pith. sign in

arxiv: 2405.06850 · v3 · submitted 2024-05-10 · 💰 econ.EM

Identifying Peer Effects in Networks with Unobserved Effort and Isolated Students

Aristide Houndetoungan , Cristelle Kouame , Michael Vlassopoulos This is my paper

Pith reviewed 2026-05-24 01:24 UTC · model grok-4.3

classification 💰 econ.EM
keywords peer effectssocial networksunobserved effortisolated studentsGPAidentificationhigh school students
0
0 comments X

The pith

Peer effects on effort appear 40 percent larger once shocks to the outcome are separated from those that change effort itself.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper develops a method to estimate peer influence on effort when only an outcome like GPA is observed. Classical approaches treat the outcome as a direct stand-in for effort, but this mixes in unrelated shocks. The new framework separates shocks that alter the outcome without touching effort from preference shocks that do change effort levels. This separation produces different peer-effect estimates whenever the network contains isolated students. In US high-school data the standard proxy approach yields estimates 40 percent smaller than the corrected ones.

Core claim

Peer effects estimates obtained using the proposed approach, which distinguishes unobserved shocks to GPA that do not affect effort from preference shocks that do affect effort levels, can differ significantly from classical estimates that approximate effort with the observed outcome if the network includes isolated students. In an application to high school students in the United States, peer effect estimates relying on GPA as a proxy for effort are 40 percent lower than those obtained using the new approach.

What carries the argument

The separation of unobserved shocks to the outcome that leave effort unchanged from preference shocks that alter effort levels, which permits identification of peer effects even when isolated students are present.

If this is right

  • Classical estimates that use the outcome as a proxy for effort will be biased when isolated students exist in the network.
  • The size of the bias depends on the share of isolated students and the structure of preference shocks.
  • The corrected estimates can be recovered from standard outcome data once the two shock types are distinguished.
  • Applications to other networks with unobserved effort will produce different results from proxy-based studies whenever isolates are present.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Earlier studies of peer effects in education may have understated the strength of peer influence on effort.
  • The distinction between shock types could be tested in other outcome measures such as wages or health behaviors.
  • Network interventions aimed at raising effort may need recalibration if the true peer multiplier is larger than previously measured.

Load-bearing premise

Shocks to the observed outcome can be divided into two distinct types, one that leaves effort unchanged and one that changes effort.

What would settle it

Estimates of peer effects remain identical when the method is applied to the same network both with and without its isolated students.

Figures

Figures reproduced from arXiv: 2405.06850 by Aristide Houndetoungan, Cristelle Kouame, Michael Vlassopoulos.

Figure 1
Figure 1. Figure 1: Solving the reflection problem Note: Ñ means that the node on the right side is a friend of the node on the left side. We employ a proof by contradiction. The nonidentification issue arises when the vector EpJsGsys|Gs, Xsq is perfectly collinear with JsXs and JsGsXs. For a non-isolated student, this suggests that there exist vector of parameters, β9 and γ9 , such that Epgs,iys´y NI s |Gs, Xsq “ pxs,i ´ˆxsq… view at source ↗
Figure 2
Figure 2. Figure 2: Effects of Shocks on the GPA This figure presents the distribution of the increase in the GPA subsequent to a 0.1-unit increase in αs and cs for the student sample (n = 68,430). “fully isolated" students allows us to conduct a robustness analysis, as it does not involve a missing network data issue. We thus define a new subsample by excluding the “fully isolated" students from our main sample, resulting in… view at source ↗
Figure 3
Figure 3. Figure 3: Illustration of the identification Note: Ñ means that the node on the right side is a friend of the node on the left side. Many other situations lead to b1 “ b2 “ b3 “ 0. In practice, one can easily verify if Js, JspGs ` G1 s qJs and JsGsG1 sJs are linearly independent. C.2 Supplementary Results on the Estimation of pσ 2 ϵ , τ, ρq In this section, we use different notations for the parameters and their tru… view at source ↗
read the original abstract

