Some statistics on the voter’s paradox

Recently, I wrote a post about the voter’s paradox in relation to GE 2015. In that post, I explained what the voter’s paradox is and how it might partly explain the results. After collecting some data, here are the statistics on the responses I got.

1. Did you vote during GE 2015

A total of 421 participants took the survey. 381, or around 90% of those who took the survey voted. For participants who did not vote, they were directed to the end of the survey. All statistics henceforth only took into account participants who voted.

2. Did WP or SDP contest in your ward?

Of the 381 participants, 196 answered that either WP or SDP contested in their ward. The remaining 185 or 49% had other opposition parties contested in their ward.

3. In this question, I asked those who answered yes to the previous, which party – WP or SDP – contested in their ward.

Most participants (74%) said that WP contested in their ward. This is to be expected since WP contested in more wards than SDP. As you can see, 134 + 46 = 180 which is less than the 196 who answered yes previously. Not everyone answered all questions (some dropped out halfway I assumed), and so in case you are wondering why some numbers don’t match, that is why.

5. For the who answered that either WP of SDP contested in their ward, I next asked them to rank their preference, amongst WP, SDP and PAP.

As you can see from the table, 68, or 40% ranked WP as their top choice. 69 ranked PAP as their top choice and 33 ranked SDP as their top choice.

6. For the 185 who did not have WP or SDP in their ward, I asked them to rank their preference for WP, SDP, PAP and XXX (whichever is the party that contested in their ward).

165 participants answered this question. XXX represents that party that contested in their ward. The table gives a detailed picture of their rankings. I want to focus on the pie chart that shows the percentage of people who ranked their top party choice. As you can see from the pie chart, 35% and 21% of the participants ranked WP and SDP as their top choice, respectively. That is 56% of 165 participants. On the other hand, only 20% ranked the party that contested in their ward as their top choice. Therefore, it seems, at least based on this data set, that some people would have preferred WP or SDP over the party that contested in their ward. It is possible that some of this 56% voted PAP over XXX but would rather have preferred WP or SDP.

This provides some evidence to support the theory of the voter’s paradox that I proposed. That PAP may not be the top choice for some, but because WP and SDP did not contest in their wards, PAP is favored over whichever party contested in their ward. That is, if they did not have the chance to vote for WP or SDP, they would rather vote for PAP.

7. Lastly, I asked everyone who voted, to rank their party preference across the 9 political parties involved in GE 2015, assuming that all parties are contesting in their ward. They do not have to rank all parties. Participants have the option to stop whenever, if they feel that they only know enough about certain parties to be able to rank them.

Here, I focus on the last pie chart. A total of 285 participants ranked their top choice. Of the 285, 40% ranked WP as their top party with 28% preferring PAP and 22% preferring SDP. No statistical analysis were ran and so we cannot infer if 40% is statistically significant. However, if we extrapolate from this limited dataset, it is possible that if we change the voting structure to be more national and less wards based, we might see a different pattern of results.

What is the make up of this group like?

Where I got my participants?
This is important because it affects the make up of the data set, which may bias the data to some extent. I posted the link in these places: 1) my facebook wall 2) any party that has a facebook page. Some smaller parties do not have a page that allows you post comments, so I was not able to post a link up 3) I also posted the link on several related groups including SMRT feedback, Fabrications about PAP, Channel News Asia and the facebook page of Calvin Cheng. I did this once every 2-3 days for a week or so.

Some caveats:
1. I asked people about their preference, not who they voted for because I wanted their decision to be private. Therefore, the assumption here is that their preference rankings do in fact mirror their voting decision.

2. This survey was done after the elections and the whole episode on RP and Kenneth Jeyaretnam may have shifted the rankings, especially on RP.

3. Participants were not randomly picked to complete the survey. More opposition supporters may have done the survey compared to PAP supporters, biasing the results.

4. From the age group pie chart, we can see that a majority of the participants are in their 20s and 30s. Therefore, the results may be more representative only for this age range. We do not know if older people think the same way as the younger generations do.

