Netspar Design Paper No. 183; September 2021

The pension industry, much like the rest of the financial industry, is increasingly adopting data science and artificial intelligence-based solutions. Applications range from leaner and faster operations (“doing the same thing better”) to completely new value propositions. However, the available literature suggests that the pension industry appears to be relatively conservative and cautious when it comes to adopting new and dynamically changing machine learning (ML) techniques. The black box nature of most ML techniques also appears to contribute to the skepticism of the pension sector. Hence, there seems to be a gap between the potential applications of data science solutions proposed by researchers and their application in the pension industry. This article provides (i) a review of what has been reported in the data science literature, (ii) a taxonomy of ML techniques that can be applied for challenges in the pension industry, and (iii) a categorization of the different aspects of the pension industry that are covered in state-of-the-art applied data science.
We surveyed 25 papers and presentations on the application of data science in the pension industry and highlight the major machine learning techniques that were used and their applicability in the pension sector. These techniques are concisely introduced to provide a basis for stakeholders to gain an understanding of their potential applicability to tackle challenges in the pension industry. Based on the existing research, three areas of the pension industry are identified as most relevant for the application of machine learning techniques: customer focus, organizational process optimization, and personnel optimization. Open issues and further opportunities regarding the application of data science in the pension sector are discussed.
We surveyed the existing body of literature to summarize how data science is being currently leveraged to deal with issues related to pensions. Prominent developments appear along the fronts of prediction and chatbot development. Our analysis suggests that there remains ample room in the pension industry to explore the use of other machine learning and data mining methodologies, such as clustering, natural language processing, and reinforcement learning. This includes gleaning insights from unconventional sources such as social media activity, and developing new customer-focused and business development applications.
:: Available at Netspar Download paper.

Netspar Survey Paper No. 58; July 2021

Many economic and financial decisions involve risk. A crucial question in this context is how much risk people are willing to take. The importance of measuring the willingness of people to take risk is acknowledged by Dutch law, which requires financial institutions to take into account their customers' risk preferences when offering products and services. Additionally, knowledge of participants’ risk preferences play an important role in the recent pension agreement in the Netherlands, since it requires pension providers to invest in line with the age composition and risk attitudes of participants. Here we discuss the concept of risk preferences and common ways to measure them as proposed in the economics literature. We summarize traditional and behavioral economics perspectives on decisions under risk. Next, we present several incentivized experimental measures and non-incentivized survey measures of risk preferences and describe how they correlate with field behavior, and we discuss advantages and disadvantages of the different methods. We also discuss what is known, according to the economics literature, about the relationship between risk preferences and age.
:: Available at Netspar Download paper.
:: Media coverage Pensioen Pro PDF

CESifo Working Paper No. 8809, IZA Discussion Paper No. 13991; December 2020

The gender gap in income and leadership positions in many domains of our society is an undisputed pervasive phenomenon. One explanation for the disadvantaged position of women put forward in the economic and psychology literature is the weaker response of women to competitive incentives. Despite the large amount of literature trying to explain this fact, the precise mechanisms behind the gender difference in competitive responsiveness are still not fully uncovered. In this paper, we use laboratory experiments to study the potential role of stereotype threat on the response of men and women to competitive incentives in mixed-gender competition. We use a real effort math task to induce an implicit stereotype threat against women in one treatment. In additional treatments we, respectively, reinforce this stereotype threat and induce a stereotype threat against men. In contrast to much of the literature we do not observe that women are less competitive than men, neither when there is an implicit nor when there is an explicit stereotype threat against women. We attribute this to two factors which differentiates our experiment from previous ones. We control, first, for inter-individual performance differences using a within-subject design, and, second, for risk differences between non-competitive and competitive environments by making the former risky. We do find an adverse stereotype threat effect on the performance of men when there is an explicit stereotype threat against them. In that case any positive performance effect of competition is nullified by the stereotype threat. Overall, our results indicate that a stereotype threat has negative competitive performance effects only if there is information contradicting an existing stereotype. This suggests that the appropriate intervention to prevent the adverse effect of stereotype threat in performance is to avoid any information referring to the stereotype.
:: Available at SSRN Download paper.

CESifo Working Paper No. 8750, IZA Discussion Paper No. 13916; November 2020

Men and women negotiate differently, which might create gender inequality in access to resources as well as efficiency losses due to disagreement. We study the role of gender and gender pairing in bilateral bargaining, using a lab-in-the-filed experiment in which pairs of participants bargain over the division of a fixed amount of resources. We vary the gender composition of the bargaining pairs as well as the disclosure of the participants' identities. We find gender differences in earnings, agreement and demands, but only when the identities are disclosed. Women in same-gender pairs obtain higher earnings than men and women in mixed-gender pairs. This is the result of the lower likelihood of disagreement among women-only pairs. Women leave more on the bargaining table, conditional on their beliefs, which contributes to the lower disagreement and higher earnings among women-only pairs.
:: Available at SSRN Download paper.