Peer influence on effort devoted to some activity is often studied when effort is unobserved, and the researcher instead observes an outcome that combines effort with other shocks. For instance, in education, achievement measures such as GPA reflect both effort and idiosyncratic GPA shocks. We propose an alternative approach that circumvents this approximation. Our framework distinguishes unobserved shocks to GPA that do not affect effort from preference shocks that do affect effort levels. We show that peer effects estimates obtained using our approach can differ significantly from classical estimates (where effort is approximated) if the network includes isolated students. Applying our approach to data on high school students in the United States, we find that peer effect estimates relying on GPA as a proxy for effort are 40% lower than those obtained using our approach.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript proposes a structural framework to identify peer effects on unobserved effort in networks, where the observed outcome (such as GPA) is a composite of effort and idiosyncratic shocks. By distinguishing effort-neutral GPA shocks from effort-affecting preference shocks and exploiting isolated students to separate their distributions, the authors derive peer-effects estimates that differ from classical proxy-based approaches when the network contains isolates. In an application to US high-school data, classical estimates using GPA as a proxy for effort are reported to be 40% lower than those obtained with the proposed method.

Significance. If the separation of shock distributions via isolates is valid, the approach offers a direct way to recover peer effects on effort rather than on a composite outcome, addressing a common measurement problem in network econometrics and education research. The reported 40% discrepancy suggests that standard approximations may systematically understate peer influence when isolates are present, which could affect policy conclusions about social multipliers in achievement.

major comments (2)
  1. [Identification section] The central identification result hinges on the claim that isolated students allow separate recovery of the distributions of GPA shocks and preference shocks (abstract and identification section). The manuscript should explicitly derive or simulate how the presence of isolates breaks the observational equivalence that otherwise confounds the two shock types, including the precise moment conditions or likelihood contributions used.
  2. [Empirical results] Table reporting the 40% difference (empirical application): the comparison between the new estimates and the classical GPA-proxy estimates must include standard errors or confidence intervals on the difference itself, as well as the exact specification of the classical benchmark (e.g., which network moments or fixed effects are held constant).
minor comments (2)
  1. [Model section] Notation for the two shock processes should be introduced with explicit subscripts or superscripts to avoid confusion between the composite outcome and the latent effort equation.
  2. [Abstract] The abstract states the 40% figure without indicating whether it is an average across specifications or a single preferred estimate; a parenthetical note on the baseline specification would improve readability.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the positive assessment and the constructive suggestions for improving the manuscript. We address each major comment below.

read point-by-point responses

  1. Referee: [Identification section] The central identification result hinges on the claim that isolated students allow separate recovery of the distributions of GPA shocks and preference shocks (abstract and identification section). The manuscript should explicitly derive or simulate how the presence of isolates breaks the observational equivalence that otherwise confounds the two shock types, including the precise moment conditions or likelihood contributions used.

    Authors: We agree that greater explicitness on this point would strengthen the paper. While the identification section outlines how isolates permit separate recovery of the two shock distributions, we will add a dedicated derivation (including the relevant moment conditions) in the revised version to show precisely how the presence of isolates breaks observational equivalence between effort-neutral GPA shocks and effort-affecting preference shocks. revision: yes

  2. Referee: [Empirical results] Table reporting the 40% difference (empirical application): the comparison between the new estimates and the classical GPA-proxy estimates must include standard errors or confidence intervals on the difference itself, as well as the exact specification of the classical benchmark (e.g., which network moments or fixed effects are held constant).

    Authors: We accept this recommendation. The revised manuscript will report standard errors (or confidence intervals) on the difference between the new estimates and the classical GPA-proxy estimates. We will also clarify the exact specification of the classical benchmark, confirming that identical network moments and fixed effects are used in both approaches. revision: yes

Circularity Check

0 steps flagged

No significant circularity detected

full rationale

The paper's identification strategy separates effort-neutral GPA shocks from effort-affecting preference shocks by exploiting isolated students in the network. This separation is presented as an external source of variation that allows recovery of peer effects on effort rather than on the composite GPA outcome. No equation or step in the provided description reduces the estimated peer effect to a fitted parameter, self-citation chain, or definitional equivalence with the input data. The reported 40% difference from classical estimates follows from the model structure once the separation is granted, without evidence of tautological construction.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The central claim depends on the validity of separating shock types and the network structure including isolated students. No numerical free parameters or invented entities are specified in the abstract.

