Testing Cognitive Hierarchy Assumptions” (Job Market Paper)

David Rojo Arjona, Daniel E. Fragiadakis and Ada Kovaliukaite.

Camerer et al. (2004) propose a Cognitive Hierarchy (CH) model that characterizes an individual as a “step k thinker” who anticipates a distribution of step 0 through step k players. In a lab experiment, we ask a subject how many participants she believes select each pure strategy in a series of games. We find strong evidence of CH beliefs that are well approximated by normalized Poisson distributions. In addition, we make a methodological contribution to the belief-elicitation literature: our results provide a proof-of-concept that our novel procedure is effective not only in theory, but also in practice.


“Measuring Trust: A Reinvestigation.”

Billur Aksoy, Haley Harwell, Ada Kovaliukaite and Catherine Eckel.

We reinvestigate the question first posed by Glaeser, Laibson, Scheinkman and Soutter (2000, GLSS hereafter): What is the best measure of trust for predicting trusting behavior? This important study, cited over 2,100 times, established that the behavior in the investment game, an incentivized measure of trust, is not correlated with the responses to the most widely used survey questions about trust, employed in the General Social Survey (GSS) and the World Values Survey (WVS). We use the GLSS protocol with one major change: we employ the original Berg et al. (1995) investment game instead of the modified version used in GLSS. The standard game endows both players, while the latter endows only the first mover, potentially changing the incentives that influence subjects’ behavior. In particular, the utility from trusting behavior for inequality averse individuals may be higher, if the second movers are not endowed. Thus, such players may appear to be more trusting even though they are simply inequality averse. This causes a distortion in the laboratory measure of trust and reduces its correlation with the survey measure of trust. In support of this concern, GLSS demonstrates that the survey measure of trust is not correlated with trusting behavior in their investment game, where the second mover is not endowed. After endowing the second mover, we find the opposite. Our finding suggests that trust is a single construct, whether measured by the survey questions or by an incentivized game. This can be masked if the incentivized measure of trust is confounded with other motives.




“Can Friendship Paradox Generate Biased Inference in Social Networks?”

Alexander L. Brown and Ada Kovaliukaite.

Literature in economics and sociology suggests that in certain social contexts individuals may possess biased beliefs regarding the behavior of the population. Jackson (2016) proposes that, in the context of socially complementary behaviors, such misperceptions may arise due to the correlation between individual preferences and social network connectivity. In this paper, we present a laboratory experiment, where we exogenously manipulate the correlation between induced individual preferences and social network connectivity. We demonstrate that subjects exhibit overconsumption of an abstract, socially-complementary good and upwards-biased beliefs regarding the consumption of others when this correlation is positive. However, participants choose optimal levels of consumption and possess accurate beliefs when this correlation is set to zero. Evidence from a continuous-time belief elicitation suggests that subjects use a naive averaging rule to form beliefs about the behavior of the population from their observation of social network neighbors.


“A (non-parameteric) Method to Evaluate The Significance and Power of Level-k Family.”

David Rojo Arjona, Daniel E. Fragiadakis and Ada Kovaliukaite.

Theories of non-equilibrium strategic thinking (Level-k and Cognitive Hierarchy) intend to describe how individuals actually behave. But how much of their descriptive accuracy is driven by being more permissive theories? We modify Selten (1991) axiomatic measure of predictive success to ensure individual consistency. By applying restrictions over observables to the individual data (echoing the revealed preference literature), we test these theories non-parametrically and quantify the economic losses for deviations from the theory. The non-parametric results are favorable for these theories and show that their predictive success is not due to their mechanical permissiveness.


Residential Electricity Consumption Field Experiments.

Ada Kovaliukaite, Steve Puller and Jeremy West.

Households are reluctant to sign up for automatic bill pay for their electricity consumption. Is this decision driven by the lack of information on the multiple benefits of automatic bill pay or by the sign-up costs? Does automatic bill pay reduce households’ responsiveness to bill shocks by reducing bill salience? In turn, does automatic bill pay reduce the probability of a household adopting energy efficient technology? We collaborate with Bryan Texas Utilities to encourage their residential customers to sign up for automatic bill pay through a random encouragement design. We randomly assign BTU customers to three treatments: a control that receives a standard news letter, an information treatment that receives a news letter with information on the benefits of automatic bill pay, and a sign-up cost treatment that receives a news letter with instructions on signing up for automatic bill pay. We use the treatment indicator as an instrumental variable for the automatic bill pay to measure the effects of auto pay on the responsiveness to the electric bill shocks in the summer months.