Archive for the ‘statistics’ Category

Storytelling can be patented

typewriter

credit: mpclemens via flickr(cc)

We already knew that computers, equipped with proper algorithms, could write stories pretty much indistinguishable from the work of professional journalists working under deadline pressure. And had I been paying attention, I’d know that the company behind this, a “spinout” from Northwestern University, is also moving into other “turn data into a story” tasks, which from the examples here seem to mainly focus on financial reporting, tho it also appears that buyers of used cars can be exposed to “automated and individualized vehicle stories” (pdf) about their cars, which presumably helps sales. And it’s no secret that In Q Tel, an affiliate of your Central Intelligence Agency, is one of several investors behind the company.

So, it’s technology, it’s government, it’s marketing– why am I surprised that it’s protected by a bunch of patents on different variations on “automatic generation of a story?” Here I am, using a computer with many automatic functions to generate a sort of story about this company, and I really haven’t time to read and try to understand all their patents. I guess I better stop before I get in more trouble.

h/t Crain’s.

More propaganda I am too slothful to review

Graffito image by Horia Varlan (cc) via flickr

Graffito image by Horia Varlan (cc) via flickr

A new report “Copyright Industries in the U S Economy” has been released by the IIPA (A conspiracy of seven associations of copyprivilege holders).  I should read and review it, but I could not do a better job than Mike Masnick, so read his review and the comments thereon.  Of course, IIPA and its members probably have several staff and/or automatons, whose duties include responding to constructive comments. Fortunately, they get responded to in turn.

Like the man said

“In order to preserve and enhance jobs, exports and economic contributions, it is critical that we have strong legal protections for U.S. creativity and innovation in the U.S. and abroad.”

Which means creators need to be free to create, with strong legal protection against those who would try to prevent their use of ideas which may have been touched by others.

I’m from the government and I’m here to deceive you

image credit: A Mina via flickr (cc)

image credit: A Mina via flickr (cc)

They call it “Factors Influencing Voluntary Compliance by Small Businesses[pdf],” and the report comes from IRS (actually “an independent organization within IRS,” whatever that means), and  so you know that it’s referring to compliance with Federal income tax laws and regulations.

The focus on small business makes sense, since folks on payrolls or pensions generally can’t hide much of their income, and large corporations can obtain legislative action or legal advice providing their own loopholes.  Two surveys were done, one of small businesspeople nationwide for whom, statistically, an audit would be expected to result in additional tax payments, and the other of small businesspeople in communities from which a high proportion of such returns was filed. (NOTE: When wooing their votes we call them “entrepreneurs” or “job creators,” but that would not be dainty when we are considering them “tax cheats.”) These “low compliance” folks were supplemented by a survey of those expected to be “high compliance.”

I see two surprising results in this report.  First, on most attitude measures there’s little difference between the “low compliance” and “high compliance” respondents.  For example, only 15% of the “low” group thought that federal tax laws were fair, and the same proportion of the “high” group agreed.   Differences that did appear were fairly small. “Wealthy taxpayers minimize their taxes in ways the average taxpayer cannot” was agreed with by 74% of the “low compliance”‘ group, and 69% of the “high compliance” people.

An even bigger surprise, to the IRS analysts as well as this blogger, is that “low compliance” seems to be associated with greater participation in their local communities, including churches, schools, volunteer organizations, and even were more likely to vote than “high compliance” people. I hesitate to speculate on what this means, but it is probably a positive for those of us who believe (along with most respondents) that federal government is not an appropriate way to deal with many of the areas it has become involved in.

The report acknowledges that the survey suffers from an indirect method: “Low compliance” merely indicates a statistical likelihood of same, not an actual lack of compliance by the respondent.   Because a random selection of taxpayers are audited simply to calibrate the compliance-prediction model, it would have been possible to target these particular taxpayers for the survey.  Of course they couldn’t be surveyed after their audits because they might be especially unhappy with IRS at that point.  But they could have been surveyed just before being notified on the audit.  IRS decided not to do that because they “deemed it overly deceptive.”  However, the actual survey (done over the phone by a contracted private company), did not reveal that this was an IRS project until the conclusion, after the data had been gathered.  That apparently was deemed just deceptive enough.

Not part of the published report is a list of “clusters of potential tax cheats” posted by AP but presumably originating with IRS. Apparently tabulated by zip code, here’s the Illinois portion of the list:

Bellwood, Calumet City, Dolton,  Grand Crossing, Hazel Crest, Matteson, Maywood, Ogden Park, Phoenix, Riverdale, Roseland, South Chicago Heights, South Holland, University Park.  I assume that folks in Oakbrook and Barrington Hills had already purchased their own loopholes.

How the CPI values our privacy– or doesn’t

wallet contents

image credit: ItalianPsycho via flickr (cc)

Measuring inflation isn’t an easy matter; I don’t think many of us can even agree on exactly what inflation is. But it’s somehow related to prices consumers pay, and the biggest operation measuring that, at least in the U S, is the Bureau of Labor Statistics’ Consumer Price Index. There are lots of issues with how they measure, and how they report, but one which I haven’t seen discussed regards “loyalty” cards and the value of privacy.

Many retailers in recent years have set up “clubs” of one sort of another to better track their customers. One of the noteworthy exceptions until last year was Walgreens, and there remain a few other respectable merchants, but nowdays one who chooses not to participate ends up giving up a fair amount of money (or convenience).  So does the Consumer Price Index recognize that privacy has a price?

The answer, at least conceptually, turns out to be straightforward.  It’s right here in Chapter 17 of the Handbook of Methods. As described on pages 31-32 of this pdf, if a discount is offered only to card users, is temporary (lasts no longer than a month or two), and at least 50% of the sales are at the discounted price, then the CPI recognizes the discounted price. (If the discount is for a longer period, then pro-rata weighting is used.)  In essence, if most people give up privacy for a discount, then the CPI doesn’t recognize this as a cost.

At least in the stores I visit, it seems that most folks are quite willing (or economically constrained) to take the discount; anyone who wants to make a purchase under the policies that were in place twenty years ago needs to pay a higher price, which is not recognized in the Consumer Price Index.