Tremors in Bank Land


Jason Baker | Flickr

Markets are putting Deutsche Bank through a safety drill. If we’re lucky, the rustle of investors fleeing Deutsche Bank late this week will prove to be a minor hiccup in the world of capital. If we’re unlucky, it will be the first domino to fall in another wave of financial disaster like the one we experienced in 2008. Deutsche Bank’s recent round of woes started with a $14 billion fine by the U.S. government for the bank’s sale of mortgage-backed securities during that last bank crisis, which sent investors for the doors (Fortune). Along with the recent scandal at Wells Fargo, which revealed that thousands of its employees had for years been opening fake accounts to meet sales incentives, Deutsche Bank’s problems remind us that, as a result of hard lessons learned a decade ago, our banking system’s safety protections got some upgrades, but they haven’t actually been tested yet. And if they’re not up to the next crisis, then we’re all going to be stuck paying the price — again.

When is a poll not a poll? When it’s an online survey in which anyone can vote and vote again. One of the foundations of data science took a beating in the political arena this week, as Donald Trump crowed over his wins in snap post-debate online “polls” even as the real polls started tilting in his opponent’s favor (CNN). Real polls, scientific ones, try to map the reality of public opinion through a process of weighted sampling. You might only ask 1000 people what they think; but if those thousand respondents match the national population in race and religion, party affiliation, income, and so on — or, if they don’t, you weight them a bit so they do — you’ve got a snapshot of something real. But if you just open the tap on the internet and ask anyone to vote, your results are nearly worthless. Trump’s wins are like Hank the Angry Drunken Dwarf’s victory in People’s “Most Beautiful People” online poll; a self-selected sample is just a gauge of partisan excitement level. In a larger sense, this overvaluing of online polls reflects our deeper confusion over how to use all kinds of metrics. By turning yardsticks into targets, we break them (Quartz). Your poll becomes someone else’s campaign; your Key Performance Indicator stops telling you anything useful because everyone in your organization is trying to game it. Without context and understanding, raw data is worse than useless — it can be toxic.

Read More