As recommender systems have increasingly become prevalent to guide consumers to find their desired products in many industries, understanding the impact of recommender systems on consumer choices is critical to the business performance and raises important policy implications. In this thesis, we examine the role of different recommendation schemes, spanning from interpersonal recommendations in social environment given by peers to product display recommendations in physical shopping environment given by sellers on consumers’ switching and search behavior in two distinct case studies. In the first study, we look at the effect of peer recommendations on subscriber churn in a large mobile network. We find that consumers’ propensity to churn increases with the number of friends that churn and in particular with the number of strong friends that churn. In the second study, we implement an in-vivo randomized field experiment to measure the effect of product display recommendations as book placement on shopper behavior in a physical bookstore. We leverage video tracking technologies to monitor how shoppers respond to random book placement, which induces random search costs. We find that books recommended at the edge of the table are more likely to be picked and taken than those placed at the center of the table. More interestingly, we also find that conditional on being picked, shoppers are equally likely to take books placed at the edge and at the center of the table, suggesting that display recommendations positively affect consumer choice mainly through its effect on the search process and not through its effect on the consideration process. Therefore, we empirically show that provision of recommendations, although in different schemes, may generally help to reduce consumers’ search costs in product or service discovery process, relative to what they would do without such an intervention. Moreover, we perform a comparative analysis between offline and online applications of recommender systems to systematically investigate the current practices, future prospects and policy perspectives when applying recommender systems in physical retailing. All these issues highlight opportunities for physical retailers to design, implement and evaluate their recommender systems that offer convenience benefits and appropriate protection to consumers.

The entertainment industry is an impactful part of the U.S. economy. My thesis explores the way Americans consume popular music and how the U.S. economic environment affects the permeability of the music industry to new artists. I use discrete-choice probit models to examine the top 10 weekly singles from the Billboard Hot 100 between 2006 and 2013. I analyze the economic factors and artist characteristics that affect an unestablished artist’s entry into the top 10 of the chart and achievement of the number one chart spot. I also use a Cox proportional hazard model to examine the effects of economic factors and artist characteristics on the number of weeks an artist’s single stays in the top 10 of the Hot 100 chart. I find that having a previous single in the top 100 decreases the predicted probability of a new artist’s song being in the top 10, and having previous singles in the top 10 or top 100 decreases number of weeks an artist’s subsequent single spends in the top 10 of the chart. Additionally, level of GDP per capita increases the number of weeks an artist’s single stays in the top 10 of the chart. Economic well-being perpetuates stability in the consumption of music, and modern culture consumers demonstrate a preference for newness in their endorsement of unestablished artists. As demonstrated by the use of singles between 2006 and 2013, new technologies decrease the costs of engaging with new artists for consumers and allow an artist to achieve success regardless of the artist’s previous success.