This survey is a simple exercise I undertook to test a theory of mine. There seems to be at least some support for it. However, a much larger survey (approx 1400 participants) by QUAD before the elections took place showed that 70% of the Singapore population would have voted for PAP, and that was in fact how the results panned out. I may be wrong after all. I urge readers to exercise caution when interpreting the results here, especially when only descriptive statistics are presented, with no formal analyses done. As I mentioned, there are several limitations to the data set as well. Regardless, for those who participated in this survey, thank you!

If there are any topics you want me to write about, from a psychological perspective or not, do not hesitate to let me know what you have in mind. If it sounds fun, I will be up for it.

Where are we headed? Regressing towards political prejudice in Singapore

As I was doing some last minute readings before my big exam on Monday, I came across a classic reading on intergroup conflict, some parts of which I thought fit considerably with what Singapore’s political climate is evolving into. Here is a quote from the paper that I thought is especially relevant to Singapore:

“Shared goals may promote intergroup conflict in another way when they are combined with shared values. When two groups are pursuing the same goals or outcome (including nonmaterial goods with positive values such as world peace and democracy), the potential for competition is enhance”.

I lifted this quote from Brewer’s classic paper in 1999, “The psychology of prejudice: Ingroup love or outgroup hate”. This quote, I feel, succinctly sums up the current state of Singapore’s political landscape. One group consists of PAP supporters while the other is made up of the opposition supporters. Both groups are essentially working towards common goals – towards a better nation, towards democracy, and towards a more compassionate Singapore. But is it also ironic (sadly) that even though they are moving in the same direction, instead of greater cooperation, you see both groups becoming more divisive, prejudical and discriminatory towards each other (calling each other dogs on facebook is one such example). Brewer’s argument is that when we have a superordinate goal or goals, social comparison becomes more salient, facilitating derogation and conflict. The process of social comparison (which is just one of the many other alternatives) that creates conflict and hate can be especially compounded in political settings as highlighted by Brewer:

“When groups are political entities, however, these processes may be exacerbated through deliberate manipulation by group leaders in the interest of mobilizing collective action to secure or maintain political power. Social differentiation provides the fault lines in any social system that can be exploited for political purposes.”

This schism is becoming more apparent in GE2015 compared to GE2011, at least in my opinion. On the side of the political parties, you see gutter politics through and through. On the side of the citizens, you see more facebook groups and maybe blogs popping up mostly for the sake of debunking the credibility of the incumbents or the oppositions, more intense name calling (e.g. dogs and IBs), telling people to leave the country and really, just putting each other down whenever and wherever possible instead of engaging in proper political discourse. Both groups are basically trying to positively differentiate themselves from each other, showcasing their group’s political prowess and acumen, at the same time devaluing the other group’s abilities (Brewer’s optimal distinctiveness theory). In the end, what happens is a perception of politics being a zero sum game instead of one where both parties can stand to gain something. This is exactly the social comparison argument made by Brewer.

However, when stripped to their core, both groups are actually more similar than they are different. Even DPM Tharman himself alluded to this during the post-elections interview, when he mentioned that the policies of both the PAP and the oppositions overlap substantially if we disregard their superficial polarity. The route may diverge somewhat along the way but the final destination is always to make Singapore a better place for all Singaporeans. If this is indeed the case, we have to make a conscious effort to minimize intergroup differences and recognize that cooperation is not only feasible but possibly pareto optimal (to borrow a term from economics).

However, based on current trends, I definitely see us regressing into a state of political prejudice and disarray instead of progress into the future. What we get eventually may well be a politically impoverished Singapore, which will likely simultaneously shred apart, or at the very least chip at our social and economic foundations. What is left, if any, will be an indelible mark of counterfactual regret that if only we had done better.

Brewer, M. B. (1999). The psychology of prejudice: Ingroup love or outgroup hate?. Journal of social issues, 55, 429-444.

Brewer, M. B. (1991). The social self: On being the same and different at the same time. Personality and social psychology bulletin, 17(5), 475-482.
(optimal distinctiveness hypothesis)

Poverty is detrimental to mental functioning

I recently read an article illustrating one negative consequence of poverty. The results might or might not come as a surprise to you but the implication for social policies in Singapore and anywhere else could be more than we can imagine.

