|
Article Excerpt
Local telecommunications competition was an important goal of the 1996 Telecommunications Act. We evaluate the consumer welfare effects of entry into residential local telephone service in New York State using household-level data from September 1999 to March 2003. We address the prevalence of nonlinear tariffs by developing a discrete/continuous demand model that allows for service bundling and unobservable provider quality. We find that the average subscriber to the entrants' services gains a monthly equivalent of $2.33, or 6.2% of her bill, in welfare from competition. These gains accrue primarily from firm differentiation and new plan introductions rather than from price effects.
1. Introduction
* Many countries, including Australia, Canada, the European Union's (EU) member countries, Japan, South Korea, and the United States, require incumbent telephone companies to allow entrants to lease individual elements of their network infrastructure at regulated rates, a practice commonly referred to as "unbundling." Regulators have imposed such requirements as a means to create competitive telecommunications markets by facilitating entry and thereby benefit consumers through adoption of new services. In this article, we study the consumer welfare implications of unbundling policies in the United States, using a household-level data set from New York State. Possible welfare effects, over the length of the time period of our study, derive from price reductions, development of new service offerings, and quality differences between the incumbent's and entrants' services.
In the United States, the markets for local residential and business telephone service were opened to competition following the 1996 Telecommunications Act (the 1996 Act). The 1996 Act allows three forms of entry: replication of the incumbent's infrastructure, lease of the incumbent's existing infrastructure (unbundling), or resale of the incumbent's services. By the late 1990s, competitive local exchange carriers had begun service in most U.S. states, primarily through leasing unbundled infrastructure or reselling service. As of December 31, 2003, entrants leased 60.6% of their residential and business lines and another 16.0% involved the resale of an incumbent's services (Federal Communications Commission, 2004).
We evaluate the consumer welfare effects of entry in New York State by quantifying the effects based on actual consumer choices. Although there are a few other estimates of the expected savings by U.S. consumers because of entry into local phone service (see Telecommunications Research & Action Center, 2000, 2001), these rely on hypothetical consumer migrations and ignore geographic differences in providers' service offerings. Instead, the household-level data that we utilize allow us to model geographic variations in service offerings as well as individual choice behavior using detailed demographics, thereby permitting a distinction between welfare effects generated by lower prices, increased product variety, and firm-level differentiation.
Providers sell residential local phone service under a menu of two-or three-part tariffs with monthly fees and per-unit prices, allowing consumers to choose their plan and their usage. Such mixed discrete/continuous choices also occur in other markets such as energy, information, and Internet access. Hanemann (1984) provides a methodology for estimating econometric models that link the discrete and continuous choices in the same utility maximization problem. We expand his framework along several dimensions. Our utility specification allows for flat-rate plans at zero marginal prices, bundling of multiple services under one plan, unobserved vertical quality differences across firms, and horizontal firm differentiation. We employ a maximum likelihood estimation technique that combines the information from the discrete and continuous choices by consumers. Our analysis focuses only on the demand effects of entry because regulation of incumbent prices for both the shared infrastructure and the final services effectively limits the role of supply-side responses.
We estimate our model using data from a random sample of New York households over a three and a half year period beginning in late 1999. We focus on New York because it was one of the first states to experience entry under the 1996 Act. Significant entry occurs by the end of our sample period, with entrants serving over 20% of the access lines. The largest entrants are AT&T and MCI, which comprise approximately 85% of the residential lines served by entrants. They compete with the incumbent, Verizon, whose customers are also included in the analysis.
Applying our econometric model to the household-level data, we are able to quantify the full distribution of welfare effects of entry. We find that these effects differ dramatically across households, underscoring the need for a household-level analysis. The addition of the entrants' plans to the choice set leads, on average, to household gains of $0.83 per month, or 2.5% of households' average monthly bill. The standard deviation of welfare gains across households is $3.42, with average gains of $2.77 and $1.13 accruing to AT&T and MCI subscribers, respectively.
Exploiting the detailed household-level data, we decompose the different sources of consumer benefits from entry and determine who gains the most from unbundling. The entrants introduce several new plans not offered by the incumbent. We find that considerable gains to consumers result from these new plan varieties. The introduction by AT&T and MCI of a fiat-rate plan previously unavailable in New York City entails a welfare increase of $0.68 per month on average, whereas AT&T's introduction of a combined local and regional calling plan results in a welfare increase of $0.21 per month on average across all households.
