Everything is logged, all the time

Assume everything you do is logged and can be accessed under the right conditions.

This is just how large computer systems work. You do something, the system takes an action, it logs the action it took. Those logs stay around for a while, to help with both diagnostics and repair. Without them, companies couldn't build or maintain large systems that span millions of devices. They couldn't detect and respond to security threats. The more reliable and high-performing the system, the more data it is likely to be logging, with more detail and more specificity.

The company may take privacy very seriously and protect those logs from improper access, but once the data is logged, the company has it, and as the article states, the government can compel companies to release it. And as a friend pointed out, that's only the company you're dealing with directly. Every web action involves a number of companies, all of which could be logging data about your actions.

This article claims that Apple should have disclosed the particular data they were logging. Maybe that would work for the article's author--a journalist who has reason to care deeply about such things--but we've all seen and ignored online Terms of Service. I think my last ToS and disclosure from Apple was 20 pages of small print, for some definition of print. Increasing that to 40 or 80 or 100 pages wouldn't help. Logging changes as engineers try to track down and fix problems, and the vast majority of users don't have the context to determine how the different pieces of logged data might be pieced together to create a larger picture.

So, just assume everything is logged. Remember that you're not operating a phone or computer by itself, but a piece of a large, highly connected system that spans continents and countries, one that records almost everything that happens, at least for a while, because those records are the nervous system and memory it needs to function.

And remember that system does not have the right to remain silent.

If your privacy needs are strict enough, learn how to protect your privacy yourself. Rather than hoping data won't be recorded, use less convenient tools that don't release the data in the first place. In the end, that will be far more effective than making sure you understand all the implications in a company's disclosures.

The "logic" of silicon valley

Robert Reich points to teachers moonlighting as Uber drivers in Silicon Valley as an example of the perverse logic of this area. There is a perverse logic here--companies like Facebook and Google serve vast number of customers with relatively few employees, making a few rich but leaving most out--but that's not what's happening with teachers. There is no logic to that. There is a combination of overwhelming economic forces and terrible laws.

The current round of high-tech companies has been so successful that they've outgrown both the local housing market and the infrastructure that supports it. Facebook, Google, Apple, Salesforce, et al continue to be successful, continue to grow, and continue to hire. Their employees continue to look for places to live. There's only so much room. There are only so many roads, schools, parks, and teachers.

Because of Prop 13 and its descendants, there's only so much money to pay for public services and employees. Property taxes are capped. Other local taxes are nearly impossible to raise. Neither the state nor the federal government are likely to come to the aid of the country's richest areas.

what brooks doesn't understand about mandatory minimums and mass incarceration

David Brooks claims that mandatory minimums have nothing to do with mass incarceration, but he misses the fact that mandatory minimums and prosecutor behavior go together. Over 95% of criminal cases are "resolved" without trial. Prosecutors use the threat of harsh mandatory minimums to coerce guilty pleas from defendants. Underfunded public defenders know that they won't have the resources to give their clients an adequate defense, so they encourage their clients to plead to lesser charges.

A couple weeks ago, a case exactly like this made national news. A teenager in North Carolina was charged as an adult felon for taking a nude selfie and sending it to his girlfriend. This caused a bit of outcry, which faded when he pled guilty to misdemeanor charges. What everyone seemed to miss is that the prosecutor never cared whether he could get a conviction on the felony charges, and probably didn't want to go to trial. The prosecutor wanted to raise the stakes for the defendant, so that the defendant would waive his fifth amendment rights and plead guilty.

It worked, and we were all happy with that.

Modern philosophy makes no sense

Kevin Drum talks about the failings of modern philosophy, including its fascination with the trolley problem. He describes a common philosophical undertaking, taking an apparently reasonable position then attempting to apply it universally and consistently, then being shocked at the results and either using that shock to discredit the initial position or claiming that the shocking final result is a valid result. The trolley problem is popular precisely because it is so easy to generate apparently reasonable initial positions, then extrapolate from them.

Peter Singer's The Life You Can Save, whether its book, web site, or organizational form is an extended exercise in this process. He takes scenarios based on a drowning child, and extrapolates to teach us how we should respond to global poverty. And it's a deeply dishonest effort.