axioms (2)
  • domain assumption The outcome (GPA) is a combination of effort and idiosyncratic shocks
    Standard in education economics, implied by the abstract.
  • ad hoc to paper Preference shocks affect effort levels while other shocks do not
    This is the key distinction introduced in the framework.

pith-pipeline@v0.9.0 · 5658 in / 1275 out tokens · 35043 ms · 2026-05-24T01:24:20.894454+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

53 extracted references · 53 canonical work pages

  1. [1]

    Baysan, M

    Alan, S., C. Baysan, M. Gumren, and E. Kubilay (2021): Building social cohesion in ethnically mixed schools: An intervention on perspective taking, The Quarterly Journal of Economics, 136, 2147--2194

    work page 2021

  2. [2]

    Albert, J. H. and S. Chib (1993): Bayesian analysis of binary and polychotomous response data, Journal of the American Statistical Association, 88, 669--679

    work page 1993

  3. [3]

    Foster, N

    Arcidiacono, P., G. Foster, N. Goodpaster, and J. Kinsler (2012): Estimating spillovers using panel data, with an application to the classroom, Quantitative Economics, 3, 421--470

    work page 2012

  4. [4]

    Calvó-Armengol, and Y

    Ballester, C., A. Calvó-Armengol, and Y. Zenou (2006): Who's Who in Networks. Wanted: The Key Player, Econometrica, 74, 1403--1417

    work page 2006

  5. [5]

    Blume, L. E., W. A. Brock, S. N. Durlauf, and Y. M. Ioannides (2011): Identification of social interactions, in Handbook of social economics, Elsevier, vol. 1, 853--964

    work page 2011

  6. [6]

    Boucher, V. and B. Fortin (2016): Some Challenges in the Empirics of the Effects of Networks , Handbook on the Economics of Networks, 45--48

    work page 2016

  7. [7]

    Boucher, V. and A. Houndetoungan (2022): Estimating peer effects using partial network data, Centre de recherche sur les risques les enjeux \'e conomiques et les politiques …

    work page 2022

  8. [8]

    Tumen, M

    Boucher, V., S. Tumen, M. Vlassopoulos, J. Wahba, and Y. Zenou (2021): Ethnic Mixing in Early Childhood: Evidence from a Randomized Field Experiment and a Structural Model, Tech. rep., CEPR Discussion Paper No.15528

    work page 2021

  9. [9]

    Djebbari, and B

    Bramoull \'e , Y., H. Djebbari, and B. Fortin (2020): Peer effects in networks: A survey, Annual Review of Economics, 12, 603--629

    work page 2020

  10. [10]

    Djebbari, and B

    Bramoullé, Y., H. Djebbari, and B. Fortin (2009): Identification of peer effects through social networks, Journal of Econometrics, 150, 41--55

    work page 2009

  11. [11]

    Patacchini, and Y

    Calvó-Armengol, A., E. Patacchini, and Y. Zenou (2009): Peer Effects and Social Networks in Education, The Review of Economic Studies, 76, 1239--1267

    work page 2009

  12. [12]

    Galeotti, G

    Conti, G., A. Galeotti, G. Mueller, and S. Pudney (2013): Popularity, Journal of Human Resources, 48, 1072--1094

    work page 2013

  13. [13]

    Dustmann, and U

    Cornelissen, T., C. Dustmann, and U. Sch \"o nberg (2017): Peer effects in the workplace, American Economic Review, 107, 425--456

    work page 2017

  14. [14]

    d'Haultfoeuille, and D

    Davezies, L., X. d'Haultfoeuille, and D. Foug \`e re (2009): Identification of peer effects using group size variation, The Econometrics Journal, 12, 397--413

    work page 2009

  15. [15]

    Pellizzari, and S

    De Giorgi, G., M. Pellizzari, and S. Redaelli (2010): Identification of social interactions through partially overlapping peer groups, American Economic Journal: Applied Economics, 2, 241--275

    work page 2010

  16. [16]

