• Tools of the trade are five – logic, ethics (broadly philosophy), statistics, writing, and skill in using tools for statistics (R), and writing (Latex). Spend time learning the tools of the trade.
• Everything takes time. Spend time wisely.
• Sub-clause: Don’t waste time on projects with no realistic future.
• Model for dissertation: Markus Prior’s Post-broadcast era democracy

In 2007, Cindy Kam, proposed using university staff as one way to move beyond the ‘narrow database’.

There exist two more convenient ways of expanding the database – alumni, and local community colleges. Using social networking, universities can also be used to tap into student, and staff networks of acquaintances.

A common (university wide) platform to recruit, and manage panels of such respondents would provide cost savings, and allow for shared access.

One benign way to understand this shift is the following – attention is a limited resource and only allocated to things deemed important. Once the importance was established, attention followed.

However –

1. Attention is not that limited a resource.
2. Attention cost is marginally minimal given I generally do look at the calorie content.
3. Given I encounter same or similar choice tasks repeatedly –cost per choice, invested once, is close to zero.

What are the reasons behind this seeming conundrum?

In the domain of food, we care about three things – price, taste, and health. Let’s factor out price for now. This leaves us with taste, and health. While making a choice, we recruit ‘relevant’ (more later) information to assess choices in a manner that maximize our utility.

Assume we have a strict preference for taste over health; more narrowly – taste always wins whatever the health information; health information comes into play only when taste is equivalent. Health information in this scenario is immaterial, and one only needs to focus on information about taste to maximize his/her utility. So, one way to explain my inattention to relevant health information is just that.

Expanding from the toy example, orderings of things we care about (utilities) dictate the way we seek information, and what information is sought. However, observation tells us that the ordinal structure of utilities is manipulable in the domain of food. For example, we prefer taste strongly but if health information were to be made salient, we would be liable to choose something healthy. One inference which we can draw from such manipulability or order is that the initial preference ordering must not have been strong. But that doesn’t seem right given our strong preference for taste ‘explains’ subdual of information seeking on health.

Rational choice assumes that if information acquisition costs are zero, more information should always be sought, and used in decision making. Rational choice seems inadequate to the task of explaining, hide and not seek.

Let’s assume we have subconscious and conscious preferences (aside from assuming a subconscious). Subconsciously we greatly prefer taste more than health. Consciously we prefer the reverse. Taste wins if health information is not made salient at the time of purchase. Assuming subconscious controls behavior, health information is deliberately not sought.
Another way to think about underlying preference structure is the following – aside from preferring taste, we also prefer feeling good about our decision. Feelings about a decision are evaluative and emerge from whether we chose wisely given information. So one way to include good feelings is to choose healthy food but that sacrifices our preference for taste. Another way this is resolved is ignoring information about health, which is much more easily ignored than information about taste.

Given this, we suppress health information. Interestingly this suppression doesn’t extend to information seeking about health on all fronts but applies only during decision time about a food.

Another interesting psychological thing to note here is that we have negative affect (feeling bad) associated with decisions that lead to negative long term consequences, but we also have ways to prevent this negative affect pathway from being triggered at all. Additionally the information suppression isn’t a one-time only but long term because we want to repeatedly ‘sin’. This in turn means that we firstly somehow ‘know’ that the food is unhealthy and hence not look at the health information, otherwise wouldn’t it just help boost one of the reasons for consuming something tasty, but don’t consciously acknowledge this information.

Yet another way to think about the problem is to assume that we have preferences for health but they are somewhat lower in order of priority. For lower order preferences (here health), information seeking becomes more passive and increasingly depends on how easy it is to acquire information, for example – how prominently it is displayed. Social desirability pressures may also play a larger role in moderating information acquisition when importance is low. For example, in US people frown upon those who look at labels in a supermarket. Thus cowed, people may be less likely to look up information. Though it was always possible to look up the information once home, and now given ease and convenience of anonymous information gathering (Internet), it is likely that social desirability issues are less of a factor (it is likely that social desirability pressures continue to apply when one is alone.) However in cases where other lower order preferences predict same choice information about them is likely highlighted. For example, if a tasty thing were healthy as well, it is likely that one reminds oneself about the health benefits while making the choice.

