Skip to main content
Log in

Estimation of Tempered Stable Lévy Models of Infinite Variation

  • Published:
Methodology and Computing in Applied Probability Aims and scope Submit manuscript

Abstract

Truncated realized quadratic variations (TRQV) are among the most widely used high-frequency-based nonparametric methods to estimate the volatility of a process in the presence of jumps. Nevertheless, the truncation level is known to critically affect its performance, especially in the presence of infinite variation jumps. In this paper, we study the optimal truncation level, in the mean-square error sense, for a semiparametric tempered stable Lévy model. We obtain a novel closed-form 2nd-order approximation of the optimal threshold in a high-frequency setting. As an application, we propose a new estimation method, which combines iteratively an approximate semiparametric method of moment estimator and TRQVs with the newly found small-time approximation for the optimal threshold. The method is tested via simulations to estimate the volatility and the Blumenthal-Getoor index of a generalized CGMY model and, via a localization technique, to estimate the integrated volatility of a Heston type model with CGMY jumps. Our method is found to outperform other alternatives proposed in the literature when working with a Lévy process (i.e., the volatility is constant), or when the index of jump intensity Y is larger than 3/2 in the presence of stochastic volatility.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Fig. 1
Fig. 2
Fig. 3
Fig. 4
Fig. 5

Similar content being viewed by others

Notes

  1. The term “tempered stable” is understood here in a more general sense than in several classical sources of financial mathematics (e.g., Applebaum (2004), Cont and Tankov (2004), Kyprianou et al. (2005)) and even more general than in Rosiński (2007). In fact, such class of Lévy processes is called the tempered-stable-like Lévy processes in Figueroa-López et al. (2016).

  2. Figueroa-López and Ólafsson (2016b) considers the asymmetric case \(\nu (dx)=C(x/|x|)\bar{q}(x)|x|^{-1-Y}\,dx\) with \(C(1)=0.015\) and \(C(-1)=0.041\). Here, we take \(C=(C(1)+C(-1))/2\) in order to simplify the simulation of the model. Our values of G, M, and Y are the same as in Figueroa-López and Ólafsson (2016b).

  3. In our code, we use the R package stabledist to generate them.

  4. In Steps 2 and 4, we choose the minimization method. We use the R function nloptr from the package nloptr with the algorithm NLOPT__LD__LBFGS.

References

  • Aït-Sahalia Y, Jacod J (2009) Estimating the Degree of Activity of Jumps in High Frequency Data. Ann Stat 37(5A):2202–2244

    Article  MathSciNet  Google Scholar 

  • Applebaum D (2004) Lévy Processes and Stochastic Calculus, 2nd Ed.. Cambridge Stud Adv Math, 116, Cambridge University Press, Cambridge, U.K.

  • Belomestny D (2010) Spectral Estimation of the Fractional Order of a Lévy Process. Ann Stat 38(1):317–351

    Article  MathSciNet  Google Scholar 

  • Bull AD (2016) Near-Optimal Estimation of Jump Activity in Semimartingales. Ann Stat 44(1):58–86

    Article  MathSciNet  Google Scholar 

  • Carr P, Geman H, Madan DB, Yor M (2002) The Fine Structure of Asset Returns: An Empirical Investigation. J Bus 75(2):305–332

    Article  Google Scholar 

  • Cont R, Tankov P (2004) Financial Modelling with Jump Processes. Chapman & Hall/CRC Financ Math Ser, Chapman & Hall/CRC, Boca Raton, FL, U.S.A.

  • Figueroa-López JE (2012) Statistical Estimation of Lévy-Type Stochastic Volatility Models. Ann Finance 8(2):309–335

    Article  MathSciNet  Google Scholar 

  • Figueroa-López JE, Gong R, Houdré C (2016) High-Order Short-Time Expansions for ATM Option Prices of Exponential Lévy Models. Math Financ 26(3):516–557

    Article  Google Scholar 

  • Figueroa-López JE, Gong R, Houdré C (2017) Third-Order Short-Time Expansions for Close-to-the-Money Option Prices under the CGMY Model. Appl Math Financ 24(6):547–574

    Article  MathSciNet  Google Scholar 

  • Figueroa-López JE, Mancini C (2019) Optimum Thresholding Using Mean and Conditional Mean Square Error. J Econom 208(1):179–210

    Article  Google Scholar 

  • Figueroa-López JE, Ólafsson S (2016a) Short-Time Expansions for Close-to-the-Money Options under a Lévy Jump Model with Stochastic Volatility. Financ Stoch 20(1):219–265

    Article  Google Scholar 

  • Figueroa-López JE, Ólafsson S (2016b) Short-Time Asymptotics for the Implied Volatility Skew under a Stochastic Volatility Model with Lévy Jumps. Financ Stoch 20(4):973–1020

    Article  Google Scholar 

  • Jacod J, Todorov V (2014) Efficient Estimation of Integrated Volatility in Presence of Infinite Variation Jumps. Ann Stat 42(3):1029–1069

    MathSciNet  MATH  Google Scholar 

  • Kawai R (2010) On Sequential Calibration for an Asset Price Model with Piecewise Lévy Processes. IAENG Int J Appl Math 40(4):239–246

    MathSciNet  MATH  Google Scholar 

  • Kyprianou A, Schoutens W, Wilmott P (2005) Exotic Option Pricing and Advanced Lévy Models. John Wiley & Sons Ltd., ChiChester, England

    MATH  Google Scholar 

  • Masuda H (2019) Parametric estimation of Lévy processes. In Lévy Matters IV, p. 179-286. Springer. Rate-Optimal Estimation of the Blumenthal-Getoor Index of a Lévy Process. Preprint.

  • Mies F (2019) Rate-Optimal Estimation of the Blumenthal-Getoor Index of a Lévy Process. Preprint. Electron J Stat 14(2):2020

  • Reiß M (2013) Testing the Characteristics of a Lévy Process. Stoch Proc Appl 123(7):2808–2828

    Article  Google Scholar 

  • Rosiński J (2007) Tempering Stable Processes. Stoch Proc Appl 117(6):677–707

    Article  MathSciNet  Google Scholar 

  • Sato K (1999) Lévy Processes and Infinitely Divisible Distributions. Cambridge Stud. Adv. Math., 68, Cambridge University Press, Cambridge, U.K.

  • Wang Q, Figueroa-López JE, Kuffner T (2021) Bayesian Inference on Volatility in the Presence of Infinite Jump Activity and Microstructure Noise. Elect J Stats 15(1):506–553

    MathSciNet  MATH  Google Scholar 

  • Zhang L, Mykland PA, Aït-Sahalia Y (2005) A Tale of Two Time Scales: Determining Integrated Volatility with Noisy High-Frequency Data. J Am Stat Assoc 100(472):1394–1411

    Article  MathSciNet  Google Scholar 

Download references

Acknowledgements

The authors are grateful to the Editor and two anonymous referees for their multiple suggestions that help to improve the original manuscript.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to José E. Figueroa-López.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Research supported in part by the NSF Grants: DMS-2015323, DMS-1613016.

