Expected Shortfall Regression for Censored Data

Abstract

Expected shortfall (ES), which conveys information regarding potential exceedances beyond the value-at-risk (VaR), is an important measure to characterize the properties of the tails of distribution. In this article, we study two two-step estimation procedures for ES regression with censored responses. Considering the potential dependence between the censoring variable and the covariates, two locally weighted estimation algorithms are proposed based on local Kaplan–Meier estimation and the joint elicitability of VaR and ES. The potential applications of this work are manifold, especially survival analysis, pharmacodynamic analysis, and sociological investigations. The resulting estimators are shown to be consistent. Extensive simulations demonstrate that the proposed method performs quite well in finite samples, with regard to estimation bias and mean squared errors. Last, the analysis of a real dataset illustrates the usefulness of our developed methodologies.

The online version contains supplementary material available at https://doi.org/10.1007/s40304-023-00357-3.

Keywords

Random censoring / Expected shortfall regression / Kaplan–Meier / Value-at-risk / 62N01 / 62G08

References

[1] BeranRNonparametric Regression with Randomly Censored Survival Data, 1981, Berkeley. University of California.

[2] BuchinskyM, HahnJ. An alternative estimator for the censored quantile regression model. Econometrica, 1998, 66(3): 653-671.

[3] ChenS. Sequential estimation of censored quantile regression models. J. Econom., 2018, 207(1): 30-52.

[4] ChenX, LintonO, Van KeilegomI. Estimation of semiparametric models when the criterion function is not smooth. Econometrica, 2003, 71(5): 1591-1608.

[5] De BackerM, El GhouchA, Van KeilegomI. Linear censored quantile regression: a novel minimum-distance approach. Scand. J. Stat., 2020, 47(4): 1275-1306.

[6] De BackerM, GhouchAE, Van KeilegomI. An adapted loss function for censored quantile regression. J. Am. Stat. Assoc., 2019, 114(527): 1126-1137.

[7] De Uña-ÁlvarezJ, Rodríguez-CamposMC. Strong consistency of presmoothed Kaplan-Meier integrals when covariables are present. Statistics, 2004, 38(6): 483-496.

[8] DimitriadisT, BayerS. A joint quantile and expected shortfall regression framework. Electr. J. Stat., 2019, 13(1): 1823-1871

[9] FisslerT, ZiegelJF. Higher order elicitability and Osband’s principle. Ann. Stat., 2016, 44(4): 1680-1707.

[10] GerdsTA, BeyersmannJ, StarkopfL, FrankS, LaanMJ, SchumacherM. The Kaplan-Meier integral in the presence of covariates: a review. Stat. Math. Financ., 2017.

[11] GneitingT. Making and evaluating point forecasts. J. Am. Stat. Assoc., 2011, 106(494): 746-762.

[12] Gonzalez-ManteigaW, Cadarso-SuarezC. Asymptotic properties of a generalized Kaplan-Meier estimator with some applications. J. Nonparamet. Stat., 1994, 4(1): 65-78.

[13] HosmerDW, LemeshowSApplied Survival Analysis: Regression Modeling of Time to Event Data, 1999, New York. Wiley.

[14] JácomeMA, GijbelsI, CaoR. Comparison of presmoothing methods in kernel density estimation under censoring. Comput. Stat., 2008, 23(3): 381-406.

[15] KalbfleischJD, PrenticeRLThe Statistical Analysis of Failure Time Data, 2011, Heidelberg. Wiley.

[16] KaplanEL, MeierP. Nonparametric estimation from incomplete observations. J. Am. Stat. Assoc., 1958, 53(282): 457-481.

[17] KoenkerR. Censored quantile regression redux. J. Stat. Softw., 2008, 27: 1-25.

[18] LeTH. Forecasting value at risk and expected shortfall with mixed data sampling. Int. J. Forecast., 2020, 36(4): 1362-1379.

[19] LengC, TongX. A quantile regression estimator for censored data. Bernoulli, 2013, 19(1): 344-361.

[20] McNeilAJ, FreyR, EmbrechtsPQuantitative Risk Management: Concepts, 2005, Princeton. Princeton University Press.

[21] MengX, TaylorJW. Estimating value-at-risk and expected shortfall using the intraday low and range data. Eur. J. Op. Res., 2020, 280(1): 191-202.

[22] MerloL, PetrellaL, RaponiV. Forecasting VAR and ES using a joint quantile regression and its implications in portfolio allocation. J. Bank. Financ., 2021, 133106248.

[23] NadarajahS, ZhangB, ChanS. Estimation methods for expected shortfall. Quant. Financ., 2014, 14(2): 271-291.

[24] NecirA, RassoulA, ZitikisR. Estimating the conditional tail expectation in the case of heavy-tailed losses. J. Prob. Stat., 2010.

[25] NeweyWK. The asymptotic variance of semiparametric estimators. Econom. J. Econom. Soc., 1994, 62(6): 1349-1382

[26] NeweyWK, McFaddenD. Large sample estimation and hypothesis testing. Handb. Econ., 1994, 4: 2111-2245

[27] Olma, T.: Nonparametric estimation of truncated conditional expectation functions. arXiv preprint arXiv:2109.06150 (2021)

[28] PattonAJ, ZiegelJF, ChenR. Dynamic semiparametric models for expected shortfall (and value-at-risk). J. Econom., 2019, 211(2): 388-413.

[29] PengL, HuangY. Survival analysis with quantile regression models. J. Am. Stat. Assoc., 2008, 103(482): 637-649.

[30] PortnoyS. Censored regression quantiles. J. Am. Stat. Assoc., 2003, 98(464): 1001-1012.

[31] PowellJL. Least absolute deviations estimation for the censored regression model. J. Econom., 1984, 25(3): 303-325.

[32] PowellJL. Censored regression quantiles. J. Econom., 1986, 32(1): 143-155.

[33] StuteW. Consistent estimation under random censorship when covariables are present. J. Multivar. Anal., 1993, 45(1): 89-103.

[34] StuteW. The statistical analysis of Kaplan-Meier integrals. Lect. Notes-Monogr. Ser., 1995, 27: 231-254

[35] StuteW. Kaplan-Meier integrals. Handb. Stat., 2003, 23: 87-104.

[36] TaylorJW. Forecast combinations for value at risk and expected shortfall. Int. J. Forecast., 2020, 36(2): 428-441.

[37] TobinJ. Estimation of relationships for limited dependent variables. Econom. J. Econom. Soc., 1958, 26(1): 24-36

[38] WangHJ, WangL. Locally weighted censored quantile regression. J. Am. Stat. Assoc., 2009, 104(487): 1117-1128.

[39] WangHJ, ZhouJ, LiY. Variable selection for censored quantile regresion. Stat. Sin., 2013, 231145

[40] YingZ, JungS-H, WeiL-J. Survival analysis with median regression models. J. Am. Stat. Assoc., 1995, 90(429): 178-184.

Funding

National Natural Science Foundation of China(Grant No. 71973133)

Anhui Provincial Natural Science Foundation(Grant No. 2208085J41)

RIGHTS & PERMISSIONS

School of Mathematical Sciences, University of Science and Technology of China and Springer-Verlag GmbH Germany, part of Springer Nature