Many critics rightly say the Federal Communications Commission’s methods for determining broadband coverage by terrestrial networks (25 Mbps minimum downstream) are wrong, and overstate the amount of access.
Virtually everyone might agree with the general statement. What has never been clear is the magnitude of the collection error. Is broadband availability wildly overstated, or just a little overstated?
George Ford, chief economist at the Phoenix Center for Advanced Legal and Economic Policy Studies, has an answer: 3.3 percent overstatement. Put another way, some four million U.S. homes are assumed to have broadband that do not. Ford uses a 126 million figure for U.S. housing units.
Using a national housing units base of 138.5 million, that is an overcount of less than three percent. Estimates of housing units are nuanced, as some units are not occupied, occupied for only parts of a year, and include rooms in other structures, boats and other unusual situations.
One might argue that the overstatement is relatively minor. At a larger level, one might also argue that getting the last few percentage points of “solution” for any “problem” tends to be wildly more expensive or difficult than producing solutions for the middle 80 percent to 90 percent of cases.
That might be as true for weight loss as it is for building or enumerating communications infrastructure.
One might argue that most of the value of a data set related to individual behavior is captured in the first 50 percent of acquired data (much of the latter behavior is repetition of past behavior). The incremental value of the last few percentage points of data is low, relative to the earlier data.
The Gaussian distribution (Bell curve or normal distribution) probably illustrates the concept as well. Most of the results of any construction project (homes connected, miles built) are likely to occur in the middle 90 percent of activities. That has the same general implications as the law of diminishing returns might suggest.
Slow progress at first results in a rapid rate of change in the middle, until incremental value slows, stops or becomes negative.
That is true for fixed communication network costs as well. Connecting the last few percent of locations, typically the most isolated and rural places, is the most expensive.
The concern about rural broadband coverage is well placed. Progress also tends to be slow, in part because we keep moving the goalposts (faster speeds over time mean the work never ends), and partly because supplying the last couple of percent of locations with fixed network access is wildly expensive to prohibitive, even with subsidies.
Rural networks might cost three times more than urban networks. Rural networks might cost four to five times more than suburban networks.
The general point is that there is a law of diminishing returns at work, as elsewhere in life.
Better data collection is better than worse data collection. But the FCC’s data seems to have captured most of the value, if inaccurate at the margins. Improvement is possible, but not as much as many seem to believe.
No comments:
Post a Comment