Proof of Theorem 1

Proof of Theorem 1

We start with the decomposition:

$$\begin{aligned} \mathbb {E}\big (b_{1}(\varepsilon )\big )&=\mathbb {E}\Big (\big (\sigma W_{h}+J_{h}\big )^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )\nonumber \\&=\sigma ^{2}\,\mathbb {E}\Big (W_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )\!+\!\mathbb {E}\Big (J_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )\!+\!2\sigma \,\mathbb {E}\Big (W_{h}J_{h}{} \mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big ).\quad \end{aligned}$$
(33)

The asymptotic behavior of each of the terms above is obtained in three steps.

Step 1

We first analyze the behavior of the first term in Eq. (33) as \(h\rightarrow 0\). By Eqs. (7) and (8), we first decompose it as

$$\begin{aligned} \mathbb {E}\Big (W_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )&=h-\mathbb {E}\Big (W_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|>\varepsilon \}}\Big )=h-\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}-\eta h}\,W_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|>\varepsilon \}}\Big ) \nonumber \\&=h-he^{-\eta h}\,\widetilde{\mathbb {E}}\Big (W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big ) \nonumber \\&\quad -he^{-\eta h}\,\widetilde{\mathbb {E}}\Big (\Big (e^{-\widetilde{U}_{h}}-1\Big )W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big ) \nonumber \\&=:h-he^{-\eta h}I_{1}(h)-he^{-\eta h}I_{2}(h). \end{aligned}$$
(34)

In what follows, we will analyze the asymptotic behavior of \(I_{1}(h)\) and \(I_{2}(h)\), as \(h\rightarrow 0\), respectively, in two sub-steps.

Step 1.1

Clearly, by Eq. (9) and the symmetry of \(W_{1}\),

$$\begin{aligned} I_{1}(h)=\widetilde{\mathbb {E}}\Big (W_{1}^{2}\,\mathbf{1}_{\{\sigma \sqrt{h}W_{1}+J_{h}>\varepsilon \}}\Big )+\widetilde{\mathbb {E}}\Big (W_{1}^{2}\,\mathbf{1}_{\{\sigma \sqrt{h}W_{1}-J_{h}>\varepsilon \}}\Big )=:I_{1}^{+}(h)+I_{1}^{-}(h). \end{aligned}$$
(35)

Denote by

$$\begin{aligned} \phi (x):=\frac{1}{\sqrt{2\pi }}e^{-x^{2}/2},\quad \overline{\Phi }(x):=\int _{x}^{\infty }\phi (x)\,dx,\quad x\in \mathbb {R}. \end{aligned}$$

By conditioning on \(J_{h}\) and using the fact that \(\widetilde{\mathbb {E}}(W_{1}^{2}{} \mathbf{1}_{\{W_{1}>x\}})=x\phi (x)+\overline{\Phi }(x)\), for all \(x\in \mathbb {R}\), we have that

$$\begin{aligned} I_{1}^{\pm }(h)=\widetilde{\mathbb {E}}\left( \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\mp \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\mp \frac{J_{h}}{\sigma \sqrt{h}}\bigg )+\overline{\Phi }\bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\mp \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) . \end{aligned}$$
(36)

Let \(p_{J,h}^{\pm }\) be the density of \(\pm J_{h}\) under \(\widetilde{\mathbb {P}}\), and recall the Fourier transform and its inverse transform defined by

$$\begin{aligned} (\mathcal {F}g)(z):=\frac{1}{\sqrt{2\pi }}\int _{\mathbb {R}}g(x)e^{-izx}\,dx,\quad \big (\mathcal {F}^{-1}g\big )(x):=\frac{1}{\sqrt{2\pi }}\int _{\mathbb {R}}g(z)e^{izx}\,dz. \end{aligned}$$

In what follows, we set

$$\begin{aligned} \psi (x)&:=\bigg (\mathcal {F}^{-1}\phi \bigg (\frac{\cdot }{\sigma \sqrt{h}}-\frac{\varepsilon }{\sigma \sqrt{h}}\bigg )\bigg )(x)=\frac{1}{\sqrt{2\pi }}\int _{\mathbb {R}}\phi \bigg (\frac{z}{\sigma \sqrt{h}}-\frac{\varepsilon }{\sigma \sqrt{h}}\bigg )e^{izx}\,dz \nonumber \\&\,=\frac{\sigma \sqrt{h}\,e^{i\varepsilon x}}{\sqrt{2\pi }}\int _{\mathbb {R}}\phi (\omega )e^{i\sigma \sqrt{h}\omega x}\,d\omega =\frac{\sigma \sqrt{h}}{\sqrt{2\pi }}\exp \bigg (i\varepsilon x-\frac{1}{2}\sigma ^{2}x^{2}h\bigg ). \end{aligned}$$
(37)

Then, we deduce that

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) =\int _{\mathbb {R}}\big (\mathcal {F}\psi \big )(z)p_{J,h}^{\mp }(z)\,dz=\int _{\mathbb {R}}\psi (u)\big (\mathcal {F}p_{J,h}^{\mp }\big )(u)\,du, \end{aligned}$$

where

$$\begin{aligned} \big (\mathcal {F}p_{J,h}^{\pm }\big )(u)&=\frac{e^{\mp iu\widetilde{\gamma }h}}{\sqrt{2\pi }}\exp \!\left( \!-(C_{+}\!+C_{-})\,\bigg |\!\cos \bigg (\!\frac{\pi Y}{2}\!\bigg )\!\bigg |\Gamma (-Y)|u|^{Y}h\bigg (\!1-i\frac{C_{+}\!-C_{-}}{C_{+}\!+C_{-}}\tan \!\bigg (\!\frac{\pi Y}{2}\!\bigg )\text {sgn}(u)\!\bigg )\!\right) \nonumber \\&=:\frac{1}{\sqrt{2\pi }}\exp \Big (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)\mp iu\widetilde{\gamma }h\Big ), \end{aligned}$$
(38)

with \(c_{1}:=(C_{+}\!+C_{-})\,\cos (\pi Y/2)\Gamma (-Y)\) and \(c_{2}:=(C_{-}-C_{+})\sin (\pi Y/2)\Gamma (-Y)\). Hence, we have