CESifo Working Paper No. 8623; October 2020

Evidence from hypothetical scenarios strongly suggests the existence of a sunk cost bias, the tendency to `throw good money after bad money.' However, the few studies using incentives are inconclusive. In addition, evidence on potential psychological channels underlying such a bias is scarce. We present a laboratory experiment designed to investigate the sunk cost bias and to test some prominent psychological mechanisms. Inspired by the hypothetical scenarios, we use a two-stage investment task in which an initial investment needs to be made to start a project. In the initial investment stage, the size of the investment and the responsibility of the investor are exogenously varied. In the second investment stage, participants can either decide to terminate the project or to make an additional investment to finish the project. We do not find evidence for the sunk cost bias. To the contrary, we observe a robust reverse sunk cost bias. That is, the larger the initial investment, the lower the likelihood to continue investing in a project. Moreover, whether or not subjects are responsible for the initial investment, does not affect their additional investment. More waste averse individuals also do not react more strongly to sunk cost whereas being in the loss domain decreases additional investment. Importantly, we replicate the sunk cost bias when using hypothetical scenarios. Surprisingly, the reverse sunk cost bias also holds for those participants who exhibit a strong sunk cost bias in the hypothetical scenarios.
:: Available at SSRN Download paper.

CESifo Working Paper No. 8058, IZA Discussion Paper No. 12912; January 2020 (revised August 2020)

Labor market opportunities and wages may be unfair for various reasons, and how workers respond to different types of unfairness can have major economic consequences. Using an online labor platform, where workers engage in an individual task for a piece-rate wage, we investigate the causal effect of neutral and gender-discriminatory unfair chances on labor supply. We randomize workers into treatments where we control relative pay and chances to receive a low or a high wage. Chances can be fair, unfair based on an unspecified source, or unfair based on gender discrimination. Unequal pay reduces labor supply of low-wage workers, irrespective of whether the low wage is the result of fair or unfair chances. Importantly, the source of unfair chances matters. When a low wage is the result of gender-discriminatory chances, workers matched with a high-wage worker substantially reduce their labor supply compared to the case of equal low wages (-22%). This decrease is twice as large as those induced by low wages due to fair chances or unfair chances coming from an unspecified source. An additional experiment confirms the deleterious effect of gender discrimination on labor supply in a work environment devoid of chances, and highlights that workers' beliefs about facing discrimination matter for their responses. Our results concerning gender discrimination indicate a new reason for the lower labor supply of women, which is a prominent explanation for the gender gap in earnings.
:: Available at SSRN Download paper.
:: Video of a presentation at Middlebury College Virtual Seminar (July 2020) Watch video.

Netspar Design Paper No. 128; 2019

Collective pension systems are based on the principle of solidarity across and within generations. Using methods from experimental economics we elicit preferences for solidarity of different age cohorts towards the same and other age cohorts for a representative sample of the Dutch population. In addition, we use survey methods to measure stated inter- and intra-generational altruism and solidarity attitudes with respect to the Dutch pension system. Finally, we analyze how revealed solidarity preferences are related to demographic and socio-economic characteristics of participants by linking the experimental and survey data to administrative data of the Dutch population maintained by Statistics Netherlands (CBS). In our study, participants made decisions in a ‘solidarity game’ where real money was at stake. They had to decide how they would share money with another participant in case they received the money while the other participant did not. We found that participants were willing to share about 40% of the money received. This shows that participants on average had a strong preference for (ex-ante) solidarity. However, the results also show large heterogeneity as well as a bias in favor of the age group that a participant belongs to. A significant percentage of young participants were prepared to share substantially more with other young participants than with other age groups. Likewise, a large percentage of old participants were prepared to share substantially more with other old participants than with other age groups. We also find that beliefs about what others will give (i.e. anticipated reciprocity) correlate strongly with solidarity preferences: participants who expect more when in need show more solidarity towards others who are likewise in need. Remarkably, participants were rather pessimistic about the solidarity of others within and across age groups in the sense that they expected others to give substantially less than they themselves would actually do. Finally, a number of demographic and socio-economic characteristics (gender, education, marital status, political involvement) of our participants are statistically significantly related to the elicited solidarity preferences. In the survey, participants of all age categories indicated that they believe that solidarity between the young and the old is under pressure. We also see clear differences between young and old participants regarding the preferred pension system. While old participants favor a collective system over an individual system, for young participants the reverse holds. Opinions on the Dutch pension system correlate to some extent with the elicited solidarity preferences. Overall, however, the correlation between survey and experimental measures is relatively weak. This indicates that pension policymakers should take into account that stated preferences on solidarity do not necessarily reflect the true solidarity preferences of individuals.
:: Available at SSRN Download paper.

arXiv:1908.02287, 2019

In this paper we present our preliminary work on monitoring data License accoUntability and CompliancE (LUCE). LUCE is a blockchain platform solution designed to stimulate data sharing and reuse, by facilitating compliance with licensing terms. The platform enables data accountability by recording the use of data and their purpose on a blockchain-supported platform. LUCE allows for individual data to be rectified and erased. In doing so LUCE can ensure subjects' General Data Protection Regulation's (GDPR) rights to access, rectification and erasure. Our contribution is to provide a distributed solution for the automatic management of data accountability and their license terms.
:: Available at arXiv.org Download paper.