What is the paper’s main idea?
The paper suggests that people who live in poverty are constantly trying to tackle many of life’s problems, especially those related to financial sustainability. Because of a myriad of difficult financial decisions (where every cent counts), mental resources are used to ensure prudence (e.g. Should I buy more rice and less vegetables? If I spend money on a soft drink, do I have enough left for lunch?). However, because mental resources are limited, if the bulk of it is apportioned to difficult financial decisions, this leaves an insufficient amount to other problems (e.g. how does photosynthesis work?).

How did the authors test their idea?
The authors ran two studies to test their idea. Below is a more detailed explanation of the studies involved

Study 1:
Participants are shoppers recruited in a mall. The median income is $70,000 USD a year with the lowest at $20,000. Participants are randomly assigned to one of two conditions: Hard or Easy. In both conditions, they are presented with hypothetical scenarios (e.g. “Your car is having some trouble and requires $X to be fixed. You can pay in full, take a loan, or take a chance and forego the service at the moment… How would you go about making this decision?”), all of which are meant to induce thoughts about their finances.

In the hard condition, the numbers are high (i.e. $1500 to fix the car) while in the easy condition, numbers are low (i.e. $150). For the easy condition, because the numbers are easier to compute, it should trigger less concerns about one’s finances for both the rich AND poor. However, the hard condition should make monetary concerns more salient for the poor BUT NOT the rich.

While thinking about how to solve the financial problems, they also had to complete two tasks thought to measure cognitive function. They are the Raven’s Progressive Matrices and a spatial compatibility task.

What did study 1 show?

Click on image to enlarge

For the poor and the rich in the easy condition, they did equally well on the cognitive tests. However, if you are poor and you are given higher numbers to compute in the scenarios, you perform poorer in both tests compared to the rich, presumably because you are more concern with the financial decisions you have to make compared to the rich.

Note: There are actually 4 experiments in this study but I describe only 1 of it. All studies are similar in design with minor changes such as extra cash for correct answers. All studies show similar results.

Study 2
This study is more naturalistic, in the sense that there wasn’t any artificial manipulation imposed on the participants. Participants in this study are sugarcane farmers from Indian. During the annual harvest, farmers receive income from sales of their produce. However, because they are unable to distribute their income well enough to last the year, they tend to be poorer just before harvest and richer right after harvest. This natural variation in income thus allows the researcher to test variation in cognitive performance as income changes.

The cognitive function tests are the Raven’s test (as in study 1) and a numeric version of the Stroop task. The farmers perform the tests pre-harvest (when they are poorer) and post-harvest (when they are richer). How did they fare?

Click on image to enlarge

Similar to study 1, during the pre-harvest (poorer) period, the farmers’ performance was worse than the post-harvest period (richer). Pre-harvest, farmers are less accurate in their performance, took a longer time to react to a reaction time task and committed more errors in the reaction time task. This is after controlling for other important variable such as nutrition, work effort and stress levels.

What are the implications for Singapore?
We have heard of how long and complicated the forms can be for the poor when they are trying to apply for welfare benefits. And they have to do it pretty often. With so much cognitive resources needed, the poor might feel discouraged, eventually deciding to give it up instead. Is there a way to make application for benefits more streamline then? There could be positive ripple effects with just a simple tweak in the current system.

Additionally, even with equal education opportunities and subsidies for education, are they sufficient? We often see cases where poor teens/young adults (usually the oldest in the family) have to juggle between school and part-time jobs because they don’t have enough to eat and go to school. With so much cognitive resources spent on having to work and worry about money issues, what is left for learning in school is, frankly speaking, not much. What can we do to ensure that not only everyone gets an education, but also an education that is being disrupted as little as possible by their social situation. I have no answers for this complex issue but what I can offer is a platform to start talking about it.

Mani, A., Mullainathan, S., Shafir, E., & Zhao, J. (2013). Poverty impedes cognitive function. Science, 341(6149), 976-980.

In search of an answer to the voter’s paradox

For those of you who read my previous post, you will know that I tried to put forward a theory of sorts to explain the election results.

So the curious me decided to come up with a quick 1 minute survey to see how accurate my theory might be. If you are also curious like me, please help me fill up this quick survey below and share it with others. Thanks!

https://uky.az1.qualtrics.com/jfe/form/SV_2s2QUU7FBcrOxuJ

I will post the results when and if I get enough participants.

Update: I need about 150 to 200 more participants. Appreciate it if you can send the link to other fellow Singaporeans.

