Research

My research work focuses on how people use technology to consume experience goods and influence others to do so. These are inextricably linked to how firms behave and how public policies affect market structures. My work focuses on the application of robust empirical identification methods to analyze large datasets obtained from organic in-vivo large-scale network-centric randomized experiments. Most of my research spans two interrelated applied areas. I am interested in the Impact of Information and Communication Technologies on Education (IICTE) and in Peer-Influence and Consumption in the Media Industry (PICMI). The bulk of my work is in medialytics — using big data to understand the future of the media industry. In addition, I have also been studying Competition, Consumer Churn and Switching Costs (CCCSC) in telecommunications.

Publications

Selected Publications

Impact of Information and Communication Technologies on Education (IICTE)

Spillovers Effects of Wiring Schools with Broadband: The Critical Role of Children.
Belo, R; Ferreira, P; Telang, R.  Management Science.  Vol 62, N. 12, pp. 3450-3471,  2016.

Abstract

Providing broadband to schools can be an effective way to foster household Internet adoption in neighboring areas. On the one hand, the infrastructure put into place to meet schools’ needs can also serve households. On the other hand, students get acquainted with Internet at school and signal its usefulness to adults at home who, consequently, can be more likely to adopt it. In this paper we model the roles that broadband use at school and Internet adoption in neighboring households play in the decision to adopt Internet at home and measure their effects empirically. We use data from Portugal between 2006 and 2009 on household Internet penetration and on how much schools use broadband. We use two different sets of instruments for the schools’ broadband use to alleviate endogeneity concerns. Both approaches yield similar results. We find that broadband use at school leads to higher levels of Internet penetration in neighboring households. Broadband use in schools was responsible for a year-over-year increase of 3.5 percentage points on Internet penetration in households with children. Across our dataset this effect accounts for about 17% of the increase in home Internet adoption. We also find evidence of regional spillovers in Internet adoption across households. These were roughly responsible for an increase of 2.1 percentage points in Internet penetration or 38% of the total increase in household Internet penetration between 2006 and 2009. These results show that wiring schools with broadband is an effective policy to lower the barriers for Internet adoption at home and as such contributes to accelerate the pace of broadband diffusion.

Broadband in School: Impact on Student Performance.
Belo, R; Ferreira, P; Telang, R.  Management Science, Lead Article.  Vol. 60, N. 2, pp. 265-282,  2014.

Abstract

This paper examines the effects of providing broadband to schools on students’ performance. We use a rich panel of data on broadband use and students’ grades from all middle schools in Portugal. Employing a first-differences specification to control for school-specific unobserved effects and instrumenting the quality of broadband to account for unobserved time-varying effects, we show that high levels of broadband use in schools were detrimental for grades on the ninth-grade national exams in Portugal. For the average broadband use in schools, grades reduced 0.78 of a standard deviation from 2005 to 2009. We also show that broadband has a negative impact on exam scores regardless of gender, subject, or school quality and that the way schools allow students to use the Internet affects their performance. In particular, students in schools that block access to websites such as YouTube perform relatively better.

Peer-Influence and Consumption in the Media Industry (PICMI)

Culling the Herd: Using Real World Randomized Experiments to Measure Social Bias with Known Costly Goods.
Matos, M; Ferreira, P; Smith, M; Telang, R.  Management Science.  Vol 62, N. 9, pp. 2563-2580,  2016.

Abstract

Peer-ratings have become increasingly important sources of product information, particularly in markets for “information goods.” However, in spite of the increasing prevalence of this information, there are relatively few academic studies that analyze the impact of peer-ratings on consumers transacting in “real world” marketplaces. In this paper, we partner with a major cable company to analyze the impact of peer-ratings in a real-world Video-on-Demand market where consumer participation is organic and where movies are costly and well-known to consumers. After experimentally manipulating the initial conditions of product information displayed to consumers, we find that, consistent with the prior literature, peer-ratings influence consumer behavior independently from underlying product quality. However, we also find that, in contrast to the prior literature, at least in our setting there is little evidence of long-term bias due to herding effects. Specifically, when movies are artificially promoted or demoted in peer-rating lists, subsequent reviews cause them to return to their true quality position relatively quickly. One explanation for this difference is that consumers in our empirical setting likely had more outside information about the true quality of the products they were evaluating than did consumers in the studies reported in prior literature. While tentative, this explanation suggests that in real-world marketplaces where consumers have sufficient access to outside information about true product quality, peer-ratings may be more robust to herding effects and thus provide more reliable signals of true product quality, than previously thought.