But why is taste implicitly prioritized over health? One explanation is that preference for taste is evolutionary – the positive immuno-response from eating calorie rich food is biologically potent. Another is that consequences on health from choosing unhealthy food are long term while gratifications from taste are instantaneous. Given that, it allows us to more readily imagine the consequences of one which in turn is perhaps define one of key ways we decide our preferences. Lastly, the preference for taste in matters of food has become likelier due to advertising and its constant valorization of taste over everything else.

It is still awe-inducing to see to what degree our brain is lazy, and inhibits acquisition of reasonably readily available information.

The above analysis assumed considerations dictating choice at point of purchase. Once we have bought something however another consideration applies – we have invested in x, so now enjoy it. What is point of reading information now that I have already spent money?

A few straight forward policy proposals emerge, given what we know about how people behave, –
• Front of package labeling
• Prominent, easy to read, comprehend, labeling
• Priming aim – be healthy
• Priming habit – look for information when buying food
• Priming consequences

Links -

http://www.wnyc.org/shows/radiolab/episodes/2008/11/14

Note -
The word ‘we’ is in quotes in the title because I do not have data to show how widespread the tendency is.

The party that dislikes the census, put a hold on the nominee for census bureau, is now sending out fraudulent mail surveys that seem as if they were from the census bureau. Read here, here, and here. Accusation for being ‘fraudulent’ stems, not only from the use of `census’, but also from it being an attempt at “frugging”, the practice of cloaking a fund raising appeal in what appears to be a research. (“Suggers” sell using surveys.)

The great burka debate
Assuming ‘God’ has recommended or even ordered that women wear burka, assuming that burka has no impact on a woman’s ability to communicate or quality of life, as has been suggested by its supporters, then here’s a suggestion – to all men, who haven’t been ordered by ‘God’ to wear burka, and who don’t see a downside to wearing it, why not voluntarily commit to wearing the burka, since no law opposes such a voluntary act, to show solidarity with the women. My sense is that even the French would come to support the burka if men en masse chose to wear it.

Copy this
Free photocopy in exchange for ad printed on the back

Making forecasts about various forms of `journalism’ has of late reduced to extending the trend line from recent past well into the future. Such exertions typically produce the following forecast – Newspapers are dead or dying. Both network newscasts, and cable news channels, are shedding audiences, and will continue to do so. Radio is generally ignored.

Continue Reading…

Reports and papers on Deliberative Polls often carry comparisons between participants to non-participants on a host of attitudinal, and demographic variables (e.g. see here, and here). The analysis purports to answer whether people who came to Deliberative Poll were different from those who didn’t and to compare participant sample to the population. This sounds about right, except this – the comparison is made between participants, and a pool of two likely distinct sub-populations – people who were never invited (probably a representative, random set), and people who were invited but never came. Under plausible and probable assumptions, such pooling biases against finding a result.

The key thing we want to measure is self-selection bias – was there a difference between people who accepted the invitation, and who did not. The correct way to estimate the bias would be as follows:
(Participant/Didn’t come) ~ socio-demographics (gender, education, income, party id, age) + knowledge + attitude extremity

Effect sizes can be provided to summarize the extent of bias. This kind of analysis can account for the fact that bias may not occur at first marginals (gender), but at second marginals (low educated females). (This all can be theory driven, or more descriptive in purpose). The analysis also allows for smaller effects to be seen, as variance within cells are reduced.

p values
When the conservative thing to do is to reject the null hypothesis, think a little less about p-values.

Assuming initial survey approximates a ‘representative’ sample of the entire population. Assuming we want inference how ‘representative’ Participants are to the entire population, it makes sense to just report mean differences without p values.

The survey sample estimates stand in for the entire population. Entire population census numbers are without standard errors or very low s.e. so comparisons are always significant.

By comparing to an uncertain estimate of the population one cannot say whether the participant sample was representative of the entire population. That estimation is without bias but suffers the following problem – the more uncertain the population estimate, the less likely one can reject null, and more likely one is to conclude that the participant sample is representative. One way to deal with this is to do the following – Have 95% conf. band of sample estimate of population and then calculate max and min difference of the sample and report that.

