Collective Exposure: Peer Effects in Voluntary Disclosure of Personal Data

This paper reports empirical evidence for peer effects in privacy behavior using field data from online social lending. Our content analysis and regression models show that individuals copy observable behavior of others in decisions on a) how much to write about oneself, b) whether to share custom pictures, c) what personal data to disclose, and d) how identifiable to present oneself. We frame this finding in the theory of descriptive social norms and analyze moderating effects, such as similarity of context, social proximity, and mimicry of success factors. The presence of peer effects in disclosure behavior can explain the formation and change of apparent social norms and attitudes towards privacy. 1 Financial Privacy and Human Behavior Information technology has created an age of perfect memory, which raises issues of information privacy and informational self-determination. With legal and technical means available that in principle empower individuals to control the distribution of their personal data, researchers of all disciplines still lack a good understanding of how individuals make use of this control. Scholars of the economics of privacy (see [1] for a survey) assume rational individuals who consider all costs and benefits when making a decision to disclose personal data. Works in this tradition largely draw on analytical economic models to study the efficiency of privacy regimes in specific market situations [2–4]. Yet it remains doubtful if individual decisions to disclose personal data can be explained sufficiently well with models of rational agents. Instead, it has been suggested to approach the subject with theories borrowed from social psychology [5] and behavioral economics [6]. In this spirit, a number of behavioral biases affecting the decision to disclose personal data has recently been identified empirically: a general discrepancy between stated preferences and behavior [7], present-based biases and discounting [8], anchor effects [9], social norms [10], perceived control [11], and contextual primes [12]. All these results have been obtained from laboratory experiments. While experiments are the method of choice for exploring causality under controlled conditions, they oftentimes suffer from small samples and questionable ecological validity. 2 R. Böhme and S. Pötzsch The contribution of this paper is to complement the laboratory studies with new evidence from field data. More specifically, we explain voluntary disclosure of personal data with peer effects, that is, the tendency of individuals to mimic other peoples’ disclosure behavior. Our data are loan applications by real users of Smava.de, the largest German platform for online social lending. Also known as “P2P lending”, “Ebay for loans”, or “crowd-sourcing of finance”, online social lending has grown rapidly over the past couple of years [13]. Drawing on concepts of (offline) micro-finance, the idea of social lending is to provide a marketplace for unsecured loans between individuals: an online platform lets borrowers advertise loan applications to lenders, who decide in which loan they invest. Each lender funds only a small share of the financed amount so that the credit risk is shared in loan-specific pools of lenders. Lenders receive interest as a compensation for taking risk, whereas the platform operators typically charge fixed (i. e., risk-free) fees. Market mechanisms differ between platforms, a fact that led to research in mechanism design [14]. Online social lending is an ideal data source for the study of behavioral aspects of financial privacy. By their very nature, loan applications contain a lot of personal details, which enable lenders to assess the associated risk [15]. Data requirements and sharing arrangements already raise privacy concerns in the traditional banking industry [16]. These issues are further exacerbated in online social lending where personal data of loan applications does not remain within heavily regulated institutions, but is accessible to all Internet users [17]. Another feature of this data source is that loan applicants disclose their personal data to this audience voluntarily. (In fact, financial regulators require the platform operator to collect additional personal data, which is not disclosed to the public though.) This enables us to look for pattern that explain the influence of peers, i. e., other applicants, on the decision to disclose personal data. There is a clear link between peer orientation and the notion of herd behavior. The latter is an active field of research in finance, often entangled with the question of rationality [18]. At some level of sophistication, models can be found that explain herding and resulting bubbles as rational action. For a first cut on the topic of peer effects in personal data disclosure, we spare us the discussion of whether peer effects are rational or not (the former requires a model of competition between borrowers). We rather see our contribution in the description and rigorous measurement of the phenomenon based on longitudinal field data. This paper is organized straight. The following Section 2 develops seven hypotheses and introduces the data and analysis method. Results are presented in Section 3, and then discussed in Section 4.