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right)&=\frac{\sigma \sqrt{h}}{2\pi }\!\!\int _{\mathbb {R}}\exp \!\bigg (\!c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{1}{2}\sigma ^{2}u^{2}h+iu\big (\varepsilon \!\pm \!\widetilde{\gamma }h\big )\!\bigg )du \nonumber \\&=\frac{\sigma \sqrt{h}}{\pi }\int _{0}^{\infty }\exp \bigg (c_{1}u^{Y}h-\frac{1}{2}\sigma ^{2}u^{2}h\bigg )\cos \Big (c_{2}u^{Y}h+u\big (\varepsilon \pm \widetilde{\gamma }h\big )\Big )du\\&=\frac{1}{\pi }\int _{0}^{\infty }\exp \!\bigg (c_{1}\cdot \frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\cos \!\bigg (c_{2}\cdot \frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}+\omega \cdot \frac{\varepsilon \!\pm \!\widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&=\frac{1}{\pi }\int _{0}^{\infty }\!\exp \!\bigg (c_{1}\!\cdot \!\frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\!-\!\frac{\omega ^{2}}{2}\bigg )\cos \!\bigg (c_{2}\!\cdot \!\frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\cos \!\bigg (\omega \!\cdot \!\frac{\varepsilon \!\pm \!\widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\quad -\!\frac{1}{\pi }\!\int _{0}^{\infty }\!\exp \!\bigg (\!c_{1}\!\cdot \!\frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\!-\!\frac{\omega ^{2}}{2}\bigg )\sin \!\bigg (\!c_{2}\!\cdot \!\frac{\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\sin \!\bigg (\!\omega \!\cdot \!\frac{\varepsilon \!\pm \!\widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega . \end{aligned}$$
(39)

By expanding the Taylor series for \(\exp (c_{1}\sigma ^{-Y}|\omega |^{Y}h^{1-Y/2})\), as well as for \(\cos (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\) and \(\sin (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\), we deduce that

$$\begin{aligned}&\widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) \nonumber \\&\,\,\,\,=\frac{1}{\pi }\!\int _{0}^{\infty }\!\cos \!\bigg (\omega \!\cdot \!\frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )e^{-\omega ^{2}/2}d\omega \!+\!\frac{1}{\pi }\!\!\sum _{(k,j)\in \mathbb {Z}^{2}_{+}:\,(k,j)\ne (0,0)}\!\!(-1)^{j}a_{k,2j,0}^{\pm }(h)\!-\!\frac{1}{\pi }\sum _{k,j=0}^{\infty }(-1)^{j}d_{k,2j+1,0}^{\pm }(h) \nonumber \\&\,\,\,\,=\frac{1}{\sqrt{2\pi }}\exp \bigg (\!\!-\!\frac{(\varepsilon \pm \widetilde{\gamma }h)^{2}}{2\sigma ^{2}h}\bigg )+\frac{1}{\pi }\!\sum _{(k,j)\in \mathbb {Z}^{2}_{+}:\,(k,j)\ne (0,0)}\!(-1)^{j}a_{k,2j,0}^{\pm }(h)-\frac{1}{\pi }\sum _{k,j=0}^{\infty }(-1)^{j}d_{k,2j+1,0}^{\pm }(h), \end{aligned}$$
(40)

where, for \(m,n\in \mathbb {Z}_{+}\) and \(r\in \mathbb {R}_{+}\),

$$\begin{aligned} a_{m,n,r}^{\pm }(h)&:=\frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\int _{0}^{\infty }\omega ^{(m+n)Y+r}\cos \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )e^{-\omega ^{2}/2}\,d\omega ,\end{aligned}$$
(41)
$$\begin{aligned} d_{m,n,r}^{\pm }(h)&:=\frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\int _{0}^{\infty }\omega ^{(m+n)Y+r}\sin \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )e^{-\omega ^{2}/2}\,d\omega . \end{aligned}$$
(42)

By applying the formula for the integrals of \(\omega ^{kY}\cos (\beta \omega )\) and \(\omega ^{kY}\sin (\beta \omega )\) with respect to \(e^{-\omega ^{2}/2}\) on \(\mathbb {R}_{+}\), as well as the asymptotics for the Kummer’s function M(abz), as \(h\rightarrow 0\), we deduce that

$$\begin{aligned} a_{m,n,r}^{\pm }(h)&=\frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\cdot 2^{((m+n)Y+r-1)/2}\,\Gamma \bigg (\frac{(m+n)Y+r+1}{2}\bigg )\\&\quad \,\cdot M\bigg (\frac{(m+n)Y+r+1}{2},\frac{1}{2},-\frac{\big (\varepsilon \pm \widetilde{\gamma }h\big )^{2}}{2\sigma ^{2}h}\bigg )\\&\sim \frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\cdot 2^{((m+n)Y+r-1)/2}\,\Gamma \bigg (\frac{(m+n)Y+r+1}{2}\bigg )\\&\quad \,\cdot \left( \frac{\Gamma (1/2)}{\Gamma \big (\!-((m+n)Y+r)/2\big )}\bigg (\frac{\varepsilon ^{2}}{2\sigma ^{2}h}\bigg )^{-((m+n)Y+r+1)/2}\right. \\&\qquad \,\,\left. +\frac{\Gamma (1/2)\,e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}{\Gamma \big (((m+n)Y+r+1)/2\big )}\bigg (\frac{\varepsilon ^{2}}{2\sigma ^{2}h}\bigg )^{((m+n)Y+r)/2}\right) , \end{aligned}$$

and that

$$\begin{aligned} d_{m,n,r}^{\pm }(h)&=\frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\cdot 2^{((m+n)Y+r)/2}\,\Gamma \bigg (\frac{(m+n)Y+r}{2}+1\bigg )\\&\quad \,\cdot M\bigg (\frac{(m+n)Y+r}{2}+1,\frac{3}{2},-\frac{\big (\varepsilon \pm \widetilde{\gamma }h\big )^{2}}{2\sigma ^{2}h}\bigg )\\&\sim \frac{c_{1}^{m}\,c_{2}^{n}\,h^{(m+n)(1-Y/2)}}{m!\,n!\,\sigma ^{(m+n)Y}}\cdot \frac{\varepsilon }{\sigma \sqrt{h}}\cdot 2^{((m+n)Y+r)/2}\,\Gamma \bigg (\frac{(m+n)Y+r}{2}+1\bigg )\\&\quad \,\cdot \left( \frac{\Gamma (3/2)}{\Gamma \big ((1-(m+n)Y-r)/2\big )}\cdot \bigg (\frac{\varepsilon ^{2}}{2\sigma ^{2}h}\bigg )^{-((m+n)Y+r)/2-1}\right. \\&\qquad \quad \left. +\frac{\Gamma (3/2)\,e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}{\Gamma \big ((m+n)Y+r)/2+1\big )}\cdot \bigg (\frac{\varepsilon ^{2}}{2\sigma ^{2}h}\bigg )^{((m+n)Y+r-1)/2}\right) . \end{aligned}$$