This doctoral dissertation documents price-setting behavior in Brazil using a unique dataset of store-level price quotes collected by the Fundação Instituto de Pesquisas Econômicas (FIPE) to construct the Consumer Price Index (CPI-FIPE) from 1989 to 2007. The dataset is extensive in terms of time (222 months), inflation variability (from hyperinflation to monthly deflation), and basket of goods and services (almost 11 million price quotes on 8,294 brands). The first chapter documents new evidence on the frequency and absolute size of price changes during the sample period. I find evidence of marked differences between hyperinflation (1989-1993) and low inflation (1995-2007) periods. During hyperinflation, the frequency and magnitude of price movements are remarkably higher. Once Plano Real took place, both statistics immediately shifted to a much lower and stable level, as did inflation. Price increases are more frequent during hyperinflation, although a small share of prices (mostly food items) drops every month. During low inflation, price decreases are almost as likely as price increases. I also document heterogeneities across different classifications of products. The second chapter investigates the relationship between inflation and relative price variability (RPV). The intramarket RPV significantly increases with the rate of inflation, but I find marked differences between the two inflationary scenarios. During hyperinflation, the relationship is roughly 70% of the magnitude of the relationship during low inflation. Higher levels of inflation are associated with higher degrees of inflation variability, yet the link is somewhat looser during the hyperinflation period. The impact of deflation (in absolute terms) is smaller than the impact of positive inflation during hyperinflation, yet stronger during low inflation. Finally, the third chapter documents the importance of Plano Real on consumers\' search costs. I estimate a nonsequential search model for homogeneous goods to structurally retrieve search costs using price data on 15 different brands of goods and services. The empirical strategy consists of using the Plano Real as a structural breakpoint in the data. I estimate the model splitting the data into before (January 1993 to June 1994) and after (August 1994 to December 1995) the plan, and I find evidence on first-order stochastic dominance of the search cost distribution of the former into the latter; that is, search costs are higher during hyperinflation. I also document evidence of the effect of the plan on shrinking price-cost margins. When searching is less costly, stores lose market power. / Esta tese de doutorado documenta comportamentos de fixação de preços no Brasil através de uma base de dados única de cotações ao nível da loja coletadas pela Fundação Instituto de Pesquisas Econômicas (FIPE) para a construção do Índice de Preços ao Consumidor (IPC-FIPE) de 1989 a 2007. Minha base de dados é extensa em tempo (222 meses), variabilidade da inflação (de hiperinflação à deflação mensal) e cesta de bens e serviços (quase 11 milhões de cotações sobre 8.294 marcas). O primeiro capítulo documenta novas evidências sobre a frequência e o tamanho absoluto das mudanças de preços durante o período da amostra. Eu encontro diferenças marcantes entre os períodos de hiperinflação (1989-1993) e baixa inflação (1995-2007). Durante a hiperinflação, a frequência e a magnitude dos movimentos de preços são notavelmente maiores. Após o Plano Real, ambas as estatísticas imediatamente mudam para um nível significativamente mais baixo e estável, seguindo o movimento da inflação. Aumentos de preços são mais frequentes durante a hiperinflação, embora uma pequena parcela de preços (principalmente de alimentos) ainda se reduza a cada mês. Sob inflação baixa, reduções de preços são quase tão prováveis quanto aumentos de preços. Eu também documento heterogeneidades presentes em diferentes classificações de produtos. O segundo capítulo investiga a relação entre inflação e variabilidade de preços relativos (VPR). A VPR intra-mercado aumenta significativamente com a taxa de inflação, mas eu encontro diferenças marcantes entre os dois cenários inflacionários. Durante a hiperinflação, a relação é aproximadamente 70\\% menor do que a estimada sob inflação mais baixa. Níveis mais altos de inflação estão associados à maior variabilidade desta, mas a ligação é um pouco mais fraca durante o período de hiperinflação. O impacto de uma deflação (em termos absolutos) é menor do que o impacto de um aumento de preço durante a hiperinflação, porém mais forte durante níveis mais baixos de inflação. Finalmente, o terceiro capítulo documenta a importância do Plano Real sobre os custos de busca (search costs}) dos consumidores. Eu estimo um modelo de busca não sequencial por bens homogêneos para recuperar estruturalmente os custos de busca dos consumidores utilizando dados de preços sobre 15 marcas diferentes de bens e serviços. A estratégia empírica consiste em usar o Plano Real como um ponto de quebra estrutural nos dados. Eu estimo o modelo dividindo os dados entre antes (de janeiro de 1993 a junho de 1994) e depois (de agosto de 1994 a dezembro de 1995) do plano e encontro evidências de dominância estocástica de primeira ordem da distribuição do custo de busca do primeiro sobre o segundo período, ou seja, os custos de busca são maiores durante a hiperinflação. Eu também encontro evidências do efeito do plano na redução da margem de preço (markup) das empresas. Quando buscar preços é menos custoso, firmas perdem poder de mercado.

Geographic space is an important friction preventing the instantaneous matching of unemployed workers to job vacancies. Cities reduce spatial frictions by decreasing the average distance between potential match partners. Owing to these search efficiencies, theories of agglomeration predict that unemployed workers in larger labor markets find employment faster than observationally similar workers in smaller markets. Existing studies rely on cross-sectional variation in aggregate unemployment rates across spatially distinct labor markets to test for scale effects in job search. A major difficulty with these studies is that the unemployment rate is, at any given time, simultaneously the incidence and duration of unemployment. Therefore, conclusions about unemployment exits using the unemployment rate are confounded by transitions into unemployment. This dissertation examines the relationship between market scale unemployment duration for permanently laid off workers in the U.S. Using a large sample of individual unemployment spells in 259 MSAs, proportional hazard model estimates predict a negative relationship between market scale and the hazard of exiting unemployment. This effect is strengthened when space is explicitly controlled for and measured with greater precision. These results are consistent with the hypothesis that search efficiencies lead workers to increase their reservation wages. 2SLS estimates show that re-employment earnings for permanently laid off workers increase with market scale after controlling for endogenous search duration. These effects are robust to standard controls, as well as controls for local labor market conditions. These results challenge the view that search efficiencies lead to lower unemployment rates through faster job-finding rates.

