Cathy O’Neil’s Weapons of Math Destruction, How Big Data Increases Inequality and Threatens Democracy, is a terrific and important book. 

O’Neil has the credentials and the cred. Her Ph.D. in mathematics is from Harvard and she subsequently taught at Barnard. She took her analytical skills to D. E. Shaw, a hedge fund and then to other private sector organizations.

O’Neil sees the risks in overuse, especially of opaque mathematical models, not so much from a technical perspective, but because of the economic and political power of such models. She also sees the risks of how using such models and analyses creates feedback loops, especially vicious cycles that disadvantage the disadvantaged. And she’s acutely critical – rightly so in my judgement – of analyses and measurement systems that may be technically sound, but are far from conceptually sound. 

Example: measurement of teacher performance based on how their students do on high-stakes, non-routine tests. O’Neil’s explanation is not mere rant. Because she understands the technical and conceptual framework, she sees and helps the reader see how fragile and unreliable (thus, unfair) the system is. (And, I would add that, as a matter of public policy, how counterproductive such systems are.) O’Neil tells the story of Tim Clifford, an experienced teacher whose model-based score in one year was 6 out of 100. Protected by tenure, but baffled by his dismal score, he continued teaching the same way. The following year his score was 96 out of 100. The arithmetic might have been right. A measurement system that produces those kinds of swings is dangerous to rely on.

(Note that education is not my domain, but health care has been and I’ve observed that education policy makers could and should have paid attention to the lessons about measurement systems that health care folks learned with a 20-30 year head start. Perhaps I’ll expand on that one of these days.)

Don’t assume that because O’Neil is very much the mathematician, that she can’t write. She can. She has a nice, accessible style and uses storytelling well to make her points. And anyone who goes by the moniker, mathbabe takes her work, but not herself too seriously.

I’d put this book in a group with Michael Lewis’s Moneyball, Nate Silver’s The Signal and the Noise as eminently readable books on modern analysis. Lewis is about “why,” Silver’s mostly about “how” and O’Neil is about “be careful” about using today’s tools.

Don’t just take my word for it. Here’s Evelyn Lamb’s review in Scientific American.

For the lay reader who wants to know what some of the downside are, for the policymaker who wants to measure performance, but wants also to be fair and responsible, and for the data maven who’s concerned about the ethical parameters of their work, Weapons of Math Destruction is a must read.

Data WMD O Neil

{ Comments on this entry are closed }

Jealous?

by John W Rodat on April 7, 2017

Hey, data mavens:

Jealous?

JWR License Plate

{ Comments on this entry are closed }

The US Senate changed their rules today, requiring only a simple majority to end debate over Supreme Court nominations. This so-called “nuclear option” reduced the votes required from 60 percent plus one to 50 percent plus one. The change was accompanied by much drama. 

It’s probably fair to assume that a similar change will come regarding legislation. (That’s already the case for Budget Reconciliation.)

Though, there’s much hue and cry, in the long run, this will make the Senate less undemocratic. As Senate seats are allocated two for each state, people in states with smaller population have proportionately more representation in the Senate that people in other states.

With the 60 percent rule, Senators (from the smallest states) representing about 11 percent of the country’s total population could block action in the Senate. With a simple majority (plus one), that number increases to about 17 percent. It’s still undemocratic, but less so.

Public Signals States Able to Block Senate Action

{ Comments on this entry are closed }

A key consideration in the current fracas, but which is not part of the discussion is that the effect of Medicaid on county finances in New York is quite variable.

Here are some graphics for background

County Cost of Medicaid Trend

 

Medicaid as Percent of Total Expenditures

 

Medicaid as Percent of Property Taxes

 

Medicaid as Percent of Sales Taxes

 

Medicaid as Percent of Sales Taxes Net of Distributions to Cities Towns Villages

 

{ Comments on this entry are closed }

I’m analyzing NYS options if the AHCA becomes law including the amendment pushed by Congressmen Collins and Faso. I’ll post that later today.

For some context and history, it will be useful to read this post from five years ago.

Note that the number in the earlier post is $8 billion while the estimated effect of the Collins/Faso amendment is around $3.2 billion. The difference is that the larger number includes New York City, for which they would provide no relief.

More to come

{ Comments on this entry are closed }

I’ve been working on health coverage issues since the late 70’s. Here are some of the key things that WE have learned during the time since.

