What is wrong with spectrum prices?

Do you know that prices emanating out of the spectrum auctions in India do not correlate with the revenue potential of the respective markets? Isn't this strange? As normally one would expect the distribution of value across circles for bands must approximately align with its (circle's) revenue distribution. In other words, these percentages must be approximately same, even if the absolute values are different (linked to band's propagation characteristics, maturity etc). Then why are these numbers oscillating violently across bands and circles? This is important, as the operators have very little room for complacency to bid irresponsibly after having spent more than 50% of the total capex spend in buying spectrum and with growth in real terms being absent. See this - Has the Indian telecom industry stagnated? The current note is an attempt to investigate this problem and identify possible actions that could have been undertaken to prevent this situation.

The following chart provides a % distribution of the telecom industry's revenue across various circles in India.

Circlewise Revenue Distribution

As expected, the Metro and the Category "A" circles, like Delhi (DL), Mumbai (MU), AP, Karnataka (KA) etc contribute significantly more compared to Category "B" and "C" circles.

Now, to measure the true value of prices emanating out of these auctions, these have been superimposed on top of the revenue distribution across various markets. For this purpose, I have taken the latest auction price of that circle in consideration. The picture emerges as under. The "blue" bars indicate "% auction price" and the "% revenue distribution" is marked by the orange rectangle.

Metro & Catagory "A" Circles

In above figure, we can see that in Delhi (DL) and Mumbai (MU), except for the 900 MHz band, the prices for all bands are way above the reference levels (% revenue distribution). Similarly, in Gujarat (GU), all spectrum bands except 2500 MHz is above the reference level. In TN, some are below and some above the reference levels. In the other circles, the situation is no different.

Category "B" Circles

Here the prices of 900 MHz band in Punjab (PU), Rajasthan (RA), UPE & UPW are way above the reference level of % revenue distribution.

Category "C" Circles

In these circles, the auction prices are generally lower than the reference levels but differ widely across bands. Also, note that this situation is same in all the 22 circles. Why? Though the absolute prices are expected to be different (across circles and bands), but the ratios must be in a ballpark range.

The reasons for this could be many. Some are purely market-determined - operator's legacy situations at the time of bidding and the desperation to get spectrum in face of predatory bidding by competitors (at a much lower risk than the incumbents who want to keep their business alive). But, the main reasons is the lack of curation by the regulators to correct these anomalies while defining the reserve prices for the future auctions.

To explain this, let's take the example of 3G auctions of 2010, which resulted in a Pan-India price of Rs 16,750 Cr for a 5 MHz block. The figure below compares the % distribution of AP (Auction Price), RP (Reserve Price) and (Market Revenue) across circles.

You can see that auction price distribution is very skewed especially for Delhi and Mumbai (marked with "red" star) which contributes to 40% of the total value when the market revenue from these two circles adds up to only 14.6%. Wasn't it the responsibility of the regulators to curate this anomaly while setting the reserve price for future auctions? But, this anomaly was allowed to persist and is clearly reflected in the reserve prices that were set for future auctions. See figure below.

But, what could be the curation strategy? As some might argue that this will empower the regulator with too many discretionary powers and disturb the level playing conditions for those who have taken spectrum in earlier auctions. This is true only if the rules for curation are not declared transparently in advance. Also, these rules have to be simple so that it does not add too much complexity into the process of calculating reserve prices and also does not dilute the value of spectrum offered.

A simple way to curate the reserve price for the forthcoming auctions is to use factors which are static, precise and do not change dynamically. These are, a) % revenue distribution, b) propagation characteristics of the band in question. Let's say the regulator is entrusted with the task of extrapolating the reserve price (RP) of 1800 MHz band from the price discovered in the auctions held in 2010 for the 2100 MHz band. The first step is to curate the anomaly in the auction prices of the 2100 MHz band. This can be easily done by multiplying the % revenue distributions of individual circles with the Pan-India price of Rs 16,750 Cr to come out with the individual reserve price. The next step is to multiply these numbers by a factor derived by dividing the uplink frequencies of the two bands. As lower frequencies travel far and therefore should be valued more. The following figure compares the curated RP of the 1800 MHz band with the revenue distribution of the individual circles.

You can now see that reserve prices are now totally aligned with the % revenue distribution of the respective circles and removes the anomaly that emanated out of the earlier bidding process. This does not impact the total Pan-India value and keeps it at constant. This method is better than current method which links the RP to the auction price of individual circles amplifying the distortion in the future bidding process. The other advantage of this process is that it does not disturb the level playing conditions as the rules are clearly laid out in advance, and also promotes responsible bidding by the bidders, since the increase in price in one circle impacts them all in other circles in the future auctions.

For spectrum auctions to function properly the incentives need to be aligned so that operators bid responsibly. There needs to be a feedback mechanism for correcting anomalies in bid held in past so that it does not get carried forward in future auctions, thereby vitiating the process. Also, a balance needs to be maintained between supply and demand, and in the case of unsold spectrum, the RP of future rounds should be readjusted proportionately by the factor which remained unsold. Curating RP by rules explicitly stated in advance can improve the auction design without questioning the integrity of the regulator taking the decision - creating a win-win situation for all.

 (Views expressed are mine and do not reflect that of my employer)