Peer Influence in the Diffusion of iPhone 3G over a Large Social Network.
Matos, M; Ferreira, P; Krackhardt, D.  Management of Information Systems Quarterly.  Vol. 38, I. 4, pp. 1103-1133,  2014.

Abstract

In this paper, we study the effect of peer influence in the diffusion of the iPhone 3G across a number of communities
sampled from a large dataset provided by a major European Mobile carrier in one country. We identify
tight communities of users in which peer influence may play a role and use instrumental variables to control
for potential correlation between unobserved subscriber heterogeneity and friends’ adoption. We provide
evidence that the propensity of a subscriber to adopt increases with the percentage of friends who have already
adopted. During a period of 11 months, we estimate that 14 percent of iPhone 3Gs sold by this carrier were
due to peer influence. This result is obtained after controlling for social clustering, gender, previous adoption
of mobile Internet data plans, ownership of technologically advanced handsets, and heterogeneity in the
regions where subscribers move during the day and spend most of their evenings. This result remains qualitatively
unchanged when we control for changes over time in the structure of the social network. We provide
results from several policy experiments showing that, with this level of effect of peer influence, the carrier
would have hardly benefitted from using traditional marketing strategies to seed the iPhone 3G to benefit from
viral marketing.

The Effect of Video-on-Demand on Piracy: Evidence From a Household Level Randomized Experiment.
Matos, M; Ferreira, P; Smith, M.  Management Science.  forthcoming.

Abstract

We partner with a major multinational telecommunications provider to analyze the effect of Subscription Video-on-Demand (SVoD) services on digital piracy. For a period of 45 consecutive days, a group of randomly selected households who used BitTorrent in the past were gifted with a bundle of TV channels with movies and TV shows that could be streamed as in SVoD. We find that, on average, households that received the gift increased overall TV consumption by 4.6% and reduced Internet downloads and uploads by 4.2% and 4.5%, respectively. However, and also on average, treated households did not change their likelihood of using BitTorrent during the experiment. Our findings are heterogeneous across households and are mediated by the fit between the preferences of households in our sample for movies and the content available as part of the gifted channels. Households with preferences aligned with the gifted content reduced their probability of using BitTorrent during the experiment by 18% and decreased their amount of upload traffic by 45%. We also show using simulation that the size of the SVoD catalog and licensing window restrictions limit significantly the ability of content providers to match SVoD offerings to the preferences of BitTorrent users. Finally, we estimate that households in our sample are willing to pay at most $3.25 USD per month to access a SVoD catalog as large as Netflix?s in the US. Together, our results show that, as a stand-alone strategy, using legal SVoD to curtail piracy will require, at the minimum, offering content much earlier, and at much lower prices than those currently offered in the marketplace – changes that are likely to reduce industry revenue and may damage overall incentives to produce new content while, at the same time, might curb only a small share of piracy.

Competition, Consumer Churn and Switching Costs (CCCSC)

Welfare Changes in the Cournot Setting: An Applications to the Telecommunications Industry.
Ferreira, P.  Journal of Industrial Economics.  Forthcoming,

Abstract

This paper characterizes the welfare efficiency of the Cournot equi- librium and provides bounds for the loss in consumer surplus, producer surplus and welfare when the number of firms in the market changes. I only assume that demand is decreasing in price and costs increasing in the quantity produced as long as Cournot equilibrium exists. I show how price, demand and average cost, before and after the number of firms in the market changes, can be used to compute these bounds. I apply these bounds to the Portuguese wireline market and conclude that the welfare loss carried by Portugal Telecom’s monopoly in 2005 reduced significantly when the company was split in 2007.


Papers Under Review

Impact of Information and Communication Technologies on Education (IICTE)

Peer-Influence and Consumption in the Media Industry (PICMI)

The Effect of Time Shift TV on TV Consumption and Ad-Avoidance.
Reis, F; Belo, R; Ferreira, P; Matos, M.