A theory on why the oppositions suffered a “bad beating” in GE 2015

Those of you who were expecting the opposition parties to have won more seats than GE 2011 in yesterday’s election might be surprised, or even shocked that the PAP won almost every seat available, save for Hougang and Aljunied. Some might have attributed the loss to the passing of LKY in galvanizing support for the PAP. Some may have reasoned that PAP’s gerrymandering worked. Others thought that the Town Council saga could have played a role (Aljunied and Hougang support for WP fell). No doubt all of these factors -and many others- likely contributed to the election results in complex ways.

I too was wondering what the reasons might be. However, the researcher in me craved for a more “academic” rationale. So I decided to do some digging of my own on voting behavior. That was when I stumbled upon a fascinating concept called the voting paradox – also known as Condorcet’s paradox. Below, I will briefly explain this paradox and try to apply it to our election results.

Condorcet’s paradox.
In order to understand this paradox, I will need to explain the term transitive and I will do so with an example below.

If you prefer apple to orange and orange to cherry, then by the property of transitivity, you must prefer apple to cherry. So your preference will look like this

Apple > Orange > Cherry

If you prefer apple to orange and orange to cherry but you prefer cherry to apple, then you are violating transitivity. When you do, your preference will look like this

Apple > Orange > Cherry > Apple

Assuming that preferences are transitive, the paradox goes like this…Imagine there are 3 voters in a really small Singapore. V1 V2 V3. There are 3 candidates they can choose from, A B and C. Below is how they rank their preference

V1: A > B > C
V2: B > C > A
V3: C > A > B

Now lets break this down a little.

Lets look at the preferences of the 3 voters in 3 scenarios;

1. A vs B

V1: A > B > C
V2: B > C > A
V3: C > A > B

You will see that V1 and V3 prefers A to B while V2 prefers B to A. A wins because 2 votes to 1.

2. B vs C

V1: A > B > C
V2: B > C > A
V3: C > A > B

Here, B wins because V1 and V2 prefers B to C.

3. A vs C

V1: A > B > C
V2: B > C > A
V3: C > A > B

In this last scenario, C wins because V2 and V3 prefers C to A.

Now you will ask, and soooo???? Well, this is a problem because if A beats B and B beats C, by the rule of transitivity, how can C beat A? This is a paradox because even if individual preference for candidates is transitive, the preference of society when added together may not be, which is puzzling and paradoxical. This is one problem with democracy’s method of majority voting with multiple parties involved.

Paradox applied to elections
How can we apply this to the Singapore context which has a GRC system and in which usually only 2 parties compete in a GRC? The GRC system actually coincides really well with this paradox I feel.

Think back on the number of times you hear people wished that “I have SDP or WP come to my ward to contest because I will definitely vote for them”. Let us now assume that the someone who said this has NSP in their GRC instead of WP. This is an indication that their preference for parties might be something like this

WP > PAP > NSP

If WP contests in their ward, they will vote for WP. But if the opposition is NSP, they prefer PAP instead. This is a valid assumption I think because the sentiments seem to be that even though we have numerous oppositions, only 1 or 2 of them are credible. The other parties seem to be weaker and if the decision is between PAP or the weaker parties, PAP is higher in the preference hierarchy.

Therefore, what the paradox shows is that when there are more than 2 parties, it may be unwise to make paired comparisons, to avoid mistakenly assuming that a chosen alternative is preferred to one that is not chosen. By considering all alternatives together instead, we might find that society has no clear preference for a party.

However, the GRC system forces us to make a choice, usually only between 2 parties, the PAP and another opposition party. But if your preference is as above, then you will vote for PAP instead because even though you want an opposition in parliament to ensure a check on PAP, you prefer WP to NSP. However, because only NSP is contesting in your ward (and you don’t think NSP is good enough for now), you will vote for PAP because you cannot vote for WP.

Of course, this is just a theoretical idea I thought was counter-intuitive, yet interesting, for thinking about why the results turned out as they were yesterday.

P.S.
Our psychological perception of groups might have played a part as well. For example, seeing all opposition parties as a single entity instead of separate parties may have resulted in voters grouping weaker parties with the stronger ones, diminishing the credibility of the strong, because we do have a negativity bias – the tendency to be more vigilant about and remember negative cues. As the saying goes, one bad apple spoils the bunch.

