Clustering Standard Errors at the "Session" Level

Kim, Duk Gyoo

doi:10.2139/ssrn.3635181

Cited by 4 publications

(5 citation statements)

References 15 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Since there is a subject effect and individuals’ residuals are correlated over rounds, we cluster the standard errors by participant. This does not affect the size of the coefficients or the model’s overall fit but robustly guards against Type I errors (Kim, 2022; Moffatt, 2016, pp. 84–86; Petersen, 2009; Van Pelt, 2020).…”

Section: Resultsmentioning

confidence: 99%

“…Like Petersen, we find clustered individual standard errors to be the most conservative approach. In selecting the level of clustering the standard errors (individual or session) we follow Kim (2022), Van Pelt (2020), and Moffatt (2016). See Supplemental Material for details on how the appropriate level of clustering the standard errors was determined for this set of data.…”

mentioning

confidence: 99%

See 1 more Smart Citation

Reciprocity or Community? Different Cultural Pathways to Cooperation and Welfare

Gunnthorsdottir

Thorsteinsson²,

Olafsson

2023

Cross-Cultural Research

View full text Add to dashboard Cite

We compare efficiency-enhancing cooperation and its underlying motives in Iceland and the US. The two countries are distinct along all measures of national culture known to us. They are however both developed democracies with similar GDP/capita (PPP adjusted). These similarities make it possible to hold constant aspects of culture related to wealth and institutions. In an experimental Voluntary Contribution Mechanism (VCM), we prime the participants with different social foci, emphasizing either their directly cooperating team or their wider social unit. With a team focus, cooperation levels do not differ between the two cultures, but this superficial similarity masks deep-seated differences: When the focus is on the wider social unit cooperation increases in Iceland and declines in the US. Both when the contribution levels are the same and when they differ, members of the two cultures differ in their motives to cooperate: Icelanders tend to cooperate unconditionally, and US subjects conditionally with a strong emphasis on reciprocity. Our findings indicate that different cultures can achieve similar economic and societal performance through different cultural norms and suggest that cooperation should be encouraged through culturally tailored persuasion tactics.

show abstract

Section: Resultsmentioning

confidence: 99%

mentioning

confidence: 99%

Reciprocity or Community? Different Cultural Pathways to Cooperation and Welfare

Gunnthorsdottir

Thorsteinsson²,

Olafsson

2023

Cross-Cultural Research

View full text Add to dashboard Cite

show abstract

“…the results section. Following the recent paper by Kim [38], we decide not to cluster the standard errors at the session level as doing so could produce false positive coefficients. Nonetheless, we have reported clustered standard errors in our pre-print version and the results are qualitatively the same [39].…”

Section: Plos Onementioning

confidence: 99%

Large losses from little lies: Strategic gender misrepresentation and cooperation

Drouvelis

Gerson²,

Powdthavee³

et al. 2023

PLoS ONE

View full text Add to dashboard Cite

This paper investigates the possibility that a small deceptive act of misrepresenting one’s gender to others reduces cooperation in the Golden Balls game, a variant of a prisoner’s dilemma game. Compared to treatments where either participants’ true genders are revealed to each other in a pair or no information on gender is given, the treatment effects of randomly selecting people to be allowed to misrepresent their gender on defection are positive, sizeable, and statistically significant. Allowing people to misrepresent their gender reduces the average cooperation rate by approximately 10–12 percentage points. While one explanation for the significant treatment effects is that participants who chose to misrepresent their gender in the treatment where they were allowed to do so defect substantially more, the potential of being matched with someone who could be misrepresenting their gender also caused people to defect more than usual as well. On average, individuals who chose to misrepresent their gender are around 32 percentage points more likely to defect than those in the blind and true gender treatments. Further analysis reveals that a large part of the effect is driven by women who misrepresented in same-sex pairs and men who misrepresented in mixed-sex pairs. We conclude that even small short-term opportunities to misrepresent one’s gender can potentially be extremely harmful to later human cooperation.

show abstract

“…These estimates are accompanied by two sets of SEs: one set is adjusted for clustering at the session level, the other is not. Clustering at the session level is common in statistical analyses of economic experiments, but Kim (2020) cautions against the practice, pointing out that in certain circumstances, the adjustment for clustering can bias the SEs towards zero. We will take a conservative approach, focusing on hypotheses that can be rejected at the 5% level with the larger of our two alternative SEs.…”

Section: The Effect Of the Social Norm On The Frequency Of Giving Typesmentioning

confidence: 99%

“…In addition, a number of studies examine the effect of having a lead donor (e.g Huck & Rasul, 2011;Karlan & List, 2007, 2020. or announcing the amount already raised (e.g.,List & Lucking-Reiley, 2002) on donations.…”

mentioning

confidence: 99%

In search of competitive givers

Fielding

Knowles

Peeters

2022

Southern Economic Journal

View full text Add to dashboard Cite

Much of the theoretical and experimental research on charitable giving allows for three main types of donor: pure altruists, impure altruists, and pure warm-glow givers. For none of these types should donations be increasing in the amount donated by others, and the fact that some experimental subjects do behave in this way suggests a fourth type, the "competitive giver". Our experimental results provide evidence for the existence of competitive givers. The results also suggest that most (but not all) competitive giving is a result of uncertainty about the social norm: when information about the norm is revealed, the incidence of competitive giving is much lower.

show abstract

Clustering Standard Errors at the "Session" Level

Cited by 4 publications

References 15 publications

Reciprocity or Community? Different Cultural Pathways to Cooperation and Welfare

Reciprocity or Community? Different Cultural Pathways to Cooperation and Welfare

Large losses from little lies: Strategic gender misrepresentation and cooperation

In search of competitive givers

Contact Info

Product

Resources

About