Name calling
Calling the analysis- ‘representativeness’ analysis – seems misleading on two counts –

1. While a clear representation question can be answered by some analysis, none such question is answered, and can be answered by the analysis presented. Moreover it isn’t clear if it relates to some larger politically meaningful variable. For example – one question that can be posed is whether participant sample resembles the population at large. For answering such a question, one would want to compare population estimates to census estimates (which have near zero variance, so t-tests etc. would be pointless.)
2. In a series of papers in the 1970s, Kruskal and Moesteller (citations at the end) rightly excoriate the use of `representativeness’, which is fuzzy and open to much abuse.

Citations
Kruskal, W; Mosteller, F. (1979) Representative sampling I: non-scientific literature. Intern Stat Rev. 47:13-24.
Kruskal, W; Mosteller, F. (1979) Representative sampling II: scientific literature. Intern Stat Rev. 47:111-127.
Kruskal, W; Mosteller, F. (1979) Representative sampling III: the current statistical literature. Intern Stat Rev. 47:245-265.

Until recently, the state of the art aggregate media use measures were Nielsen ratings, which put a `meter’ in a few households, or asked people to keep a diary of what they saw. In short, the aggregate measures were pretty bad as well. Digital media, which allows for effortless tracking, and the rise of Internet polling however for the first time provides an opportunity to create `panels’ of respondents for whom we have near perfect measures of media use. The proposal is quite simple: create a hybrid of Nielsen on steroids and YouGov/Polimetrix or Knowledge Network kind of recruiting of individuals.

Logistics: Give people free cable and Internet (~ 80/month) in return for 2 hours of their time per month and monitoring of media consumption. Pay people who already have cable (~100/month) for installing a device and software. Recording channel information is enough for TV, but Internet equivalent of channel – main Domain – clearly isn’t as people can self-select within websites. So we only need to monitor channel for TV but more for Internet.

While number of devices on which people browse Internet, and watch TV has multiplied, there generally remains only one `pipe’ per house. We can install a monitoring device at the central hub for cable, and automatically install software for anyone who connects to the Internet router, or do passive monitoring on the router. Monitoring can also be done through applications on mobile devices.

Monetizability: Consumer companies (say Kellog’s, Ford), Communication researchers, Political hacks (e.g. how many watched campaign ads) will all pay for it. The crucial innovation (modest) is addition of the possibility to survey people on a broad range of topics, in addition to getting great media use measures.

Addressing Privacy concerns:

1. Limit recording information to certain channels/websites – ones which on which customers advertize etc. This changing list say can be made subject to approval by the individual.
2. Provide for a web-interface where people can look/suppress the data before it is sent out. Of course reconfirm that all data is anonymous to deter such censoring.

Ensuring privacy may lead to some data censoring and we can try to prorate the data we get it a couple of ways -

• Survey people on media use
• Use Television Rating Points (TRP) by sociodemographics to weight data.

ANES cumulative file (1948-2004) carries a variety of indices. In creating some of the indices, it appears pre-election measures have been combined with post-election measures in some of the years. If that wasn’t enough, at least one of the times, the same index in some years has pre-election measure combined with post-election measure, while using only post measures in other years. Here’s an example –

‘External Efficacy Index’ (VCF0648) is built out of two items –
http://www.electionstudies.org/studypages/cdf/anes_cdf_var.txt

Item 1: Public officials don’t care much what people like me think.
Item 2: People like me don’t have any say about what the government does

Item 2 is asked both pre and post election in some cycles. In 1996, efficacy is built out of -

960568 (pre), 961244 (post)

http://www.electionstudies.org/studypages/1996prepost/nes1996var.txt

[you can ID post-election wave questions through the following coding category - Inap, no Post IW]. Post version of 960568 is 961245

While in 2000 it is built out of – 001527 (post) ,001528 (post)

http://www.electionstudies.org/studypages/2000prepost/anes2000prepost_var.txt

I have alerted the ANES staff and it is likely that the new iteration of cumulative file will fix this particular issue.