    (2017): Econometrics of network models, in Advances in economics and econometrics: Theory and applications, eleventh world congress, Cambridge University Press, Cambridge, 268--323

    De Paula, \'A . (2017): Econometrics of network models, in Advances in economics and econometrics: Theory and applications, eleventh world congress, Cambridge University Press, Cambridge, 268--323

    work page 2017

  17. [17]

    --- -.1pt --- -.1pt --- (2020): Econometric models of network formation, Annual Review of Economics, 12, 775--799

    work page 2020

  18. [18]

    Duncan, G. J., J. Boisjoly, and K. Mullan Harris (2001): Sibling, peer, neighbor, and schoolmate correlations as indicators of the importance of context for adolescent development, Demography, 38, 437--447

    work page 2001

  19. [19]

    Durlauf, S. N. and Y. M. Ioannides (2010): Social interactions, Annu. Rev. Econ., 2, 451--478

    work page 2010

  20. [20]

    (2019): An empirical model of dyadic link formation in a network with unobserved heterogeneity, Review of Economics and Statistics, 101, 763--776

    Dzemski, A. (2019): An empirical model of dyadic link formation in a network with unobserved heterogeneity, Review of Economics and Statistics, 101, 763--776

    work page 2019

  21. [21]

    Epple, D. and R. E. Romano (2011): Peer effects in education: A survey of the theory and evidence, in Handbook of Social Economics, Elsevier, vol. 1, 1053--1163

    work page 2011

  22. [22]

    Fortin, B. and M. Yazbeck (2015): Peer effects, fast food consumption and adolescent weight gain, Journal of Health Economics, 42, 125--138

    work page 2015

  23. [23]

    Fruehwirth, J. C. (2013): Identifying peer achievement spillovers: Implications for desegregation and the achievement gap, Quantitative Economics, 4, 85--124

    work page 2013

  24. [24]

    --- -.1pt --- -.1pt --- (2014): Can achievement peer effect estimates inform policy? a view from inside the black box, Review of Economics and Statistics, 96, 514--523

    work page 2014

  25. [25]

    Goldsmith-Pinkham, P. and G. W. Imbens (2013): Social networks and the identification of peer effects, Journal of Business & Economic Statistics, 31, 253--264

    work page 2013

  26. [26]

    Graham, B. S. (2008): Identifying social interactions through conditional variance restrictions, Econometrica, 76, 643--660

    work page 2008

  27. [27]

    --- -.1pt --- -.1pt --- (2017): An econometric model of network formation with degree heterogeneity, Econometrica, 85, 1033--1063

    work page 2017

  28. [28]

    (2022): Name your friends, but only five? the importance of censoring in peer effects estimates using social network data, Journal of Labor Economics, 40, 779--805

    Griffith, A. (2022): Name your friends, but only five? the importance of censoring in peer effects estimates using social network data, Journal of Labor Economics, 40, 779--805

    work page 2022

  29. [29]

    Hastie, T. J. (2017): Generalized additive models, in Statistical models in S, Routledge, 249--307

    work page 2017

  30. [30]

    Hausman, J. A. (1978): Specification tests in econometrics, Econometrica: Journal of the econometric society, 1251--1271

    work page 1978

  31. [31]

    Hong, S. C. and J. Lee (2017): Who is sitting next to you? Peer effects inside the classroom, Quantitative Economics, 8, 239--275

    work page 2017

  32. [32]

    Horn, R. A. and C. R. Johnson (2012): Matrix analysis, Cambridge University Press

    work page 2012

  33. [33]

    Houndetoungan, E. A. (2022): Count data models with social interactions under rational expectations, Tech. rep., SSNR

    work page 2022

  34. [34]

    Hsieh, C.-S. and L. F. Lee (2016): A social interactions model with endogenous friendship formation and selectivity, Journal of Applied Econometrics, 31, 301--319

    work page 2016

  35. [35]

    Hsieh, C.-S. and X. Lin (2017): Gender and racial peer effects with endogenous network formation, Regional Science and Urban Economics, 67, 135--147

    work page 2017

  36. [36]

    Jackson, M. O. and Y. Zenou (2015): Games on networks, in Handbook of game theory with economic applications, Elsevier, vol. 4, 95--163

    work page 2015

  37. [37]