In the asymptotic formulas for the Kummers function in the expression of \(a_{n,m,r}^{\pm }(h)\) above, the first term vanishes if \(\Gamma (-((m+n)Y+r)/2)\) are infinity. This happens when \(-((m+n)Y+r)/2\) is a nonpositive integer. Similarly, in the asymptotic formulas for the Kummers function in \(d_{m,n,r}^{\pm }(h)\), the first term vanishes if \((1-(m+n)Y-r)/2\) is a nonpositive integer. Hence, for \(m,n\in \mathbb {Z}_{+}\) and \(r\in \mathbb {R}_{+}\), as \(h\rightarrow 0\),

$$\begin{aligned} a_{m,n,r}^{\pm }(h)&=O\bigg (\frac{h^{m+n+(r+1)/2}}{\varepsilon ^{(m+n)Y+r+1}}\bigg )\!+O\bigg (\frac{\varepsilon ^{(m+n)Y+r}e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}{h^{(m+n)(Y-1)+r/2}}\bigg )\!=O\bigg (\frac{h^{m+n+(r+1)/2}}{\varepsilon ^{(m+n)Y+r+1}}\bigg ), \end{aligned}$$
(43)
$$\begin{aligned} d_{m,n,r}^{\pm }(h)&=O\bigg (\frac{h^{m+n+(r+1)/2}}{\varepsilon ^{(m+n)Y+r+1}}\bigg )\!+O\bigg (\frac{\varepsilon ^{(m+n)Y+r}e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}{h^{(m+n)(Y-1)+r/2}}\bigg )\!=O\bigg (\frac{h^{m+n+(r+1)/2}}{\varepsilon ^{(m+n)Y+r+1}}\bigg ). \end{aligned}$$
(44)

Therefore, by combining Eqs. (40), (43), and (44), we obtain that

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) =\frac{e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}{\sqrt{2\pi }}+O\big (h^{3/2}\varepsilon ^{-1-Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(45)

Next, we note that

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \mp J_{h}\,\phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) =\int _{\mathbb {R}}(\mathcal {F}\psi )(z)zp_{J,h}^{\mp }(z)\,dz=\int _{\mathbb {R}}\psi (u)\mathcal {F}\big (zp_{J,h}^{\mp }(z)\big )(u)\,du, \end{aligned}$$

where by Eq. (38),

$$\begin{aligned} \mathcal {F}\big (zp_{J,h}^{\pm }(z)\big )(u)=i\frac{d}{du}\big (\mathcal {F}p_{J,h}^{\pm }\big )(u)&=\frac{i}{\sqrt{2\pi }}\exp \Big (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)\mp iu\widetilde{\gamma }h\Big )\\&\quad \,\cdot \Big (c_{1}Y|u|^{Y-1}\,\text {sgn}(u)h+ic_{2}Y|u|^{Y-1}h\mp i\widetilde{\gamma }h\Big ). \end{aligned}$$

Together with Eq. (37), we obtain that

$$\begin{aligned}&\widetilde{\mathbb {E}}\left( \mp J_{h}\,\phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) \nonumber \\&\quad =\frac{ic_{1}\sigma Yh^{3/2}}{2\pi }\int _{\mathbb {R}}\,{\mathrm{sgn}}(u)|u|^{Y-1}\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du \nonumber \\&\qquad -\frac{c_{2}\sigma Yh^{3/2}}{2\pi }\int _{\mathbb {R}}\,|u|^{Y-1}\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du \nonumber \\&\qquad \mp \frac{\widetilde{\gamma }\sigma h^{3/2}}{2\pi }\int _{\mathbb {R}}\,\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du \end{aligned}$$
(46)

The asymptotics of the last term above follows from Eqs. (39) and (45), namely

$$\begin{aligned}&\frac{\widetilde{\gamma }\sigma h^{3/2}}{2\pi }\int _{\mathbb {R}}\,\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du \nonumber \\&\quad =O\Big (h\,e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )\!+\!O\big (h^{5/2}\varepsilon ^{-1-Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(47)

For the first term in Eq. (46), by expanding the Taylor series for \(\exp (c_{1}\sigma ^{-Y}|\omega |^{Y}h^{1-Y/2})\), as well as for \(\cos (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\) and \(\sin (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\) below, and using Eqs. (41) and (42), we have

$$\begin{aligned}&\frac{ic_{1}\sigma Yh^{3/2}}{2\pi }\int _{\mathbb {R}}\,{\mathrm{sgn}}(u)|u|^{Y-1}\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du \nonumber \\&\quad = -\frac{c_{1}\sigma Yh^{3/2}}{\pi }\int _{0}^{\infty }u^{Y-1}\exp \bigg (c_{1}u^{Y}h-\frac{\sigma ^{2}u^{2}h}{2}\bigg )\sin \Big (c_{2}u^{Y}h+u\big (\varepsilon \pm \widetilde{\gamma }h\big )\Big )du \nonumber \\&\quad = -\frac{c_{1}Yh^{(3-Y)/2}}{\sigma ^{Y-1}\pi }\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\sin \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}+\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\quad = -\frac{c_{1}Yh^{(3-Y)/2}}{\sigma ^{Y-1}\pi }\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\sin \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\cos \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\qquad -\frac{c_{1}Yh^{(3-Y)/2}}{\sigma ^{Y-1}\pi }\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\cos \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\sin \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\quad = -\frac{c_{1}Yh^{(3-Y)/2}}{\sigma ^{Y-1}\pi }\left( \sum _{k,j=0}^{\infty }(-1)^{j}a_{k,2j+1,Y-1}^{\pm }(h)+\sum _{k,j=0}^{\infty }(-1)^{j}d_{k,2j,Y-1}^{\pm }(h)\right) =O\big (h^{3/2}\varepsilon ^{-Y}\big ), \end{aligned}$$
(48)

as \(h\rightarrow 0\), where we have used the asymptotic formulas (43) and (44) in the last equality. Finally, for the second term in Eq. (46), again by expanding the Taylor series for \(\exp (c_{1}\sigma ^{-Y}|\omega |^{Y}h^{1-Y/2})\), \(\cos (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\), and \(\sin (c_{2}\sigma ^{-Y}\omega ^{Y}h^{1-Y/2})\) below, and using Eqs. (41), (42), (43), and (44), we deduce that