Although the entrants employ the same technological infrastructure as the incumbent, their services are vertically and horizontally differentiated due to nontechnological factors, such as customer support and marketing. Observed and unobserved firm effects play an important role in households' choices. Prior experience with an entrant's long-distance services is particularly important, suggesting that households consider a prior relationship in forming their assessment of the entrant's local service quality. We find that AT&T households enjoy monthly benefits of $1.84 per month from such differentiation, whereas MCI households receive monthly benefits of $0.64. This compares to savings of $1.15 and $0.40 over what the average AT&T and MCI household, respectively, would have paid on Verizon based on pure price effects alone. Thus, a focus only on prices, disregarding firm differentiation, significantly understates household welfare gains.
Relating household-level welfare benefits to household characteristics, we find that larger households in New York City benefit more from entry, presumably due to the entrants' introductions of new flat-rate plans there. The fact that high-usage households, which on average are lower income, black, and larger, gain more than other households do suggests that competition intensifies primarily for high-valuation customers. Customers of the entrants' long-distance services also gain more, consistent with the importance of horizontal firm differentiation.
Our results are particularly interesting because, according to the provisions of the 1996 Act, incumbents and entrants were to compete starting with identical inputs and on an equal footing. Our finding that entrants offer differentiated services and introduce meaningful new plan types is thus important for quantifying the benefits of regulatory measures to increase competition when entrants and incumbents share infrastructure. We find that these effects are more important than the pure price effects. This is consistent with results in Greenstein and Mazzeo (2006), who find that entrants' initial market selection depends on differentiation strategies, in addition to factors highlighted by Crandall and Sidak (2002) and Zolnierek, Eisner, and Burton (2001), such as demand and cost differences across markets, economies of geographic scope, and regulatory stringency.
In March 2004, the DC Court of Appeals reversed the Federal Communications Commission's (FCC) implementation of unbundling policies, finding that the absence of nationwide access to the local telecommunications network at "cost plus reasonable profit" did not "impair" entrants, as the 1996 Act required. Thereafter, incumbents were not required to follow cost-based pricing in leasing the local network. As a result, the largest entrants stopped accepting new customers (see Economides, 2005 for more detail). Despite the short-lived effects of unbundling in the United States, the experience through 2004 is instructive for several reasons. First, it offers a unique opportunity to inform decisions of whether to initiate or continue unbundling policies in local telecommunications markets in other countries or settings. These include unbundling cable modem Internet access, presently awaiting implementation in Canada, cable television service, under consideration in The Netherlands, and DSL Internet access, mandated by the EU's 2000 Unbundling Regulation. Second, the detailed U.S. data available allow an assessment of the economic implications of unbundling policies, such as the sources and magnitudes of any welfare gains and the types of households that are likely to benefit. Although the specifics of settings differ, our results suggest that initiatives to promote unbundled access in other countries should consider broader implications beyond price effects.
2. New York's local telephone markets
* New York has two sizeable incumbents, whose geographic territories are displayed in Figure 1. Frontier Communications of New York controls 8% of incumbent business and residential access lines as of December 31, 2001, while Verizon controls 89%. The 1996 Act introduces local telecommunications competition in Verizon's region, which had not previously opened its markets to competition.
The New York Public Service Commission (NYPSC) serves two roles related to our study in regulating the local phone market. First, it sets infrastructure lease rates that entrants pay to the incumbent. These rates remain stable during our sample period with the exception of a decrease of about 30% in July 2002. The lease rates vary by geographic zone within the state, averaging approximately $20 per line served as of early 2001 and failing to approximately $15 by the middle of 2002. (1) Second, the NYPSC regulates the retail rates that incumbents charge for local phone services. We describe the retail prices faced by customers in our sample in Section 3.
[FIGURE 1 OMITTED]
The entrants' market share of residential lines increases steadily from 6% in 1999 to 22% in 2002. The main entrants in New York during our sample period are AT&T and MCI, both of which expand into local service from the long-distance market. As of December 31,2001, AT&T and MCI hold shares of 58% and 27%, respectively, of residential lines served by entrants in New York. AT&T's initial entry into New York occurs in late 1999 and focuses on the state's metropolitan areas, as displayed in Figure 2. In 2001, AT&T enters into the remainder of Verizon's territory. MCI enters Verizon's entire territory in 1999.