Update.
I decided to test the accuracy of this theory. If you are curious like me, please follow this link to complete a short 1 minute survey. Additionally, you can also share this new post which has a link to the survey as well. Thanks! I will post the results when and if I have enough participants.

Making a case for greater distributive justice in Singapore.

If you are, like many others, keeping abreast of the political happenings in Singapore, you should have heard the viral speech by DPM Tharman about taxes and the “bad-ass” reply by our Ex-GIC economist Lam Keong Yeoh. Tharman basically implied that taxing the middle class is inevitable in order to help everyone, especially the poor. LKY (not Lee Kuan Yew) is saying we can do more – or greater economic redistribution, especially to the poor and underprivileged without being fiscally draining.

Singapore is, broadly speaking, fiscally conservative as you can imagine. Our distributive policies are mediocre at best (in my opinion) but it seems like since GE 2011, there have been improvements (as LKY mentioned in his bad-ass reply). But we definitely can do more.

There have been calls for greater redistribution to the lower-incoming citizens but at the same time, others are hitting out against unemployment and welfare benefits. Their argument is that we either have to increase taxes (as per Tharman) or “plunder” from our reserves to make this happen, both of which are economical dangerous. LKY in his reply has made an economic case (I leave it up to you to decide if it is convincing or not) that this doesn’t have to be so. That is, we can increase social protection while still being fiscally responsible.

Here, I want to offer up a moral case for redistribution. To do this, I like to engage you in a thought experiment. Humor me a little. Some of you may find this familiar at the get go and if you do, please bare with me for a little bit.

Imagine yourself being temporarily stripped of your citizenship and status (social, political and every kind imaginable) in Singapore. Tabula rasa. Everyone of you will now enter into a new social contract with each other and decide together on a set of principles that can serve as the basis for a new Singapore. However, before you re-enter into the new Singapore, you will have to randomly pick a ball from a box (blinded) that will determine your new identity. That means if you were once a doctor, you could now be a factor worker. If you were Chinese, you could now be Indian. If you were poor, you could now be rich. Your past can be entirely different from your future. Knowing that you will have no clue of your new identity, how will you decide on the principles that are the foundation of Singapore2?

Yes, this thought experiment is based on Rawls’s original position and veil of ignorance.

Rawls argued that because we are deprive of information as to our new identity, essentially behind a veil of ignorance, selection of principles will be guided by rationality and impartially, as a result of our risk aversion. Because of this impartially, we should derive principles that will 1. ensure equal liberties 2. ensure fair equality of opportunities (jobs, education etc) and 3. regulation of inequalities ensuring a fair deal for the worst-off.

For those of you who are against more redistribution, take a moment to think hard about this thought experiment. Please think about how much we are doing to help the less fortunate right now. Do you think we are doing enough? Is the current state acceptable to you when you are behind the veil? Lam Keong Yeoh’s remarked that “Many Middle class members in developed countries don’t mind paying higher taxes to have the security of a higher level of social protection and mobility as well as to look after the less privileged” really struck a chord. Our middle class is being taxed too little. The middle income should, and I believe we are willing to, be taxed more.

The middle class may be taxed more, but the social changes that occur may in fact leave us happier as a nation than if we kept to the status quo. Why? Because money is not the only measure of happiness. In fact, many psychological studies have provided evidence that this is the case. According to Baron Layard, “Family and personal life come top in every study, and work and community life rank high. Health and freedom are also crucial, and money counts too, but in a very specific way”. I also happen to believe that a happier country will become a more productive country, with the (unintended) effect that our economy will naturally grow with it.

This post is not an economic barrage for increasing taxes in Singapore. Regardless of whether we need to increase taxes, we have a social and moral responsibility toward distributive justice which is frankly speaking, sorely lacking in Singapore. I may not have convinced you just yet but I am hoping to have tugged you toward me, even if for just an inch or two.

Cheers

“Vote wisely”. How potential cognitive pitfalls can bias your decision making process in the Singapore election

Vote wisely.