Abstract

We partner with a major telecommunications provider to study the effect of Time-Shift TV (TSTV) on TV consumption. TSTV automatically records in the cloud pro- grams that were broadcasted live for the past few days and thus allows users to watch the programs they want when they want. In 2012, our industrial partner deployed TSTV to half the TV channels it offered to premium consumers. Using difference-in-differences with Inverse Probability of Treatment Weighting we find that, on average, the introduction of TSTV increased the consumption of daily TV by 11 minutes (p < 0.01), from a baseline of 3.4 hours, and did not change the consumption of live TV. Furthermore, we find that the concentration of TV consumption increased after TSTV was available. In 2015, our industrial partner run a randomized experiment in which a random set of households was selected to obtain access to a new set of TV channels broadcasting movies and TV shows. A random subset of these households obtained access to these channels with TSTV and another random subset of them obtained access to these channels without TSTV. Using difference-in-differences we find that, on average, the former set of households consumed 4.5 more minutes of TV per day (p < 0.01), from a baseline of 5.0 hours, and as much live TV as the latter set of households. In addition, the consumption of TV by the former set of households was more concentrated towards the most popular programs. Finally, we show that households do not seem to use TSTV as a new tool to strategically avoid ads. In particular, and in 2015, households given access to the new TV channels with TSTV exit ads in the original TV channels as much as the households given access to these channels without TSTV. Therefore, the concern of advertisers that TSTV may reduce their revenues is unwarranted. How- ever, advertisers do not have data on TV audiences. Instead, content distributors do. Therefore, the latter are in a unique powerful position to deploy auction-based systems to sell TV ad slots taking advantage of the fine-grained information they have on TV viewership, much like websites do today online using cookies.

Peer Influence in Viral Products: Empirical Evidence from a Large Mobile Network.
Belo, R; Ferreira, P.

Abstract

This paper analyzes the role of peer influence in the diffusion of telecom-related products across social networks. We study a subset of the products deployed by a large European mobile carrier and look at how their viral characteristics affect diffusion. We develop a theoretical model of the consumers’ incentives to adopt these products and find that apparently similar viral designs can result in very different adoption dynamics. We use randomization to identify peer influence from observational data. In short, we shuffle adoption dates to create pseudo worlds where peer influence is inexistent. This allows us to capture, and thus separate, homophily. We prove that this approach provides a lower bound for the effect of peer influence. We find that for some products more adopters lead to more adoption. However, for other products more adopters lead to less adoption. These empirical results come in line with our theoretical model and show empirical evidence of peer effects that reduce adoption even for products with positive network effects. These effects might limit considerably the diffusion of some products countering the potential benefits from viral designs.

The Welfare Effects of Recommender Systems: Results from a Randomized Experiment with Video on Demand.
Zhang, X; Ferreira, P; Belo, R; Matos, M.

Abstract

Recommender systems assign products to slots in ways that improve consumers’ experience when choosing what to buy. They usually lead to more sales, which increases both consumer surplus and profit. However, firms may also choose which recommender system to use to maximize profit. Furthermore, consumers tend to exhibit a lower price elasticity of demand towards products placed in salient slots. We show that a firm using this knowledge may increase its profit hurting both consumer surplus and total welfare. We use data from a large scale field experiment ran using the video-on- demand system of a large telecommunications provider to measure the price elasticity of demand for movies placed in salient and non-salient slots on the TV screen. During this experiment, the firm randomized the slots in which movies were recommended to consumers as well as their prices. This readily allows for identifying the effects of price and slot on demand and thus compute consumer surplus. We find empirical evidence that indeed consumers are less price elastic towards movies placed in salient slots. Using the outcomes of this experiment we simulate how consumer surplus and welfare change when the firm implements the recommender system that maximizes profit. We also show that, at least in our setting, this system still yields higher consumer surplus than some recommender systems often used in practice, such as lists of most sold, most rated and highest rated products. Our results question whether recommender systems embed mechanisms that extract excessive surplus from consumers, which may need to be better scrutinized.

Price Discounts and Peer Effects in Information Goods: Results from an In-vivo Organic Randomized Experiment.
Matos, M; Ferreira, P; Belo, R.

