Correlates of Quality Improvement in U.S. Broadband Service

Newly introduced and improved official government price indexes show both residential and business wired internet access prices essentially flat since 2007. In stark contrast, prices for wireless telecommunications services have been falling, at a consistent rate of about 2 to 4 percent per year over this period, and mobile broadband data prices have been falling at rates a full order of magnitude greater. Several studies which attempt to correct official price indexes for unmeasured quality improvement using data on actual or advertised download speed,  conclude that, overall, quality‐adjusted wired broadband service prices did fall, but at disappointingly slow, single‐digit rates.

This paper is concerned with understanding whether or not broadband quality improvement is systematically related to competitive conditions in regional broadband markets, making use of rich new data sources that allow us to directly measure the evolution of many dimensions of broadband service quality (though not price) over the thirty month period from March of 2011 through September of 2013. Using these data, I construct household‐specific weekly average download speed indexes for a sample of over 9,000 individual households. Household‐specific hourly variation in local network broadband download speed is controlled for using hour‐pair dummy variables in a first‐stage fixed effects regression model of changes in weekly download speed. The estimated first‐stage weekly dummy variables can then be interpreted as weekly measures of household download speed adjusted for time and day of week effects.

The first stage estimates of hourly effects provide compelling evidence that variation in download speed within a speed tier primarily reflects variation in local network traffic. The pattern of these temporal effects strongly suggests that it is traffic (and congestion) on local network infrastructure, rather than global (or national) backbone infrastructure, that is primarily responsible for reductions in speed and service quality.

A second stage fixed effects model is then applied to these estimates of weekly household download speed to estimate the impact of changes in time‐varying shifters of broadband demand and supply on weekly download speed. Preliminary results show that increased numbers of fixed wireline broadband providers generally have no statistically or economically significant impact on download speeds. Quite interestingly and perhaps provocatively, the same model shows statistically significant impacts on fixed connection broadband service quality associated with the presence of larger numbers of wireless mobile internet providers. We are left with a provocative and interesting empirical relationship that is consistent with the view that mobile is where the quality and price competition is most likely going to be coming from in the future, absent any major structural change in the current U.S. broadband market.