This is the popular political mantra everyone – regardless of their political affiliation – has been advocating for the last week or so (Some with implied bias). But what does wisely mean? Perhaps it means to deliberate about the policies of each party before deciding. In other words, to be rational and unbiased, and to not vote for the oppositions or the incumbents just because you don’t like them. But we know from Herbert Simon’s theory of bounded rationality that humans are not fully rational beings but rather bounded creatures. We are not optimizers but satisficers because we are unable to digest the vast amount of information available to make the optimal decision. Instead we often make decisions that are satisfactory.

This idea I would say, almost single-handedly resulted in the blossoming of a subfield in psychology called Judgment and Decision Making (JDM). And one of the cornerstones of JDM is research in cognitive biases as a result of our bounded rationality. Cognitive biases affect many of our decisions in life and not surprisingly, it also affects our political decision making process.

Hence, in the midst of the political fever, it might be “wise” to be aware of the potential biases that might affect (your) thoughts about different parties and policies and subsequently who gets elected into parliament.

1. Conservatism or Bayesian Inference
What is it?
The tendency for people to insufficiently revise their beliefs in light of new evidence.

Example
Your initial preference for minimum wage may be positive, but when others present you with data on the harmful effects of minimum wage, you might not adjust your preference for minimum wage because you already believe that it is good. This works both ways. If you initially believe that minimum wage is a bad policy, new evidence about how good it is might not sway you or barely nudge you at all. For you to make a better informed decision, be open to all policies and make sure to adjust your beliefs based on evidence.

2. Confirmation bias
What is it?
The tendency to search for and interpret information that conforms to one’s preconceptions

Example
If your initial position on the CPF is that interests should be higher than the current rate because the government is using it to earn higher interest elsewhere, you will find yourself going to places/websites that confirm your views. By allowing yourself to fall into the bias, you essentially prevent yourself from being open to be persuaded otherwise because every incoming information you find will only confirm your initial position. I am not saying that your position is definitely wrong. I am saying you are not allowing yourself for the possibility that there are better arguments against this position.

Confirmation bias can also lead people to perverse conclusions depending on how the question is being framed. Imagine that the decision in this scenario is of giving custody to either Parent A or B. Parent A is remarkably mediocre in remarkable ways. Parent B has great strengths but also apparent weaknesses. In a study, when asked which parent should be GIVEN custody of the child, people tend to choose B. However, when asked which parent should be DENIED custody, people tend to deny B custody.

The logic here is that when framed as giving custody, the parenting skills of parent B become salient because it aligns with how the question is being framed. You want to give custody to a parent who has some parenting skills. However, denying of custody is a negative frame, resulting in the weaknesses of parent B becoming more apparent while undermining the good qualities of B.

We can see how parents A and B can be easily substituted with the words party or policy. When we evaluate two parties, decisions about whether you should give your vote to one party or deny your vote to one party can affect voting behavior, especially for swing voters.

3. Above average effect/Illusory superiority
What is it?
The tendency for people to overestimate their abilities and performance, relative to others.

Example.
This example does not relate directly to voting behavior but I brought it up because I think it clouds our perception in important ways which may bias the way we think of others, especially foreigners in the case of PMETs.
How many times have you read about how people always feel that companies are hiring FTs that are “less qualified and less hardworking than Singaporeans”? Some of the cases might be true but some of it may be attributed to this bias.

For example, how did you assess your ability against your competitor? Were you able to look at his/her CV. Were you able to compare levels of industriousness? Were you present at the interviews? Could it be the case that because your competitor is not local, he/she is mostly probably less qualified than you? Maybe your competitor attended a university you have never heard of? These points however are not sufficient assessments of ability and qualification.

Just how pervasive is this bias you might think? Well, I recently read an article where the authors were investigating this effect in prisoners. In the experiments, prisoners were asked to rate themselves on a series of traits compared to other prisoners and non-prisoners:

The traits are: morality, kindness to others, trustworthiness, honesty, dependability, compassion, generosity, self-control, and the ability to abide by the law.

What were their ratings like?
“Surprisingly, they rated themselves superior to the average community member on all traits as well, with one exception. Prisoners considered themselves as law-abiding as the average community member”.

Maybe we should give priority to locals before offering jobs to FTs but we should not make and generalize the claim that we are better than them in every other way based on anecdotal evidence. That is wrong. This only creates animosity, dividing people into us and them, eventually leading to stereotypes, prejudice and discrimination.