  1. Prior to Medicare, private insurers did not offer coverage to the elderly. Most elderly simply got dropped when they turned 65. That’s why we have Medicare.
  2. Prior to the Affordable Care Act (ACA), most people without coverage were in families with at least one working person. Many were in families with two working persons. But the employer/s didn’t offer coverage or the wages were too low to take advantage of what was offered. And, in many states, Medicaid did not offer coverage either.
  3. Those uncovered families included a lot of kids.
  4. There were lots of people without coverage and the numbers were growing, even after Medicare and Medicaid and until the ACA was implemented
  5. Coverage makes a difference in whether and when you get health care. And the more subtle (or insidious) the condition, the bigger the difference.
  6. If you seek care without coverage, it’s usually later in the disease process and thus, worse and more expensive to care for. It’s usually later because you don’t know you’re sick or the nature of your illness. Think late-stage cancer vs. cancer that’s detected early.
  7. Getting care makes a difference in whether and how well you live.
  8. If you get care without coverage, you’re more likely to go bankrupt.
  9. If you get care without coverage, people with coverage cover the cost through higher prices that eventually insurers and government pay.
  10. If you seek care without coverage, you usually go to the least efficient provider, namely a hospital emergency room.
  11. Not covering kids, especially with broad coverage that includes preventive care, is especially dumb.
  12. Higher numbers and percentages of uninsured patients undermine the financial health and viability of hospitals – including those that serve the insured. So coverage won’t protect you from indirect effects.
  13. People with infectious diseases, but without coverage, are still infectious and more likely to spread the infection. So coverage won’t protect you from indirect effects.You can’t hide from all those diseases in some gated community. 

Well, I hope I’ve learned other things too, but I think this list covers the key things. 

So repealing the Affordable Care Act without a credible replacement is downright foolish. 

And, no, I’m not a liberal and not even a fan of single-payer structures. But increasingly, I think I’d take single payer over a system collapsing of its own weight, both financial and political.

{ Comments on this entry are closed }

We’ve updated our visualized Medicaid enrollment data. These data run through May of 2016.

{ Comments on this entry are closed }

From The Secret War: Spies, Cyphers, and Guerrillas, 1939-1945, by Max Hastings:

 

“The first requirement for successful use of secret data is that commanders should be willing to analyze it honestly. Herbert Meyers, a veteran of Washington’s National Intelligence Council, defined his business as the presentation of ‘organized information’; He argued that ideally intelligence departments should provide a service for commanders resembling that of ship and aircraft navigation systems. Donald McLachlan, a British naval practitioner, observed: ‘Intelligence has much in common with scholarship, and the standards which one demanded in scholarship are those which should be applied to intelligence.’ “ 

Hastings goes on to reference German commanders after the war blaming their intelligence failures on Hitler’s “refusal to countenance objective assessment of evidence” – especially if the reports were unfavorable [to his views].’

 

Useful lessons here even outside the military and national security domains, but seems especially timely these days. Bad enough to try to fool others. Fooling yourself is especially dangerous.

{ Comments on this entry are closed }

Visualized Medicaid enrollment trends, by health plan (and fee-for-service) along with mix by demographic characteristics

{ Comments on this entry are closed }

This is as we – or at least most of us – would expect.

Compared to people in states that did not expand Medicaid under the ACA, newly available Medicaid in States that expanded the program enabled newly covered people to reduce their unpaid bills and debts. 

Luojia Hu, Robert Kaestner, Bhashkar Mazumder, Sarah Miller, and Ashley Wong wrote: 

The Effect of the Patient Protection and Affordable Care Act Medicaid Expansions on Financial Well-Being (PDF)

NBER Working Paper No. 22170, April 2016

Their abstract:

We examine the effect of the Medicaid expansions under the 2010 Patient Protection and Affordable Care Act (ACA) on financial outcomes using credit report data for a large sample of individuals. We employ the synthetic control method (Abadie et al., 2010) to compare individuals living in states that expanded Medicaid to those that did not. We find that the Medicaid expansions significantly reduced the number of unpaid bills and the amount of debt sent to third- party collection agencies among those residing in zip codes with the highest share of low income, uninsured individuals. Our estimates imply a reduction in collection balances of around $600 to $1,000 among those who gain Medicaid coverage due to the ACA. Our findings suggest that the ACA Medicaid expansions had important financial impacts beyond health care use 

Good to have evidence

{ Comments on this entry are closed }