Abstract

We look at the interplay between price promotions and peer influence in the case of movies in Video-on-Demand (VoD), which are non-storable information goods. We analyze the outcomes from a randomized experiment run for 3 consecutive months using the VoD system of a large telecommunications provider. We show that households with access to movies priced 25% lower than usual lease 11.1% more of these movies than households that never had access to movies at reduced prices. However, they also lease 3.3% fewer of the movies without price discounts during the entire experiment, which reduces aggregate sales by 2.9% hurting the provider’s profitability. We use cell phone call detail records from this provider to infer a graph of social proximity across households. The average degree in this graph is 10.23 friends. Using this graph, we find a positive effect of peer influence in the consumption of movies in this VoD system, which can be strategically used by the firm to issue price promotions minimizing profit losses. Firms can break-even if they offer price promotions to households with enough connections to generate enough sales through peer influence to counter the undesirable effect of price promotions. We show how the ability of the firm to break-even depends on the magnitude of the effect of peer influence, the discount offered and the markup factor. At the average of the covariates observed in our setting, the firm breaks-even if it offers price promotions to households with more than 4 connections.

Competition, Consumer Churn and Switching Costs (CCCSC)

Target the Ego or Target the Group: Evidence from a Randomized Experiment in Proactive Churn Management.
Matos, M; Ferreira, P; Belo, R.

Abstract

This paper analyzes outcomes from a randomized field experiment run with major telecommunications provider aimed at measuring the effect of proactive churn management among triple-play users and at determining whether using social network data may help retain them. We used state-of-the-art machine learning algorithms to develop a model to identify likely churners. We also used millions of call detailed records to identify their friends. A random subset of likely churners were selected to be contacted by the firm’s call center. We also randomly selected whether their friends would be contacted by the call center. We find that proactively listing likely churners to be contacted reduced their propensity to churn by 1.9 percentage points from a baseline of 17.2%. When their friends were also listed to be contacted the likelihood of churn reduced an additional 1.3 percentage points. We find that despite the discounts offered by the firm to retain likely churners and the false positives from our predictive model of churn, the NPV of likely churners increased 2.2% with proactive churn management. This statistic becomes 6.4% when their friends were also considered for churn management purposes. Furthermore, we provide some evidence that when likely churners obtain retention offers from the company they confer with their friends before accepting them. Our results show that managers should consider proactive churn management and that using social network data to design churn management campaigns is likely to increase retention rates.

The Impact of Mobile Number Portability on Price and Consumer Welfare.
Cho, D, Ferreira, P, and Telang, R.

Abstract

This paper examines the effect of Mobile Number Portability (MNP) on market price and consumer surplus. MNP reduces switching costs by allowing consumers to keep their phone number when they change service provider. Most European countries introduced MNP in the early 2000s as a result of a mandate from the European Commission. This supra-national legislative shock provides a unique opportunity to study the relationship between switching costs, price and consumer surplus. Theory shows that market prices can either increase or decrease when switching costs reduce and that followers may try to take advantage of decreasing switching costs to attract consumers from the market leaders. Using quarterly data from 47 wireless service providers in 15 EU countries between 1999 and 2006, we find that MNP decreased market price by at least 4.15% and increased consumer welfare by at least 2.15 euros per person per quarter on average during our period of analysis. This amounts to 880 Million euros per quarter across the 15 EU countries analyzed in this paper and 15% of the observed increase in consumer surplus during the period of analysis. This result is obtained using cost-shifters to instrument changes in price, which allows for determining the demand curve in the markets we analyze. Our study shows how the European experience with regards to the introduction of MNP can be used as an example of best practice by other countries that plan to introduce it in the near future.


Working Papers

Impact of Information and Communication Technologies on Education (IICTE)

Peer-Effects in Student Performance at the University Level.
Turner, R, Ferreira, P and Belo, R.  INFORMS Annual Meeting.  San Francisco, CA,  November 6 - 12, 2014.

Abstract

There is mixed evidence on the effect of peers on student performance and little evidence at all on whether friends influence grades at the University level. Yet, peer-based education has been pursue in recent times has a natural mechanism to improve student performance. In this paper we use randomization techniques to measure the effect of peers on student performance. Randomization allows us to separate homophily from peer influence under very mild assumptions. We use a panel of data on more than 10k students in an Engineering school between 2008 and 2010. We have data on grades per course passed and we use data on Wi-Fi sessions to proxy friendships. We find positive peer effects for worse student in all curricular years and for better students in later curricular years. We find no peer effects for better students in early curricular years. More interestingly, we find that the positive effect on worse students is greater than that on better students in later curricular years but not in early ones. This provides evidence in favor of tracking students in early years and grouping them in later years, that is, peer influence might only be beneficial in later stages of education when maturity abounds. A draft of this paper will be available soon.

