Some Internet stores manage to charge prices that are significantly higher than market averages, therefore, obtaining some sort of price premium. This paper is dedicated to building a model that can be used to explain and predict a typical price premium that an Internet store charges for a specific product based on the information about the characteristics of the store and the features of the market for this product. Such models can provide support for pricing and assortment decisions: in particular, they allow detecting products that a store is likely to sell with the highest or the lowest markup based on price premia that are charged by stores with similar characteristics on similar markets.

В данной работе представляется подход для автоматического определения сегментов, в которых регрессионная модель имеет существенно меньшую точность, и для выявления сегментов с систематической недооценкой или переоценкой предсказания. Данный сегментационный подход применим к различным экспертным системам, включая, но не ограничиваясь, те, которые используются для массовой оценки. Представленный подход может быть полезен для различных применений регрессионного анализа, особенно для тех, в которых сильная гетероскедастичность. Он помогает выявлять сегменты, для которых желательно использование отдельных моделей или помощь оценщика. Сегментационнный подход был применен к модели оценки недвижимости, основанной на алгоритме Random Forest.

In this paper we study the influence of superstitions related to numbers 13 and 7 on people's buying behavior in the apartment market. A unique feature of our methodology is that we use real sales data instead of survey or pricing data. Based on the dataset from Saint-Petersburg primary real estate market we compare the share of sold apartments on floor 7 with that of on floors 6 and 8, whereas floor 13 is benchmarked to floors 12 and 14. As floors are comprised by exactly the same apartments we manage to isolate the effects of the “lucky” and “unlucky” numbers. We have found a significantly negative effect of the 13th (“unlucky”) floor on demand for apartments in new apartment houses, but no significant positive effect of the 7th (“lucky”) floor. Possible implications of this result and directions for future research are discussed.

In this study we demonstrate how publicly available data can be used to work out the indirect importance of various hotel attributes for their visitors. We apply Shapley value decomposition of the recommendation rate to compute the percentage contributions of various attributes to the overall loyalty, which helps us explain why some of Cyprus hotels have higher satisfaction ratings than the others. It appeared that satisfaction with gastronomy is the key driver of the overall satisfaction with Cyprus hotels. We conduct importance-performance analysis for one of the hotels to demonstrate a strategic management application of our empirical analysis. Directions for other potential studies using user-generated content are proposed.

В статье исследуется влияние отдачи от человеческого капитала на самосохранительное поведение. В качестве одной из предпосылок исследования выступает трактовка здоровья как инвестиционного блага, комплиментарного по отношению к человеческому капиталу, вкупе с трактовкой последнего как фактических навыков и знаний, приносящих денежную премию. Предложена модель, связывающая спрос на здоровье с человеческим капиталом. Согласно модели, человеческий капитал определяет самосохранительное поведение через ожидаемое влияние здоровья на отдачу от человеческого капитала. Главный прикладной вывод из модели заключается в том, что образованные люди, если их образование не обеспечивает им денежной премии, не будут сильно отличаться от необразованных людей по параметрам самосохранительного поведения.

This paper compares relative performance evaluation via tournaments to absolute performance evaluation via piece rates when agents are heterogeneous ex post, to make the point that agent heterogeneity compromises the insurance function of tournaments. In particular, we show that the more heterogeneous agents are the less insurance can be offered through tournaments and the less dominant tournaments are over piece rates. Thus, absolute performance piece rates should be preferred when agents are highly heterogeneous. However, even with heterogeneous agents, tournaments become more desirable when the number of agents or the uncertainty about the common shock increases sufficiently/

In this paper we demonstrate how publicly available data from location-based social networks can be used to model the popularity of different places. Our empirical analysis is based on a sample of 112 shopping malls located in Saint- Petersburg, the second largest city in Russia. Regression models that explain various measures of shopping malls' popularity using the characteristics of a place are built. It appeared that the number of visitors, check-ins, tips and even the frequency of visit can be largely explained by the basic features of a shopping mall, while successful modeling of a place's user rating requires more detailed data. Combining the data on the features of places, text reviews and popularity indicators from social networks is a promising approach for building sales, traffic, satisfaction and loyalty projection models, which are especially useful in business planning.