Cooking the Books While Startups Languish

Since leaving the FCC, former chairman Tom Wheeler and Special Advisor Gigi Sohn have relentlessly attacked current chairman Ajit Pai. This is an unusual breach of decorum because former FCC chairs, like former presidents, generally leave the politicking against their successors to others.

Wheeler’s and Sohn’s criticisms also tend to be factually deficient. The most recent example is a blog post on the Brookings site criticizing America’s broadband speeds. The post, Defining Digital Down, asserts that the US has fallen behind Kenya in Internet connectivity speeds.

This is presumably shameful for reasons that Wheeler doesn’t feel a need to spell out, but this fact – if it is a fact – is more an indictment of Wheeler’s leadership than any actions the current FCC may take with the current method of assessing broadband progress, the general subject of the post.

Measuring Broadband Speed

Wheeler’s post displays confusion about how we measure broadband speeds. He supports the claim with a link to a Kenyan blog that cites data from Akamai about mobile network speeds in the first quarter of 2017. The blog cites the mobile data taken by measuring TCP performance, which is not the same as broadband speed.

According to the Akamai data, Kenya’s average mobile TCP download speed was 13.7 Mbps while the corresponding figure for the US was 10.7 Mbps. This is mobile application speed rather than a network speed, but let’s go with it.

The Kenyan figure represents a 274% increase from the oldest data Akamai has for that country, Q4 2015, to the most recent data, Q1 2017. During the same period, the US showed a 223% increase. In absolute terms, Akamai’s dataset shows Kenya improving from 5.0 to 13.7 Mbps while the US improved from 4.8 to 10.7.

Mobile data is always somewhat dubious, but if we take it at face value we see that Kenya’s mobile applications outperformed those in the US while Mr. Wheeler led the FCC. This can’t very well be laid at the feet of Chairman Pai because he was only on the job for one month of the measurement period.

Pity Me, I’m an Orphan

So this is a very curious argument. As FCC chairman, Wheeler enacted regulations that caused America’s networks to advance less rapidly than those of Kenya, and now he complains about the fruits of his labor. Wheeler is like the child who murders his parents and then begs the court for mercy because he’s an orphan.

But I said mobile is hard to measure, so let’s see what the other datasets show. Speedtest Global Index only shows the most recent year, and over that period Kenya has outperformed the US 161% to 115%.

But this dataset has the US running at a mobile download speed of 23 Mbps vs. 15 for Kenya. Wheeler was careful to avoid it, perhaps because it doesn’t really work for his denunciation of Pai. But he may not have been aware of it either; data has never really been his thing.

What is Wheeler Trying to Say?

Apart from the clumsy data, the theme of the post seems to be that the US needs to have high aspirations for broadband speeds if we’re ever to become a world leader in Internet applications. He doesn’t state this clearly, but we presumably want fast networks for some reason beyond bragging rights over Kenya.

And indeed, Kenya is powerhouse nation in mobile payments. Its leading payment processor, M-PESA, was launched in 2007 and now serves 96% of Kenyan households outside of Nairobi. A recent study shows that access to mobile money is making an impact on poverty:

The study found that increased access to mobile money has reduced poverty in Kenya, particularly among female-headed households. Rapid expansion of mobile money has lifted an estimated 2 percent of Kenyan households (some 194,000) out of extreme poverty. It has also enabled 185,000 women to move out of subsistence farming and into business or sales occupations.

So there can be a connection between app development, network quality, and poverty. It’s just not as simplistic as a “virtuous cycle” where investment in one part of the network economy makes everything else better.

Magic Yardsticks

It’s also not the case that playing with definitions has a meaningful impact on network quality. Network startups pay less attention to government statistics than to unmet needs such as those of the Kenyans who didn’t have bank accounts before M-PESA.

The US is falling behind in startup formation. In 1980, one in eight Americans was employed by a firm less than one year old, but the current figure is one in 12. Eight of the top ten mobile apps are owned by Google and Facebook, which makes things tough on startups.

There are a number of reasons for the concentration of market power in the hands of these dominant firms, such as network effects. But this is an area of the economy that’s received scant attention during the net neutrality decade. Pursuing the strategy of limiting the power of broadband service providers has taken regulators’ eyes off the market concentration ball.

Perhaps it’s better to pay more attention to economic analysis of Internet markets than to pseudo-technical performance goals and dubious service definitions. If we were to do that, we might be able to find a way to do something about the stagnation in startup formation.

Addendum: I’d be remiss if I didn’t tout my conference paper on broadband and web measurement, You Get What You Measure: Internet Performance Measurement as a Policy Tool. It covers both wireless and wired, and broadband as well as web speeds.