    (2023): Peer effects and endogenous social interactions, Journal of Econometrics, 235, 1203--1214

    Jochmans, K. (2023): Peer effects and endogenous social interactions, Journal of Econometrics, 235, 1203--1214

    work page 2023

  38. [38]

    Johnsson, I. and H. R. Moon (2021): Estimation of peer effects in endogenous social networks: control function approach, Review of Economics and Statistics, 103, 328--345

    work page 2021

  39. [39]

    Kelejian, H. H. and I. R. Prucha (1998): A generalized spatial two-stage least squares procedure for estimating a spatial autoregressive model with autoregressive disturbances, The Journal of Real Estate Finance and Economics, 17, 99--121

    work page 1998

  40. [40]

    Kline, B. and E. Tamer (2020): Econometric analysis of models with social interactions, in The Econometric Analysis of Network Data, Elsevier, 149--181

    work page 2020

  41. [41]

    (2000): The incidental parameter problem since 1948, Journal of Econometrics, 95, 391--413

    Lancaster, T. (2000): The incidental parameter problem since 1948, Journal of Econometrics, 95, 391--413

    work page 2000

  42. [42]

    (2004): Asymptotic distributions of quasi-maximum likelihood estimators for spatial autoregressive models, Econometrica, 72, 1899--1925

    Lee, L.-F. (2004): Asymptotic distributions of quasi-maximum likelihood estimators for spatial autoregressive models, Econometrica, 72, 1899--1925

    work page 2004

  43. [43]

    --- -.1pt --- -.1pt --- (2007): Identification and estimation of econometric models with group interactions, contextual factors and fixed effects, Journal of Econometrics, 140, 333--374

    work page 2007

  44. [44]

    Liu, and X

    Lee, L.-f., X. Liu, and X. Lin (2010): Specification and estimation of social interaction models with network structures, The Econometrics Journal, 13, 145--176

    work page 2010

  45. [45]

    (2010): Identifying Peer Effects in Student Academic Achievement by Spatial Autoregressive Models with Group Unobservables , Journal of Labor Economics, 28, 825--860

    Lin, X. (2010): Identifying Peer Effects in Student Academic Achievement by Spatial Autoregressive Models with Group Unobservables , Journal of Labor Economics, 28, 825--860

    work page 2010

  46. [46]

    Manski, C. F. (1993): Identification of Endogenous Social Effects: The Reflection Problem, The Review of Economic Studies, 60, 531--542

    work page 1993

  47. [47]

    Mas, A. and E. Moretti (2009): Peers at work, American Economic Review, 99, 112--145

    work page 2009

  48. [48]

    Newey, W. K. and D. McFadden (1994): Large sample estimation and hypothesis testing, Handbook of Econometrics, 4, 2111--2245

    work page 1994

  49. [49]

    Rose, C. D. (2017): Identification of peer effects through social networks using variance restrictions, The Econometrics Journal, 20, S47--S60

    work page 2017

  50. [50]

    (2011): Peer effects in education: How might they work, how big are they and how much do we know thus far? in Handbook of the Economics of Education, Elsevier, vol

    Sacerdote, B. (2011): Peer effects in education: How might they work, how big are they and how much do we know thus far? in Handbook of the Economics of Education, Elsevier, vol. 3, 249--277

    work page 2011

  51. [51]

    (1957): Specification errors and the estimation of economic relationships, Revue de l'Institut International de Statistique, 25, 41--51

    Theil, H. (1957): Specification errors and the estimation of economic relationships, Revue de l'Institut International de Statistique, 25, 41--51

    work page 1957

  52. [52]

    (1980): A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity, Econometrica: journal of the Econometric Society, 817--838

    White, H. (1980): A heteroskedasticity-consistent covariance matrix estimator and a direct test for heteroskedasticity, Econometrica: journal of the Econometric Society, 817--838

    work page 1980

  53. [53]

    Jiang, S

    Yan, T., B. Jiang, S. E. Fienberg, and C. Leng (2019): Statistical inference in a directed network model with covariates, Journal of the American Statistical Association, 114, 857--868

    work page 2019