No description available.

The Law of One Price (LOP) is one of the most frequently tested economic laws. In simple terms the LOP states that once price is converted to a common currency, the same goods should sell for the same price in different countries. We have extracted the data from the price comparison website PriceSpy over seven different countries to analyze the trend in price dispersion. The methodologies we have used were regression for comparison of three different country pairs along with descriptive statistics to analyze the price dispersion trends related to the exchange rate. We concluded that there are many factors to make the fluctuation in prices and deny the LOP.

[pt] Este trabalho procura entender padrões de busca do consumidor e se fricções informacionais desempenham um papel na dispersão de preços no varejo brasileiro de gasolina. Na nossa abordagem, os consumidores devem se engajar em busca custosa para obter informação sobre os preços cobrados pelos postos de gasolina. Empiricamente, dividimos nossa análise em duas partes. Na primeira, utilizamos um modelo estrutural que nos permite estimar pontos da distribuição dos custos de busca. Estimamos o modelo usando dados de preços no nível do posto para vários mercados no Brasil. Na segunda parte, em duas análises independentes, investigamos os determinantes da proporção de consumidores com baixa quantidade de busca por OLS e construímos uma estimativa para o custo médio de busca por mercado encaixando nossas estimativas pontuais em uma distribuição paramétrica por NLS. Nossas descobertas revelam uma variação significativa na busca do consumidor entre os mercados. Além disso, nossos resultados revelam que a maioria dos consumidores não compara muitos preços antes de comprar gasolina. Ademais, nossas estimativas indicam que o número de postos de gasolina em um mercado, a distância média entre os postos, a renda e a população são fatores importantes para explicar a proporção de consumidores que procuram em apenas um posto antes de comprar. Por fim, o custo médio estimado de busca representa 3 por cento dos preços da gasolina, proporção esta não desprezível. Portanto, os resultados indicam que os atritos de informação são importantes para explicar a dispersão de preços no varejo brasileiro de gasolina. / [en] This paper seeks to understand consumer search patterns and whether information frictions play a role in price dispersion in Brazilian gasoline retail. In our setting, consumers must engage in costly search to gain information about the prices charged by gas stations. Empirically, we divide our analysis into two parts. In the first part, we use a structural model that permits us to estimate points of the distribution of search costs. We estimate the model using price data at the station level for multiple markets in Brazil. In the second part, in two independent analyzes, we investigate the determinants of the proportion of consumers with a low amount of search by OLS and construct an estimate for the average search cost per market by fitting our point estimates into a parametric distribution by NLS. Our findings reveal significant variation in consumer search across markets. Furthermore, our results reveal that most consumers do not compare many prices before buying gasoline. Moreover, our estimates indicate that the number of gas stations in a market, the average distance between gas stations, income, and population are important drivers of the proportion of consumers that search in only one gas station before buying. Finally, the estimated average search cost represents 3 percent of gasoline prices, a non-negligible proportion. Therefore, the results indicate that information frictions are important to explain price dispersion in Brazilian gasoline retail.