The trolley problem (and Singer's drowning child variation) are exercises not in moral philosophy, but in moral psychology. They investigate the boundaries of our moral sense, not in the realm of logic but in the realm of emotion. And it turns out that our emotional moral senses are complicated, even chaotic. Like public opinion surveys, the results you get from exercises like this depend a great deal on precise wording, the exact types of actions (or inactions) taken (or not). What they show is that our moral sense is weird, strange, and inconsistent, and that we don't really understand it.

What Singer does with this stew of inconsistent results is to pick a particular result that supports his conclusion, then extrapolate from it, ignoring and sometimes dismissing contradictory scenarios, as well as common patterns from the same sets of morality exercises.

For example, Singer tells the story of someone driving a luxury car that represents not just a current source of great personal joy, but his retirement savings, then asks whether the owner should direct a train away from a child to save the child's life, even at the expense of his own joy and future well-being. Unsurprisingly, most people think he should, and from that Singer concludes that we should all forgo all but the bare essentials to prevent avoidable deaths from poverty.

Yet if you change the story a bit, and ask questions like "Should someone mortgage their house to pay for life-saving treatment of a child who lives down the street?" you would get different answers.

In Singer's book, all his scenarios support his conclusions. Moreover, after noting that trolley problem thought experiments suggest that people view action and inaction differently (perhaps recognizing that in a finite life, we cannot do all things), he dismisses that general observation, consistent across a wide range of moral psychology scenarios, as irrelevant.

Unsurprisingly, while noting that poverty has decreased dramatically over the past five decades, both in real and relative terms, the book spends almost no time investigating that phenomenon, insisting instead that large increases in aid are the best answer going forward.

The biggest disappointment for me is that I agree with Prof. Singer that we should and could be doing far more, and yet I found the arguments in his book depressingly weak.

Federal fund rate, median income, and core inflation.

Here, I've added a graph of core inflation to the previous graph. While the correlation between median income growth and the federal funds rates is not perfect, it's far stronger than the correlation between the federal funds rate and inflation. In 1994-95, for example, the FFR doubled, even though inflation (already at a near 10-year low), dropped slightly.

Federal Funds Rate vs Median Household Income

Sustained drops in median income seem to be the rule rather than the exception. Once median income starts to fall, the drop seems to gather momentum and be hard to arrest, at least with monetary policy. What starts the drops? Is it policy or a natural cycle?


Here we have a graph of the federal funds rate against median income. While the graphs are not completely synchronized, their shapes are largely similar. While median income rises, the Federal Funds Rate increases. When median income falls, the Federal Funds rate also falls. The exception to this pattern is 1995-1998, when the Fed allowed median income to rise without responded with interest rate increases. Perhaps not coincidentally, that was the longest period of sustained income growth covered by this data series.

inequality increases, recession vs non-recession

Following up on that last post, depending on how you calculate it, the convergence between median family income and per capita GDP since 1984 occurs either completely or predominantly during the median income drops that surround recognized recessions.

Arithmetically, the difference between median family and per capita income dropped by $14,500 between 1984 and 2012. $14,500 of that change was associated with recessions.

Geometrically, per capita GDP increased 61% from '84 to '12. Median income increased by 8.1%. Between median income recessions, per capita income outpaced median income 43% to 30%, significant but not startling. During median income recessions, per capita income increased 12%. Median income dropped 17%.

By the way, I realize that per capita GDP and median household income measure slightly different things. Average household size, for example, decreased by about 7% over this time period, which would explain some of the difference in non-recession growth. There are, no doubt, other corrections to the baseline trends that explain some differences in growth rates without corresponding increases in inequality.

It seems unlikely, however, that those trends are related to economic cycles. It's hard to believe that average family sizes dropped 8% from 2007 to 2012, or that benefits increased about 15% from 1999-2004 to compensate for the drop in median income. And to the extent that those trends explain the baseline, non-recession convergence of the curves, it only further emphasizes the role of recessions in increasing inequality.

It hasn't always been this way. Prior to 1980, per capita GDP and median household income tracked one another very closely.

more fun with graphs

Median incomes decline before a recession, and continue to decline after the recession is over. What about per capita incomes?



Average incomes drop less steeply, for a shorter time.

It appears, in fact, that the median income drops surrounding recessions account for much of the rise in income inequality over the past thirty years. During full recovery periods (where median income is rising), the curves in the graph track fairly well. During recessions (broadly defined), the curves diverge.