Effect of Wi-Fi on Student Performance in a University Campus.
Turner, R, Ferreira, P, and Belo, R.  INFORMS Annual Meeting.  Minneapolis, MN,  October 6 - 9, 2013.

Abstract

The Internet can be beneficial to students but can also be a disruptive distraction. Many schools are anxious to deploy Information and Communications Technologies (ICTs) such as Wi-Fi networks but the literature remains undecided as to whether, and in what contexts, Wi-Fi is academically productive. Using a panel of data on Wi-Fi sessions and grades of more than 10k students in an Engineering school between 2008 and 2010, we show that students who use more Wi-Fi have better grades. Using minimal Wi-Fi usage as a proxy for having a laptop, we show that laptop ownership alone is not associated to changes in performance. These results are obtained using dynamic propensity score matching and are robust across majors, curricular years and cohorts. A draft of this paper will be available soon.

Peer-Influence and Consumption in the Media Industry (PICMI)

Can Personalization be as Bad as Price Discrimination? A Theoretical Model.
Zhang, X; Ferreira, P; Matos, M; Belo, R.
Binge Yourself Out: The Effect of Binge Watching on the Consumption of Media.
Matos, M; Ferreira, P.  Conference on Digital Experiments (CODE@MIT).  Massachusetts Institute of Technology, Cambridge, MA,  October 14-15, 2016.
The Effect of Friends and Crowds Throughout the Consumer Shopping Journey.
Yang, C; Matos, M; Ferreira, P.  Workshop on Information Systems and Economics (WISE).  Dublin, Ireland,  December 10-12, 2016.
The Effect of Website Blocking on Piracy.
Reis, F; Matos, M; Ferreira, P.  Workshop on Information Systems and Economics (WISE).  Dublin, Ireland,  December 10-12, 2016.

Abstract

In this work we study the effectiveness of website blocking, in particular, DNS block- ing on deterring piracy activity and promoting the use of legal alternatives. In collaboration with a large telecommunications provider, we measure the impact of a website blocking protocol on household’s Internet and paid Video on Demand use. We find that after the start of the protocol, daily download traffic households who had previously used BitTorrent regularly decreased by 35% (p < 0.01) and upload traffic decreased by 40% (p < 0.01). We find no statistical significant impact on these households’ VoD expenditure except for a slight increase of 1.5 minutes/week (p < 0.1) of VoD view time. We complement this analysis with a district level analysis exploring some of the mechanisms affecting website blocking effectiveness. We find that as side effect of the blocking protocol, part of the BitTorrent user population became more sophisticated in their piracy behavior by searching for and adopting strategies to bypass blocks. Our results shows that districts where this behavior was more prevalent presented significantly higher torrent activity. Our study contributes to the literature on piracy control strategies by being the first to conduct a household level analysis of the effectiveness of batch DNS blocking of websites providing copyright infringing content. We also provide evidence on the heterogenous effect of this measure and the associated learning and increased sophistication of BitTorrent users in what regards block circumvention.

Understanding the Role of Social Networks on Labor Market Outcomes Using a Large Dataset from a Mobile Operator.
Reis, F; Ferreira, P.  International Conference on Information Systems (ICIS).  University of Texas at Dallas, Fort Worth,  December 13-16, 2015.

Abstract

We use a new and unique dataset combining social network data from Call Detail Records with employment information on mobile phone subscribers to study the role of informa- tion networks on job market outcomes. The novel contribution of our work is to focus on the effect of actual social connections beyond that associated to living in the same neigh- borhood. We find that the propensity to work together is two orders of magnitude greater for a pair of neighbors who call each other than that for a pair of neighbors who do not, suggesting that actual social ties play a significant role in learning about job opportunities. We also find that social networks play a stronger role in less privileged neighborhoods, which provides some evidence that social networks may be unable to mitigate the insulation problems of such neighborhoods.

Competition, Consumer Churn and Switching Costs (CCCSC)

Measuring Switching Costs in the Telecom Industry.
Yang, C; Matos, M; Ferreira, P.  Workshop on Information Systems and Technology (WITS).  University of Texas at Dallas, Fort Worth,  December 13-16, 2015.

Other Refereed Publications

Impact of Information and Communication Technologies on Education (IICTE)

Peer-Influence and Consumption in the Media Industry (PICMI)

Competition, Consumer Churn and Switching Costs (CCCSC)