Submitted by Manuela Lopes (manu.hlopes@hotmail.com) on 2012-03-23T05:02:09Z No. of bitstreams: 1 discount.pdf: 259598 bytes, checksum: 22629d2c1b0a6a09240f05febeeddef1 (MD5) / Approved for entry into archive by Andrea Virginio Machado (andrea.machado@fgv.br) on 2012-03-23T12:49:52Z (GMT) No. of bitstreams: 1 discount.pdf: 259598 bytes, checksum: 22629d2c1b0a6a09240f05febeeddef1 (MD5) / Made available in DSpace on 2012-03-27T18:24:13Z (GMT). No. of bitstreams: 1 discount.pdf: 259598 bytes, checksum: 22629d2c1b0a6a09240f05febeeddef1 (MD5) Previous issue date: 2011-12-19 / Prices of a homogenous package of services present considerable dispersion across banks. A simple model that uses consumer search costs and loyalty is devolped in order to reproduce this empirical observation. Using data for the brazilian banking sector, empirical exercises are performed to allow for: (i) the estimation of consumer search costs by xing the values of other parameters; and (ii) the estimation of the corresponding deadweightlosses imposed by costly search. When only 80% of the population is free to engage in search activity, at a 0.5% monthly interest rate, the upper limit of the support of the search cost distribution is found to be 1805.80 BRL. In this case, the corresponding estimate of the deadweightloss imposed on an average searcher is 233.71 BRL. / Pode-se observar uma considerável dispersão entre os preços que diferentes bancos comerciais no Brasil cobram por um mesmo pacote homogêneo de serviços— dispersão esta que é sustentada ao longo do tempo. Em uma tentativa de replicar esta observação empírica, foi desenvolvido um simples modelo que lança mão do arcabouço da literatura de custos de procura (search costs) e que baseia-se também na lealdade por parte dos consumidores. Em seguida, dados de preços referentes ao setor bancário brasileiro são aplicados ao modelo desenvolvido e alguns exercícios empíricos são então realizados. Esses exercícios permitem que: (i) os custos de procura incorridos pelos consumidores sejam estimados, ao fixar-se os valores dos demais parâmetros e (ii) as correspondentes perdas de peso-morto que surgem como consequência dos custos de procura incorridos pelos consumidores sejam também estimadas. Quando apenas 80% da população é livre para buscar por bancos que cobrem menores tarifas, à taxa de juros mensal de 0,5%, o valor estimado do custo de procura médio incorrido pelos consumidores chega a 1805,80 BRL, sendo a correspondente perda de peso-morto média na ordem de 233,71 BRL por consumidor.

近年來台灣房屋仲介市場可說發展的相當快速，許多賣方和買方常基於成本的考量，而透過仲介公司來買賣房屋。當然，委託給仲介公司來買賣房屋。其中可能主要因素即著眼於縮短交易的時間或者是提高成交的機率，本研究的整個重心將圍繞在賣方的訂價高低(底價)賣屋動機與房屋的屬性來解釋銷售期間和成交機率的高低，並且本文擬得用搜尋理論和代理人理論來詮釋房地產仲介市場的交易行為。 本研究主要分成六章，第一章為緒論。第二章從搜尋理論的觀點來研究訂價與成交價和銷售期間的關係。我們從賣方決定訂價之後，買方會依據訂價要求折扣率，而賣方也會決定一願意給買方的最小折扣率的角度出發，來建構房屋搜尋模型。由理論得知，賣方折扣率底線愈大，則預期銷售期間愈長；銷售期間愈長，則預期賣方折扣率底線愈低；搜尋成本愈大，則預期賣方折扣率底線愈小；買方要求折扣率分配的平均數愈大，則預期賣方折扣率底線愈大。在實證研究上，利用信義房屋仲介公司所提供的資料(1990－1993)，以銷售期間和訂價相對於成交價的比例為應變數，採用聯立方程式的方式來估計分析。實證結果發現上述幾項論點皆獲得驗證與支持。 第三章模型同時考慮賣方和仲介業的行為，一方面說明賣方搜尋成本和買方所要求折扣率分配對銷售期間和成交機率的影響，另一方面提供未來修正理論模型的基礎。利用存活分析法(survival analysis)來估計銷售期間。一方面，我們要探究影響房屋交易之銷售期間的可能原因為何？另一方面，我們也想了解銷售期間是否具有時間相依性(time dependence)，是否銷售期間越長者，越不易賣出？或越容易賣出？ 第四章利用logit模型來估計房屋成交的機率，由於危險模型為純粹之計量模型，而非由理論模型所導出，所以分配之假設將限制其估計模型與理論的關連性，而logit模型則無此問題，因此本研究亦嘗試用logit模型來做估計。 第五章利用代理與搜尋模型來分析賣方和仲介業間的行為關係。不同的仲介收費制度--固定百分比收費(fixed-percentage commission)、定額制(flat－fee)、代銷(consignment sale)，對雙方利益衝突的衝擊為何，是否潛藏著嚴重的道德危險(moral hazard)。透過本篇的分析，可瞭解何以固定百分比收費制度是台灣房屋仲介市場的收費趨勢。最後一章為結論與未來研究方向。 / This paper employs search theory to study the re1ationships between the list price, the transaction price,and marketing duration in the Taiwan real estate market. Theoretically, buyer uses the a set of criteria together,with the listing price to develop an offer which is based on a (guest) minimurn discount rate guideline form the listing price to determine a price that will be acceptable to the seller. We attempt to describe the impact of pricing strategies (seller's minimum discoun rate) and marketing duration by incorporating the minimurn discount rate in a search model. The derived search model indicates the presence of a positbive retalionship between the minimurn discount rate and marketing duration;an inverse relationship beteween marketing duration and munimurn discount rate; an inverse relationship between searching costs and mininurn discount rate; and a positive relationship between the buyer's average discount rate of distribution function and minimum discount rate. The study uses data collected during the l990-1993 time period and provided by Hsin Yi Realty Co., with the dependent variables being the marketing duration and the ratio of the listing price to the transaction price. A simultaneous equation is developed and used to analyze the following hypotheses: firstly, the higher the ratio of the listing price against the transaction price, the longer it takes to reach the marketing duraion; secondly, the longer it takes to reach the marketing duraion, the higher the ratio of the listing price against the transaction price; thirdly, the longer the period of consignment, the longer it takes to reach the marketing duration; and the fourth,if the seller is not in a hurry to sell,then the marketing duration becomes longer. Our empirical findings verify and support all the above stated analyses.