GSB_RM13034, 2013 (revised June 2017)

We elicit reciprocal preferences in a firm-worker gift-exchange setting and relate them to actual behavior in a repeated gift-exchange game. We find that only a small minority of 10 percent of workers is materially selfish whereas 90 percent exhibit reciprocal preferences. However, the intensity of reciprocal preferences is weak in the sense that firms maximize profits by not relying on gift-exchange but by offering the lowest possible wage. Workers behavior in the repeated gift-exchange game is predicted by their elicited preferences, but the correlation between preferences and behavior is imperfect. Together with profit maximizing behavior of firms these observations can explain the observed unraveling of gift-exchange over time in our experiment and some recent field experiments.
:: Download paper.

Maastricht University, GSBE Research Memorandum No. 16024, 2016

Social identity has been shown to successfully enhance cooperation and effort in cooperation and coordination games. Little is known about the causal effect of social identity on the propensity to engage in group conflict. In this paper we explore theoretically and experimentally whether social identity increases investments in group contests. We show theoretically that increased social identity with the own group implies higher investments in Tullock contests. Empirically we find that induced social identity does increase group closeness but does not increase conflict investments.
:: Download paper.

(July 2013)

To accurately predict behavior economists need reliable measures of individual time preferences and attitudes toward risk and typically need to assume stability of these characteristics over time and across decision domains. We test the reliability of two choice tasks for eliciting discount rates, risk aversion, and probability weighting and assess the stability of these characteristics over time and across situations. We find high reliability and that individual characteristics are remarkably stable over time. The estimated parameters correlate well with self-reported decisions in financial domains, but are largely uncorrelated with decisions in other important life domains involving intertemporal trade-offs and risk.
:: Download paper.

(May 2013; first version: December 2010 entitled "Directed Generosity and Network Formation: Network Dimension Matters")

We empirically analyze the relation between resource sharing, social proximity, and structural network positions in different network dimensions. We elicit socio-economic characteristics of all household heads in a rural village in Nicaragua, map their complete network in the dimensions friendship, social-public activities, and economic exchange, and conduct a sequence of field dictator experiments to measure the willingness to share resources in a controlled way. Different network dimensions differ substantially in structure and show little overlap. Relational and structural positions of individuals in these network dimensions correlate strongly with important socio-economic characteristics. Resource sharing is positively related to social proximity in friendship networks but not in other network dimensions. In all network dimensions resource sharing correlates with structural network variables, such as centrality and closure. These relations vary across network dimensions suggesting that for theoretical analysis as well as policy applications both network structure and network dimensions have to be taken into account.
:: Download paper.

(revised December 2008; first version: April 2004)
An abridged version of this paper is published in Journal of Public Economics under the title
We study the effects of competition in a context in which people's actions can not be contractually fixed. We find that in such an environment the very presence of competition does neither increase efficiency nor does it yield any payoff gains for the short side of the market. We also find that competition has a strong negative impact on social well-being, the disposition towards others, and individually experienced well-being, the emotional state, of those on the long side of the market. We conjecture that this limits the possibilities of satisfactory interaction in the future and, hence, has negative implications for efficiency in the longer-run.
:: Download paper. : Experimental instructions available upon request.

(Revised June 2004)

We investigate experimentally whether emotions affect bidding behavior in a first price auction. To induce emotions, we confront subjects after a first auction series with a positive or negative random economic shock. We then explore the relation between emotions and bidding behavior in a second auction series. Our main results are: (i) the economic shock has a substantial impact on the experienced emotions of bidders; (ii) the emotional state systematically influences bidding behavior. Our results clearly suggest that emotions should not be excluded as an explanatory factor of behavior in competitive environments like auctions.
: Paper available upon request.

(2003)

We extend the standard ultimatum game to a three person game where the proposer chooses a three-way split of a pie and two responders independently and simultaneously choose to accept or reject the proposal. We investigate whether a responder perceives the other responder as a reference person. We do this by varying the other responder's payoff in case the responder rejects. Hence, we explore whether reciprocal behavior towards the proposer is affected by the presence of the third player. In three treatments, the third player is either negatively affected, unaffected, or positively affected by the responder's choice to punish the proposer. We find that responders are very heterogeneous in their actions. Around one half of subjects submit strategies showing no concern for the other responder's payoffs. Another half of the subject pool submits strategies sensitive to the distribution of the pie among all three players. Preferences for equal splitting of the pie are expressed by less than 10 percent of all responders.
: Paper available upon request.