4. Fundamental attribution error
What is it?
The tendency for people to make dispositional attributions (personality) with regards to other’s behavior while at the same time discounting or even disregarding situational contrains placed on the person.

Example
Unemployment benefits is something I want highlight. I see many people on facebook commenting on policies related to welfare/unemployment benefits, especially the bad of it . For those who are against it, their argument is that those who claim benefits are in fact lazy and not motivated to work because if they are, there are plenty of jobs out there. Because people who claim welfare must be lazy, it is not a feasible idea.

This epitomizes what FAE is about. If you are unemployed, you must be lazy because there are so many jobs out there. Situational constrains like illness, family members who are ill and so need constant care, therefore restricting ability to work are not taken into account or discounted. While it is true that some people are in fact lazy, it is not true that ALL of them are lazy.

This in no way speaks of the economic viability of unemployment benefits. It speaks of our psychological tendency to think that people who needs unemployment welfare are dispositional lazy. This is obviously not a neutral example because I am advocating for some form of unemployment benefits. However, the crux is that we should not vote on policies because we think we understand the character of someone who is unemployed.

There are of course many other biases that can influence the way you think about parties and policies but here I highlighted 4 that I think are more ubiquitous, more likely to impact on people and potentially easier to recognize and avoid.

NHST vs Bayesian approach: Which side are you on?

One of the assignments in my methods class is writing up a short 2-3 page report on your views about the current debate over the limitations of our orthodox statistics and the push for a new approach – Bayesian statistics. Where do you stand on this? Is the new method superior to NHST? Why or why not? What are the strength and weaknesses of both methods of inference. This is one of the rare occasions where I am actually excited about writing a paper for class. Also, if you are an ISCON (International Social Cognition Network) on facebook, you can tell that the discourse is nothing short of vibrant. After writing the paper, I felt a strong urge to share it with interested others, and so below is a copy of my class paper:

Putting NHST on trial

Should Null Hypothesis Significance Testing (NHST) be declared obsolete, to be overtaken by the Bayesian approach? While the discourse between both statistical approaches has a long history, their antagonism became more prominent at a time when discovery of scientific frauds, questionable research practices and failed replications have caused quite a stir in the field of psychology. My stand on the whole issue follows the suggestion of Gigerenzer & Marewski (2015): There is no such thing as the ideal method to statistical inference. The better approach is to think of the various statistical devices we have as a “toolbox”, and the tool should be appropriate to the problem at hand.

One reason why I think (and rightly so) NHST has been receiving a heavy dose of criticisms is the disconnect between the conclusions we want out of it versus the conclusions we can get out of it. What researchers want to know is the probability of the theory being true given data, P(Theory │Data). However, what NHST tells us is the probability of data being true given the theory P(Theory │Data) – the inverse conditional probability (Dienes, 2011). The only conclusion we can get out of NHST is whether we can reject the null or not. But researchers tend to see a significant p value as evidence for the alternative hypothesis. And because Bayesian statistics require you to pit one hypothesis against another, you are able to assess and adjust your confidence in your hypothesis against the competing hypothesis (null), instead of being able to only evaluate the null. This is a very convincing and logical argument for me in favor of Bayes over NHST.

However, one of the primary reason why there is so much resistance against Bayesian statistics is the setting of a prior. For people against Bayes, their argument is the subjective nature of priors. How a prior should be set is the biggest concern for most people. The Bayesian counterargument I have read against this concern is that essentially, this should force researchers to develop a deeper understanding of the theory behind the phenomenon in order to set an appropriate prior before testing it. While I agree with the Bayesian camp, I think there will be times when even with theory, it is not easy to set up a proper prior because of reasons such as which manipulation is used and what dependent variable is being measured. For example, you can get different effect sizes if your prime is subtle versus supraliminal or if your dependent variable is the noise blast task versus the hot sauce paradigm.

In such cases, the Bayesian default alternative can be a way to get around the uncertainty. The Bayesian default alternative is a distribution rather than a point estimate of a prior, say Ha: d~N(0,1). However, Uri Simonsohn in a recent blogpost (http://datacolada.org/2015/04/09/35-the-default-bayesian-test-is-prejudiced-against-small-effects/) made a case that this default approach is biased against small effects. When d=.28, the default approach supports the null 20% of the time compared to 5% when d=.64. Given that psychology mainly deals with a small to moderate effect size, the default alternative might lead to higher false negative rates.