$$\begin{aligned}&\frac{c_{2}\sigma Yh^{3/2}}{2\pi }\int _{\mathbb {R}}\,|u|^{Y-1}\exp \bigg (c_{1}|u|^{Y}h+ic_{2}|u|^{Y}h\,\text {sgn}(u)-\frac{\sigma ^{2}u^{2}h}{2}+iu\big (\varepsilon \pm \widetilde{\gamma }h\big )\bigg )du\nonumber \\&\quad =\frac{c_{2}\sigma Yh^{3/2}}{\pi }\int _{0}^{\infty }u^{Y-1}\exp \bigg (c_{1}u^{Y}h-\frac{\sigma ^{2}u^{2}h}{2}\bigg )\cos \Big (c_{2}u^{Y}h+u\big (\varepsilon \pm \widetilde{\gamma }h\big )\Big )du \nonumber \\&\quad =\frac{c_{2}Yh^{(3-Y)/2}}{\pi \sigma ^{Y-1}}\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\cos \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}+\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\quad =\frac{c_{2}Yh^{(3-Y)/2}}{\pi \sigma ^{Y-1}}\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\cos \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\cos \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\qquad -\frac{c_{2}Yh^{(3-Y)/2}}{\pi \sigma ^{Y-1}}\int _{0}^{\infty }\omega ^{Y-1}\exp \bigg (\frac{c_{1}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}-\frac{\omega ^{2}}{2}\bigg )\sin \bigg (\frac{c_{2}\,\omega ^{Y}h^{1-Y/2}}{\sigma ^{Y}}\bigg )\sin \bigg (\omega \cdot \frac{\varepsilon \pm \widetilde{\gamma }h}{\sigma \sqrt{h}}\bigg )d\omega \nonumber \\&\quad =\frac{c_{2}Yh^{(3-Y)/2}}{\pi \sigma ^{Y-1}}\left( \sum _{k,j=0}^{\infty }(-1)^{j}a_{k,2j,Y-1}^{\pm }(h)-\sum _{k,j=0}^{\infty }(-1)^{j}d_{k,2j+1,Y-1}^{\pm }(h)\right) =O\big (h^{3/2}\varepsilon ^{-Y}\big ), \end{aligned}$$
(49)

as \(h\rightarrow 0\). Therefore, by combining Eqs. (46), (47), (48), and (49), we obtain that

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \mp J_{h}\phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) =O\Big (he^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h^{3/2}\varepsilon ^{-Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(50)

It remains to analyze the asymptotic behavior of \(\mathbb {E}(\overline{\Phi }((\varepsilon \pm J_{h})/(\sigma \sqrt{h})))\). We first note that there exists a universal constant \(K>0\), such that

$$\begin{aligned}&\widetilde{\mathbb {E}}\left( \overline{\Phi }\bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\mathbf{1}_{\{\varepsilon \pm J_{h}\ge 0\}}\right) \le K\,\widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\mathbf{1}_{\{\varepsilon \pm J_{h}\ge 0\}}\right) \nonumber \\&\quad =O\left( \widetilde{\mathbb {E}}\left( \phi \bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) \right) =O\Big (e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h^{3/2}\varepsilon ^{-1-Y}\big ),\quad \text {as }\,h\rightarrow 0, \end{aligned}$$
(51)

where the last inequality above follows from Eq. (45). Moreover, by Eq. (11), as \(h\rightarrow 0\),

$$\begin{aligned}&\widetilde{\mathbb {E}} \ \left( \overline{\Phi }\bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\mathbf{1}_{\{\varepsilon \pm J_{h}\le 0\}}\!\right) =\int _{\mathbb {R}}\phi (u)\,\widetilde{\mathbb {P}}\Big (\varepsilon \pm J_{h}\le \sigma \sqrt{h}u,\,\varepsilon \pm J_{h}\le 0\Big )du \nonumber \\&\quad =\int _{0}^{\infty }\phi (u)\,\widetilde{\mathbb {P}}\Big (\pm Z_{h}\le -\varepsilon \mp \widetilde{\gamma }h\Big )du+\int _{-\infty }^{0}\phi (u)\,\widetilde{\mathbb {P}}\Big (\pm Z_{h}\le \sigma \sqrt{h}u-\varepsilon \mp \widetilde{\gamma }h\Big )du \nonumber \\&\quad =\frac{1}{2}\,\widetilde{\mathbb {P}}\bigg (\!\pm Z_{1}\le \frac{-\varepsilon \mp \widetilde{\gamma }h}{h^{1/Y}}\bigg )+\int _{-\infty }^{0}\phi (u)\,\widetilde{\mathbb {P}}\bigg (\pm Z_{1}\le \frac{\sigma \sqrt{h}u-\varepsilon \mp \widetilde{\gamma }h}{h^{1/Y}}\bigg )du \nonumber \\&\quad \le \frac{1}{2}\,\widetilde{\mathbb {P}}\bigg (\!\pm Z_{1}\le \frac{-\varepsilon \mp \widetilde{\gamma }h}{h^{1/Y}}\bigg )+\frac{\widetilde{K}h}{\varepsilon ^{Y}}\int _{-\infty }^{0}\phi (u)\bigg (1-\frac{\sigma \sqrt{h}\pm \widetilde{\gamma }h}{\varepsilon }u\bigg )^{-Y}du=O\big (h\varepsilon ^{-Y}\big ). \end{aligned}$$
(52)

Therefore, by combining Eqs. (51) and (52), we obtain that

$$\begin{aligned} \widetilde{\mathbb {E}}\left( \overline{\Phi }\bigg (\frac{\varepsilon }{\sigma \sqrt{h}}\pm \frac{J_{h}}{\sigma \sqrt{h}}\bigg )\right) =O\Big (e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h\varepsilon ^{-Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(53)

Finally, by combining Eqs. (35), (36), (45), (50), and (53), we conclude that

$$\begin{aligned} I_{1}(h)={\frac{\sqrt{2}}{\sigma \sqrt{\pi }}\cdot \frac{\varepsilon }{\sqrt{h}}\,e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}}+O\Big (e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h\varepsilon ^{-Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(54)

Step 1.2

We now study the asymptotic behavior of \(I_{2}(h)\), defined in Eq. (34), as \(h\rightarrow 0\). Let us first consider the following decomposition

$$\begin{aligned} I_{2}(h)&=\widetilde{\mathbb {E}}\Big (\Big (e^{-\widetilde{U}_{h}}-1+\widetilde{U}_{h}\Big )W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big )- \widetilde{\mathbb {E}}\Big (\widetilde{U}_{h}W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big ) \nonumber \\&=:I_{21}(h)-I_{22}(h). \end{aligned}$$
(55)

