see also

Things I notice:

  • teachers and nurses are not low earners. that would be cooks, cashiers, janitors, customer service reps, secretaries, housekeepers, and teachers’ & nurses’ aides. And the people who are “below” those job-titles, as well.
  • "confidence intervals": if you click on any job-title in the NPR link, it highlights all occurrences of that job-title in yellow. Some job titles span wide across several income categories (truck drivers, sales supervisors) whereas others (software developers) are more concentrated/narrow in quantile.
  • vagueness of data: Data data data. Data. By the way, did I mention data? There’s a lot of it, you know. But check it out: measurement entails the creation of categories and sometimes those categories are so vague that “sales supervisor” shows up in the $12k–$26k level, whilst “salesperson” shows up in the $207k+ category. Wait, aren’t the supervisors supervising those salespersons? Turns out, “salespersons” show up ≥ “sales supervisors”, who show up ≥ “sales”. Oops.
  • histogram: look at the distance between incomes for each (constant-size) decile bracket. +12, +9, +5, +8, +8, +10, +14, +31, +104, +∞. This matches Catherine Mulbrandon’s picture:
    where the latter leaps capture the convexity of US earning power.
  • jobs you can just get: Many of the dark-teal jobs are ones with a low barrier to entry. I call these “private-sector welfare”. The jobs that, if able-bodied I found myself unemployed and needed money, I could apply and believe that I actually have a chance of being hired—as opposed to the less-than-1% response rate from sending CV’s to large corporations for salaried office work.

The Australian state of Victoria implemented the world’s first … sin-tax … hypothecated for health in 1987. It came in the form of … a 5% levy on tobacco products … whose revenue was then used to fund a newly formed independent health promotion foundation called VicHealth.

Apart from increasing cigarette prices, the legislation banned most tobacco advertising and formed the basis for later rules to create smoke-free workplaces and public venues.

Meanwhile, VicHealth bought-out all tobacco industry sponsorships of the arts and sports. This proved less costly and easier than anticipated, as most preferred non-tobacco sponsors. Among the foundations other activities are more than AUS$ 20 million annually in funding for health research and in support of anti-smoking and other public health campaigns.

Until 1997, all of these activities were funded exclusively from the hypothecated tax on cigarettes. Since then, the hypothecation aspect has been weakened as states are no longer allowed such tobacco levies. However, tax funding from the national level from sin-taxes and others is transferred to states to compensate.

Prior to the Victorian tobacco legislation, a survey found 47% of respondents in favour of an increase in tobacco taxes (including 20% of smokers). If hypothecated for health or other community benefits, this support surged to 84%.

To retain such support and realise the benefits in terms of accountability and public trust the hypothecation must be strict, i.e. no topping-up from general taxation and no siphoning off to other purposes.

Ole Doetinchem

Beyond the government-finance issues of hypothecation I’m just so fascinated that smokers want to raise taxes on (only) themselves. It’s unsurprising from real-life experience but does not fit into the standard microeconomics utility theory.


High modernist subjectivity gives an extraordinary privilege … to judgement and especially to cognition…. The modern predominance of reading….

High [modernism] … furthermore … privileges the cognitive and moral over the aesthetic and the libidinal, the ego over the id, the visual over touch, and discursive over figural communication.

…the individual [is] somehow ‘closed’ instead of open; to be somehow obsessed with self-mastery and self-domination.

Lash, S. & Friedman, J. (Eds.). (1993). Modernity & Identity. Massachusetts: Blackwell, pg. 5

via writingcapital



Carey Davenport


Statistics vs Machine-Learning
by Rob Tibshirani


The Hunt of the Unicorn tapestry, dyed with weld (yellow), madder (red), and woad (blue).


Big, long cycle = trend.

Hello R people.

In December of 2013 I posted a cheap-o wiki-editable (thank you github) contact list which recruiters can use to find you, if they’re looking for R programmers.

In what I consider a resounding success, within a few weeks it got onto the first page of google (thank you github), and within a month or two it was the first result in Google.

So I would say this is the best (also to my knowledge the only) place to put your name if you want to be found for R work.

Posting this again because some peoples’ situations may have changed, and others may have not seen the first notice. Also because a little bird told me about a recruiter who wants to hire people for full-time ggplot work in London asap.

I’ll be checking github for pull requests in the wake of this posting, to make sure your details don’t linger in github limbo.