Perhaps all is not lost and some compromises can be made. The Fisherian framework explicitly states that null hypothesis testing can be useful when there is little information about we are trying to study (Gigerenzer, 2004). It seems acceptable then that we should be able to use null hypothesis testing when doing exploratory analyses or when devising a new method to study an old phenomenon? When doing more confirmatory research (e.g. Minimal group paradigm), perhaps a good first step for psychology is to report both p values and Bayes Factor. In cases where contradictions occur, it might be wise to do a high powered replication of the study. As Lakatos elucidated in his idea of a scientific research program (1976), when a researcher encounters refuting evidence, assumptions about the “protective belt” can be revised without having to declare that the research program is dead based on a single failure. Only with multiple failures over time should such considerations be made. Similarly, when a single study finds evidence for a phenomenon, or fails to replicate, whether under the NHST or Bayesian approach, it is not instructive to make a definite claim about it. Therefore, we should be mindful not to equate finding positive results using the “better” Bayesian statistics to proving a theory true. As Gigerenzer & Marewski (2015) aptly puts it, “proponents of the “Bayesian revolution” should be wary of chasing yet another chimera: an apparently universal inference procedure”.

Dienes, Z. (2011). Bayesian versus orthodox statistics: Which side are you on?.Perspectives on Psychological Science, 6(3), 274-290.

Gigerenzer, G. (2004). Mindless statistics. The Journal of Socio-Economics,33(5), 587-606.

Gigerenzer, G., & Marewski, J. N. (2014). Surrogate Science The Idol of a Universal Method for Scientific Inference. Journal of Management, 0149206314547522.

Lakatos, I. (1976). Falsification and the methodology of scientific research programmes (pp. 205-259). Springer Netherlands.

What do you think?

Methods class

Hi everyone

My first random musing will be about today’s methods class.

Today, we had the honor of having Mark Schaller skype with us during class. The topic for this week was about collecting data outside the lab. If you know who Mark Schaller is, you will know that he has a line of research investigating pathogen prevalence. How do you manipulate pathogen? Well, you can, but it is not easy, which is why he has had papers using data from outside of the lab. One of his lesser known but nonetheless fun paper is on the psychology of fame. In the paper, he argues that the attainment of fame can lead to chronic self-consciousness. How do you study famous people? You cannot possibly invite them to the lab to participant in your study for a paltry $5. One clever way of measuring their self-consciousness is by examining the use of pronouns in their lyrics/monographs/diaries before and after they turn famous. I am not going to go into the details but go have a look if you interested.

What was memorable for me however was not his paper and his experience with collecting data outside of the lab. Rather, it was his way of studying a psychological phenomenon. Many graduate students – me included – come into graduate school with tons and tons of ideas (hopefully). Where do they all come from. Many times, we simply pair an independent variable with a dependent variable that we found in some article and viola, you have study 1. When the pressure to publish gets into you, there is less thought going into the theoretical understanding of a phenomenon and more into what variables I can use to make this study work. I’ll admit, up till now, at least 60% of my ideas came about as a variant of what I just described.

But Schaller says to hold our horses. When we have a research question in mind, the first step is not what variables I can use to test my claim. What we first need is a thorough understanding of  the theoretical framework supporting our claim. The theory is the backbone of our hypothesis, something that should not be taken lightly. Even more so, it should not be after the fact. That is, you do not work the theory around the data you have. This is not what science is about. Similarly, it is plain wrong to tinker with your researcher degrees of freedom to fit the data into your theory. Once we pass the theoretical hurdle can we then move on to the methods. Even during the design of the study we have to be mindful of variables selection. With so many measures out there, which is the best. The best is not the most common or popular ones. The best should be the one that most appropriately answers the hypothesis. Sounds simple but when the momentary excitement of an idea hits you, theory often takes a back sit and variables are being mindlessly abused.

Rushing into the study design is a bane, not a boon, and I am guilty as charged. After today’s class, I am going to keep reminding myself not to fall into the temptation of doing a sloppy job in theory formulation and methods design. Time to kick my old habit away.