The first term \(I_{21}(h)\) can be bounded as follows: as \(h\rightarrow 0\),

$$\begin{aligned} 0\le I_{21}(h)\le \widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}\!-1+\widetilde{U}_{h}\Big )=\exp \bigg (h\int _{\mathbb {R}_{0}}\!\Big (e^{-\phi (x)}\!-1+\phi (x)\Big )\widetilde{\nu }(dx)\bigg )-1=O(h). \end{aligned}$$
(56)

To deal with \(I_{22}\), for any \(t\in \mathbb {R}_{+}\), we further decompose \(\widetilde{U}_{t}\) as

$$\begin{aligned} \widetilde{U}_{t}=\int _{0}^{t}\int _{\mathbb {R}_{0}}\big (\varphi (x)+\alpha _{\text {sgn}(x)}x\big )\widetilde{N}(ds,dx)-\int _{0}^{t}\int _{\mathbb {R}_{0}}\alpha _{\text {sgn}(x)}x\widetilde{N}(ds,dx)=:\widetilde{U}^{\text {BV}}_{t}-\alpha _{+}{Z}_{t}^{+}-\alpha _{-}{Z}^{-}_{t}, \end{aligned}$$

where the first integral is well-defined in light of Assumption 1-(i) & (ii), so that

$$\begin{aligned} I_{22}(h)&=\widetilde{\mathbb {E}}\Big (\widetilde{U}^{\text {BV}}_{h}W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big )-\alpha _{+}\widetilde{\mathbb {E}}\Big (Z^{+}_{h}W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big ) \nonumber \\&\quad -\alpha _{-}\widetilde{\mathbb {E}}\Big (Z^{-}_{h}W_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|>\varepsilon \}}\Big )=:I_{22}^{\text {BV}}(h)-\alpha _{+}I_{22}^{+}(h)-\alpha _{-}I_{22}^{-}(h). \end{aligned}$$

For the first term \(I_{22}^{\text {BV}}(h)\), note that

$$\begin{aligned} \big |I_{22}^{\text {BV}}(h)\big |\le \widetilde{\mathbb {E}}\Big (\big |\widetilde{U}^{\text {BV}}_{h}\big |\Big )\le 2h\int _{\mathbb {R}_{0}}\big |\varphi (x)+\alpha _{\text {sgn}(x)}x\big |\,\widetilde{\nu }(dx), \end{aligned}$$

where the last integral is finite since in a neighborhood of the origin,

$$\begin{aligned} \big |\varphi (x)+\alpha _{\text {sgn}(x)}x\big |=\big |-\ln q(x)+\alpha _{\text {sgn}(x)}x\big |=O\Big (\big |1-q(x)+\alpha _{\text {sgn}(x)}x\big |\Big ), \end{aligned}$$

which is integrable with respect to \(\widetilde{\nu }(dx)\) in view of Assumption 1-(ii). As for the terms \(I_{22}^{\pm }\), due to the self-similarity of \(Z_{t}^{\pm }\) and the fact that \(\varepsilon h^{-1/Y}\rightarrow \infty\) (since \(Y\in (1,2)\)), the monotone convergence theorem implies that \(I_{22}^{\pm }(h)=o(h^{1/Y})\), as \(h\rightarrow 0\). Hence, we obtain that

$$\begin{aligned} I_{22}(h)=o\big (h^{1/Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(57)

By combining Eqs. (55), (56), and (57), we conclude that

$$\begin{aligned} I_{2}(h)=o\big (h^{1/Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(58)

Finally, from Eqs. (34), (54), and (58), we obtain that

$$\begin{aligned} \mathbb {E}\Big (\!W_{h}^{2}{} \mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\!\Big )\!=\!h\!-\!\frac{\sqrt{2h}\,\varepsilon }{\sigma \sqrt{\pi }}e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\!+\!O\Big (\!he^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\!\Big )\!+\!O\big (h^{2}\varepsilon ^{-Y}\big )\!+{\!o\big (h^{1+1/Y}\big )},\quad \end{aligned}$$
(59)

as \(h\rightarrow 0\), which completes the analysis in Step 1.

Step 2

In this step, we will study the asymptotic behavior of the second term in Eq. (33), as \(h\rightarrow 0\). By Eqs. (7), (8), and (9), we first have

$$\begin{aligned} \mathbb {E}\Big (J_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )&=\widetilde{\mathbb {E}} \ \Big (e^{-\widetilde{U}_{h}-\eta h}\,J_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big ) \nonumber \\&=e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}Z_{h}^{2}{} \mathbf{1}_{\{|W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )+2\widetilde{\gamma }he^{-\eta h}\,\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}Z_{h}{} \mathbf{1}_{\{|W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big ) \nonumber \\&\quad +\widetilde{\gamma }^{2}h^{2}e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}\,\mathbf{1}_{\{|W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big ) \nonumber \\&=:e^{-\eta h}I_{3}(h)+2\widetilde{\gamma }he^{-\eta h}I_{4}(h)+\widetilde{\gamma }^{2}h^{2}e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}\,\mathbf{1}_{\{|W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big ). \end{aligned}$$
(60)

Clearly,

$$\begin{aligned} \widetilde{\gamma }^{2}h^{2}e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}}\,\mathbf{1}_{\{|W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )=O\big (h^{2}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(61)

It remains to analyze the asymptotic behavior of the first two terms in Eq. (60).

Step 2.1

We begin with the analysis of \(I_{3}(h)\). Clearly,

$$\begin{aligned} I_{3}(h)&=\widetilde{\mathbb {E}}\Big (Z_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )+\widetilde{\mathbb {E}}\Big (\Big (e^{-\widetilde{U}_{h}}-1\Big )Z_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big ) \nonumber \\&=h^{2/Y}\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )+\widetilde{\mathbb {E}}\Big (\Big (e^{-\widetilde{U}_{h}}-1\Big )Z_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\Big ) \nonumber \\&=:h^{2/Y}I_{31}(h)+I_{32}(h). \end{aligned}$$
(62)

By the symmetry of \(W_{1}\), we note that

$$\begin{aligned} \widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{-\varepsilon \le \sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h\le 0\}}\Big )=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}-h^{1/Y}\!Z_{1}-\widetilde{\gamma }h\le \varepsilon \}}\Big ). \end{aligned}$$

In what follows, we let \(h>0\) small enough so that \(\varepsilon -|\widetilde{\gamma }|h>0\).