近年來，在世界各地金融公司所謂的「指數股票型基金」已經成為主流投資標的。指數股票型基金讓投資人能夠將其資金直接投資於國外股市，同時避免匯率問題。目前，台灣投資人藉由所謂的中國指數股票型基金來直接投資中國股市。 身為中國指數股票型基金的先驅者，台灣匯豐中華證券投資信託股份有限公司於2009年夏天創造了第一恆生跨境指數股票型基金。就像其他服務業，匯豐中華必須向投資人推銷該指數股票型基金以及其他金融產品。有鑑於行銷對於資產管理公司的成敗扮演著很重要角色，因此本研究以策略行銷分析為主要研究基礎。並且，藉由公司訪談和邱志聖（2006)的四個成本分析架構(4C Analysis)，來探討台灣匯豐中華證券投資信託股份有限公司在台灣的行銷活動。所謂的4C分析包含外顯單外效益成本、資訊搜尋成本、道德危機成本及專屬陷入成本。本研究為台灣匯豐中華提供服務、媒體廣告、品牌形象及行銷策略的改善之道。 關鍵字：策略行銷、四個成本分斯、外顯單外效益成本、資訊搜尋成本、道德危機成本及專屬陷入成本、指數股票型基金 / Since the early 2000's, Exchange Traded Funds (ETFs) have become a mainstream investment product in the portfolios of financial companies across the world. In recent years, ETFs have been created to allow investors to allocate their capital in the stock market of other countries while at the same time avoiding the currency exchange problem. Now, Taiwanese people can invest in the China stock market indexes via China ETFs. A pioneer in China ETFs, HSBC Global Asset Management Taiwan created the first Heng Seng cross border ETFs in the summer of 2009. As with any service-oriented business, HSBC Taiwan must advertise and promote these new investment products. In light of the fact that marketing plays a large role in the success of asset management companies, hence this thesis adopts strategic marketing analysis as the backbone of the research process. Utilizing professional interviews and Chiu's (2006) four cost (4C) analysis, this thesis examines HSBC Global Asset Management Taiwan's marketing activities via a constructively critical lens. These four costs include external unit costs, information search costs, moral hazard costs and firm-specific costs. This thesis focuses on providing solutions for enhanced customer service, media advertising, brand image as well as marketing strategies for HSBC Taiwan. Key Words: strategic marketing, four cost structure, information search costs, moral hazard costs, firm-specific costs, exchange-traded funds