support of the populace. Obama’s approval rating in live poles trended higher against anonymous poles while coverage of these major news items continued and then started to drift back down as coverage lessened.
An important point here is not simply that the bias began around important shifts in policy but that the bias had a limited life span. It seems that, for politics, people will have a tendency to change their answer to a socially more acceptable one as long as it is on the top of mind – which requires in most cases reinforcement by the media. Once the media move on to another topic, the impact starts to fade.
The second occurred directly after Obama began his aggressive attacks against Trump. A sitting president going after a presidential candidate is fairly rare and Obama doing so is newsworthy. These high-profile attacks began in January 2016 at Obama’s last State of the Union address. Almost immediately, the gap between live and anonymous polls began to increase implying an increasing Social Desirability Bias in favor of Obama. Implied in these numbers are that Trump had been branded as socially undesirable by the media and Obama aggressively attacking Trump would result in a benefit to Obama whereby his approval ratings received a superficial short-term bump. Again, this bump will likely last as long as Obama continues upping his attacks and as long as media coverage continues to be biased against Trump. In August 2016, Obama flatly stated that Trump is ‘unfit to serve’ and that Republicans should withdraw their support of his candidacy. Topping this might be difficult, but very possible given the environment.
These larger trends can be seen in the following chart.
Chart 1: Obama’s Net Bias Approval Rating, as calculated by Obama’s Approval Rating from Live Polls minus Obama’s Approval Rating from Anonymous Polls, 12 month moving average
Source: RealClearPolitics
In a normal environment, Obama’s Social Desirability Bias might average around 3.0% to 3.5%, using 12 month moving averages. He is fairly well liked by the media and coverage tends to be more supportive. Furthermore, he has strong support within different demographic groups which likely puts social pressure on others within those groups to publically support him though they might not equally support him in private. A 3% positive bias seems fairly acceptable for a historically important president – if you like him or not, you should recognize that history books will be mentioning his name for many centuries to come.
When jumping above this supposed norm, however, there is presumably a catalyst. At the very beginning of the chart, the bias was likely a bit higher due to his reelection. It has been shown that people like supporting winners in elections, even if they did not vote for them. If you ever want to get a laugh look how consistently polls show a higher percentage of people having voted for the winning candidate in the previous election than actually did. Likewise, you will see the percent of people admitting to voting for the losing candidate to be even lower, with an unusually high ‘don’t recall’ category. In short, it is not surprising that Obama or any president would receive a net positive bias soon after winning an election.
Then, the bias settles into its presumed longer term norm by declining to around 3.3% on a 12 month moving average. It remains bouncing around this level until the second half of 2014 when it starts to spike higher. The timing coincides with the Republicans unexpectedly winning control of the Congress in the 2014 mid-term elections in November 2014. Obama seems to have benefited in live polls through a positive bias thanks to an assumed spike in positive coverage for Obama and negative coverage for Republicans winning control of the Congress. The reverse of this phenomenon can be seen in the bias against Congress deteriorating sharply at the same time (covered elsewhere).
Additionally, around the same time Obama came out with somewhat surprising initiatives. He pushed hard for an agreement with Iran over nuclear negotiations and in opening diplomatic relations with Cuba during primarily the second half of 2014 and into 2015.
As the data in the previous chart is smoothed using a 12 month moving average, some of the movements are hard to identify. Next is a chart that is unsmoothed.
Chart 2: Obama’s Net Bias Approval Rating, as calculated by Obama’s Approval Rating from Live Polls minus Obama’s Approval Rating from Anonymous Polls, no moving average
Source: RealClearPolitics
Although more choppy, the unsmoothed data is more exact. You can see the bias really beginning to take off from the start of 2016, or when Obama gave his SOTU speech. This was his entrance into the race, not as an observer but as an active participant. Though not the norm for a lame-duck president to become as active as he has in influencing the race, his attacks have been generally accepted by the media. And, combined with the media’s negative bias towards Trump has created an almost super-charged Social Desirability Bias.
With the media extremely positive on Obama and extremely negative on Trump to an extent far in excess of the average viewpoints of the populace, it has created an environment of excessive bias. In fact, Obama’s positive bias (as measured by his net bias approval rating) stood at 10.5% in August 2016, the highest recorded.
In other words, Obama’s approval rating measured in live polls averaged 10.5 percentage points higher than the average as measured in anonymous polls. This is simply astounding. You might shrug your shoulders assuming that it will not play into the election and Obama is out of office soon anyway, so what difference does such a large bias mean anyway? Actually, people are using this information to make decisions so it makes a very large difference. Many political analysts point to Obama’s approval rating as an indicator pointing to a Clinton victory. The argument goes that a follow-on candidate should be elected if the two-term successor is very popular.
Well, looking just at anonymous data which in the current environment is far more accurate, Obama’s approval rating is about -1%. In comparison, his approval rating using live polls is about +10%. This swing is enormous and is the difference between a president who is very popular within his own party and one who is just plain popular. Assuming the former is true, Clinton does not have an obvious advantage in the general election as most assume.
Additionally, such an exaggerated approval rating, even if based on an artificial bias, likely has impacted Obama’s own actions. Though there is no evidence that he follows his approval rating or that it would impact his actions, we can be fairly sure that he or one of his staff has put together his rising approval ratings coinciding with attacks on Trump. Again, there is no evidence, but it seems given the lack of precedence of former lame-duck presidents attacking candidates and his generally less aggressive personality, that he might not have ratcheted up his Trump attacks to such a degree if he was not receiving positive feedback through an increasing approval rating. Now, whether he understands that 10.5 percentage points of this can be explained by bias is another story. The point is that once Social Desirability Bias reaches such a high level it produces a lot of collateral damage and modified perceptions of reality.
Lastly, the bias shown to exist in Obama’s Approval Rating helps to support the larger concept of Social Desirability Bias within Politics. Also, the fact that the abrupt increase in the bias coincides with attacks on Trump further supports the idea of a negative bias against Trump, without using any Trump specific data.