To study the asymptotic behavior of \(I_{31}(h)\), as \(h\rightarrow 0\), let us first consider

$$\begin{aligned} E_{1}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\ge 0,\,\pm Z_{1}\ge 0\}}\Big )\!=\!\!\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}\!\!\bigg (\!\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!\!u^{2}p_{Z}(\pm u)du\!\bigg )\phi (x)dx \nonumber \\&\,\,=\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{2}p_{Z}(\pm u)\,du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \nonumber \\&\,\,=\frac{C_{\pm }\big (\varepsilon \mp \widetilde{\gamma }h\big )}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{1-Y}du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \nonumber \\&\,\,\quad +\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{2}\Big (p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\Big )du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega . \end{aligned}$$
(63)

For the first term in Eq. (63), we have

$$\begin{aligned} \int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}\!u^{1-Y}du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega&=\frac{\big (\varepsilon \mp \widetilde{\gamma }h\big )^{2-Y}}{(2-Y)h^{(2-Y)/Y}}\int _{0}^{1}(1-\omega )^{2-Y}\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \nonumber \\&\sim \frac{\big (\varepsilon \mp \widetilde{\gamma }h\big )^{2-Y}}{2(2-Y)h^{(2-Y)/Y}}\cdot \frac{\sigma \sqrt{h}}{\varepsilon \mp \widetilde{\gamma }h},\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(64)

For the second term in Eq. (63), since \(Y\in (1,2)\), we first observe that, for any \(z>0\),

$$\begin{aligned} \int _{0}^{z}u^{2}\big (u^{-Y-1}\!\wedge u^{-2Y-1}\big )du=\frac{z^{2-Y}}{2\!-\!Y}{} \mathbf{1}_{(0,1]}(z)+\frac{1\!-\!z^{2-2Y}}{2(Y\!-\!1)}{} \mathbf{1}_{(1,\infty )}(z)\le \frac{Y}{2(Y\!-\!1)(2\!-\!Y)}. \end{aligned}$$
(65)

Hence, we deduce from Eq. (13) that

$$\begin{aligned}&\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{2}\big |p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\big |du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \\&\quad \le \frac{\widetilde{K}Y}{2(Y-1)(2-Y)}\int _{0}^{1}\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega =O\big (\sqrt{h}\,\varepsilon ^{-1}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$

Therefore, we obtain that

$$\begin{aligned} E_{1}^{\pm }(h)=\frac{C_{\pm }}{2(2-Y)}\,h^{1-2/Y}\varepsilon ^{2-Y}+O(1),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(66)

Using the same argument as above and since \(\varepsilon \gg h\), we also obtain that, when \(\pm \widetilde{\gamma }>0\), as \(h\rightarrow 0\),

$$\begin{aligned} E_{2}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\le 0,\,\pm Z_{1}\le 0\}}\Big ) \nonumber \\&\,\,=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\le \mp \widetilde{\gamma }h,\,W_{1}\le 0,\,\pm Z_{1}\le 0\}}\Big )=O\big (h^{3-Y-2/Y}\big )+O(1). \end{aligned}$$
(67)

Next, we consider

$$\begin{aligned} E_{3}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\ge 0,\,\pm Z_{1}\le 0\}}\Big ) \nonumber \\&\,\,=\!\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}\!\!\bigg (\!\int _{-\frac{\sigma \sqrt{h}x}{h^{1/Y}}}^{0}\!\!u^{2}p_{Z}(\pm u)du\!\bigg )\phi (x)dx\!+\!\!\int _{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}^{\infty }\!\!\bigg (\!\int _{-\frac{\sigma \sqrt{h}x}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!\!\!u^{2}p_{Z}(\pm u)du\!\bigg )\phi (x)dx.\quad \end{aligned}$$
(68)

By Eq. (12), the first term in Eq. (68) is such that

$$\begin{aligned}&\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}\bigg (\int _{-\frac{\sigma \sqrt{h}x}{h^{1/Y}}}^{0}u^{2}p_{Z}(\pm u)\,du\bigg )\phi (x)\,dx\le \widetilde{K}\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}\bigg (\int _{0}^{\frac{\sigma \sqrt{h}x}{h^{1/Y}}}u^{1-Y}du\bigg )\phi (x)\,dx\\&\quad =\frac{\widetilde{K}\sigma ^{2-Y}h^{2-Y/2-2/Y}}{2-Y}\int _{0}^{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}x^{2-Y}\phi (x)\,dx=O\big (h^{2-Y/2-2/Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$

Similarly, the second term in Eq. (68) can be estimated as follows:

$$\begin{aligned}&\int _{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}^{\infty }\bigg (\int _{-\frac{\sigma \sqrt{h}x}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{2}p_{Z}(\pm u)\,du\bigg )\phi (x)\,dx\le \widetilde{K}\int _{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}^{\infty }\bigg (\int _{\frac{\sigma \sqrt{h}x-(\varepsilon \mp \widetilde{\gamma }h)}{h^{1/Y}}}^{\frac{\sigma \sqrt{h}x}{h^{1/Y}}}u^{1-Y}du\bigg )\phi (x)\,dx\\&\quad \le \frac{\widetilde{K}\sigma ^{2-Y}h^{2-2/Y-Y/2}}{2-Y}\int _{\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}}^{\infty }x^{2-Y}\phi (x)\,dx=o\big (h^{2-2/Y-Y/2}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$

Therefore, we obtain that

$$\begin{aligned} E_{3}^{\pm }(h)=O\big (h^{2-Y/2-2/Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(69)

To complete the analysis for \(I_{3}(h)\), it remains to study

$$\begin{aligned} E_{4}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\le 0,\,\pm Z_{1}\ge 0\}}\Big )=\!\int _{-\infty }^{0}\!\bigg (\!\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!u^{2}p_{Z}(\pm u)du\!\bigg )\phi (x)dx \nonumber \\&\,\,=C_{\pm }\int _{-\infty }^{0}\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{1-Y}du\bigg )\phi (x)\,dx \nonumber \\&\quad \,\,+\int _{-\infty }^{0}\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{2}\Big (p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\Big )du\bigg )\phi (x)\,dx. \end{aligned}$$
(70)

For the first term in Eq. (70), we have

$$\begin{aligned}&C_{\pm }\int _{-\infty }^{0}\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{1-Y}du\bigg ) \ \phi (x)\,dx \nonumber \\&\quad =\frac{C_{\pm }\,\varepsilon ^{2-Y}}{(2-Y)h^{2/Y-1}}\int _{-\infty }^{0}\left( \bigg (1-\frac{\sigma \sqrt{h}x\pm \widetilde{\gamma }h}{\varepsilon }\bigg )^{2-Y}\!-\bigg (\!-\frac{\sigma \sqrt{h}x\pm \widetilde{\gamma }h}{\varepsilon }\bigg )^{2-Y}\right) \ \phi (x)\,dx \nonumber \\&\quad \sim \frac{C_{\pm }\,h^{1-2/Y}\varepsilon ^{2-Y}}{2(2-Y)},\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(71)