By 2001, each of the three carriers offers three types of phone service: local, regional, and long-distance. Local service applies to phone calls in the household's local calling area. The carriers define local calling areas identically. These differ significantly in the population served, ranging from 1,873 people in Indian Lake to 7.98 million people in New York City, with an average of 304,597 people. Local calling areas are grouped into seven regional calling areas. The local service provider generally handles calls within these regions and outside of the household's local calling area, but imposes higher prices. All remaining calls qualify as long-distance calls. Although Verizon also began offering long-distance service in 2001, it and the entrants provided long-distance service primarily under separate calling plans from local and regional service during our sample period. As a result, we concentrate on estimating the benefits from competition in local and regional service.
The carriers provide local and regional phone services through monthly calling plans, of which three types are available to New York customers. Metered plans include a monthly fee to obtain service and a per-call fee for usage. Flat-rate plans specify a monthly fee for unlimited calling. Hybrid plans are three-part tariffs where a household pays a monthly fee to obtain a certain number of calls at zero marginal price. For calls above the predefined allowance, the household pays a positive marginal price. The FCC requires local carriers to offer additional calling plans at reduced rates to qualifying low-income households.
[FIGURE 2 OMITTED]
Local phone service providers also offer add-on features, such as call waiting, call forwarding, three-way calling, and speed dialing, to their customers. During the time period of our study, the carriers do not bundle these features with local or regional service. Instead, they offer them as optional features that households purchase along with basic service.
3. Data
We analyze plan choice and usage behavior using data on a sample of New York households collected by TNS Telecoms (TNS), complemented with data on the geography and structure of local telephone markets from various sources. TNS's Bill Harvesting data contain survey data from residential customers. TNS gathers demographic information on the households and asks them to submit their actual phone bill. Because willingness to respond varies by household characteristics, TNS employs oversampling to obtain a random sample. Most households participate in the survey only once. Our sample runs from September 1999 through March 2003 and covers a total of 7222 cross-sectional household observations.
For each household, the data set contains information on its local carrier choice and any detailed line items recorded on its local telephone bill, including the services the household purchased, total amount paid, and a breakdown of the bill into services, fees, and regulatory charges. For households that subscribe to metered service or exceed their call allowance on a hybrid plan, we observe the number of calls made during the billing period. Local usage is generally not available for the remaining households, notably those on flat-rate plans.
The TNS data do not directly identify the calling plan chosen by the household. Instead, we use descriptions of purchased services and associated expenditures to identify the household's chosen calling plan based on publicly available information. The NYPSC requires all carriers to publicly file retail price information whenever prices change, allowing us to construct the universe of calling plans and their prices at all times during our sample period. Because providers adjust consumers' bills to reflect price changes as they go into effect, we identify each household's plan among contemporaneous plans. Across providers, we are able to match 97.2% of households to a calling plan.
Because AT&T and MCI did not enter into Frontier Communications's territory, we focus on households that reside in Verizon's territory. We use the household's location, available at the zip-code level, together with detailed data on the availability of AT&T's and MCI's local service, to construct the household's choice set of local service carriers and plans at the time of its bill. We employ the Center for Communications Management Information's QTEL Local Calling Area Database, which maps the first six digits of a telephone number to the set of telephone numbers that can be called locally, to identify each household's local calling area based on its telephone number. To measure the size of the potential calling area for local and regional service, the "coverage area," we use mapping software to measure the calling area's population based on the 2000 Census.
Of the 7222 households in the data, 696 subscribe to AT&T, 931 to Frontier Communications, 362 to MCI, and 5233 to Verizon. We exclude the Frontier customers and the remaining 547 subscribers to other, small carriers from the sample. The entrants' representation in the resulting sample is similar to aggregate market shares for New York. In 2001, for example, the NYPSC reports market shares based on number of lines of 11.8% and 5.6% for AT&T and MCI, respectively, while in the TNS sample for the same year, 12.2% of households are AT&T customers and 5.4% are MCI customers. Due to missing information, reporting errors by TNS, or other data issues, we are not able to use all available observations in our estimation. The sample observations with complete demographic and usage information total 592 (85%) for AT&T, 218 (62%) for MCI, and 3981 (76%) for Verizon. (2)
The TNS data contain information on the household's basic demographic profile as well as use of other telecommunications services and technology products. We summarize the available variables in Table 1. Household size and income are likely the strongest proxies for demand for usage. TNS reports the household's income as a categorical variable. To transform it into a single continuous variable, we assign to each household an income equal to the predicted average income level in its category, as in Miravete (2002b). Compared to the state's aggregate distribution, the income distribution is slightly skewed toward lower-income households. Other demand shifters include whether the household moved in the past year and had to choose a new service provider, whether at least one member of the household subscribes to cellular service, and whether...
|