For the second term in Eq. (70), we deduce from Eq. (65) that

$$\begin{aligned} \int _{-\infty }^{0}\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{2}\Big |p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\Big |du\bigg )\phi (x)\,dx=O(1),\quad h\rightarrow 0. \end{aligned}$$

Therefore, we obtain that

$$\begin{aligned} E_{4}^{\pm }(h)=\frac{C_{\pm }}{2(2-Y)}h^{1-2/Y}\varepsilon ^{2-Y}+O(1),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(72)

By combining Eqs. (66), (67), (69), and (72), we conclude that

$$\begin{aligned} I_{31}(h)=\sum _{i=1}^{4}\big (E_{i}^{+}(h)\!+\!E_{i}^{-}(h)\big )=\frac{C_{+}\!+\!C_{-}}{2-Y}h^{1-2/Y}\varepsilon ^{2-Y}+O\big (h^{2-Y/2-2/Y}\big ),\quad \text {as }\,h\rightarrow 0.\quad \end{aligned}$$
(73)

Next, we will study the asymptotic behavior of \(I_{32}(h)\), as \(h\rightarrow 0\). Clearly, by Cauchy-Schwarz inequality and self-similarity of \(Z_{h}\) and \(W_{h}\) under \(\widetilde{\mathbb {P}}\), we have that

$$\begin{aligned} I_{32}(h)\le h^{2/Y}\left( \widetilde{\mathbb {E}}\bigg (\!\Big (e^{-\widetilde{U}_{h}}-1\Big )^{2}\bigg )\right) ^{1/2}\bigg (\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )\bigg )^{1/2}. \end{aligned}$$
(74)

By Assumption 1-(v) and denoting \(\widetilde{C}_{\ell }=\int _{\mathbb {R}_{0}}\big (e^{-\ell \varphi (x)}-1+\ell \varphi (x)\big )\tilde{\nu }(dx)\), \(\ell =1,2\), we first have

$$\begin{aligned} \widetilde{\mathbb {E}}\bigg (\Big (e^{-\widetilde{U}_{h}}-1\Big )^{2}\bigg )=e^{\widetilde{C}_{2}h}-2e^{\widetilde{C}_{1}h}+1\sim \big (\widetilde{C}_{2}-2\widetilde{C}_{1}\big )h,\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(75)

The analysis of the asymptotic behavior, as \(h\rightarrow 0\), of the second factor in Eq. (74) is similar to that of \(I_{31}(h)\). More precisely, we first consider

$$\begin{aligned} F_{1}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\ge 0,\,\pm Z_{1}\ge 0\}}\Big )\\&\,\,=\frac{C_{\pm }\big (\varepsilon \mp \widetilde{\gamma }h\big )}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{3-Y}du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \\&\,\,\quad +\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{4}\Big (p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\Big )du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega . \end{aligned}$$

A similar argument as in Eq. (64) shows that

$$\begin{aligned} \frac{C_{\pm }\big (\varepsilon \mp \widetilde{\gamma }h\big )}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{3-Y}du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \sim \frac{C_{\pm }}{2(4-Y)}h^{1-4/Y}\varepsilon ^{4-Y},\quad \text {as }\,h\rightarrow 0, \end{aligned}$$

and by Eq. (13),

$$\begin{aligned}&\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\int _{0}^{1}\bigg (\int _{0}^{\frac{(\varepsilon \mp \widetilde{\gamma }h)(1-\omega )}{h^{1/Y}}}u^{4}\big |p_{Z}(u)-Cu^{-1-Y}\big |du\bigg )\phi \bigg (\frac{\varepsilon \mp \widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega \\&\quad \le \frac{\varepsilon \!\mp \!\widetilde{\gamma }h}{\sigma \sqrt{h}}\cdot \frac{\widetilde{K}\big (\varepsilon \mp \widetilde{\gamma }h\big )^{4-2Y}}{2(2-Y)h^{(4-2Y)/Y}}\!\int _{0}^{1}(1-\omega )^{4-2Y}\phi \bigg (\frac{\varepsilon \!\mp \!\widetilde{\gamma }h}{\sigma \sqrt{h}}\omega \bigg )d\omega =O\big (h^{2-4/Y}\varepsilon ^{4-2Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$

Hence, we obtain that

$$\begin{aligned} F_{1}^{\pm }(h)=\frac{C_{\pm }}{2(4-Y)}h^{1-4/Y}\varepsilon ^{4-Y}+O\big (h^{2-4/Y}\varepsilon ^{{4}-2Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(76)

Using the same argument as above and since \(\varepsilon \gg h\), we also obtain that, when \(\pm \widetilde{\gamma }>0\), as \(h\rightarrow 0\),

$$\begin{aligned} F_{2}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\le 0,\,\pm Z_{1}\le 0\}}\Big ) \nonumber \\&\,\,=\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\le \mp \widetilde{\gamma }h,\,W_{1}\ge 0,\,\pm Z_{1}\le 0\}}\Big )=O\big (h^{6-4/Y-2Y}\big ). \end{aligned}$$
(77)

Moreover, using arguments similar to those for \(E_{3}^{\pm }(h)\), we deduce that

$$\begin{aligned} F_{3}^{\pm }(h):=\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\ge 0,\,\pm Z_{1}\le 0\}}\Big )=O\big (h^{3-4/Y-Y/2}\big ),\quad \text {as }\,h\rightarrow 0.\quad \end{aligned}$$
(78)

Finally, we consider

$$\begin{aligned} F_{4}^{\pm }(h)&:=\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{0\le \sigma \sqrt{h}W_{1}\pm h^{1/Y}Z_{1}\pm \widetilde{\gamma }h\le \varepsilon ,\,W_{1}\le 0,\,\pm Z_{1}\ge 0\}}\Big )=\!\int _{-\infty }^{0}\!\bigg (\!\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!u^{4}p_{Z}(\pm u)du\!\bigg )\phi (x)dx\\&\,\,=C_{\pm }\!\int _{-\infty }^{0}\!\!\!\bigg (\!\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!\!u^{3-Y}\!du\!\bigg )\phi (x)dx\!+\!\!\int _{-\infty }^{0} \left( \!\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!\!\!u^{4}\bigg (\!p_{Z}(\pm u)\!-\!\frac{C_{\pm }}{u^{1+Y}}\!\bigg )du \right) \ \phi (x)dx. \end{aligned}$$

A similar argument as in Eq. (71) shows that

$$\begin{aligned} C_{\pm }\int _{-\infty }^{0}\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}u^{3-Y}du\bigg )\phi (x)\,dx\sim \frac{C_{\pm }}{2(4-Y)}h^{1-4/Y}\varepsilon ^{4-Y},\quad \text {as }\,h\rightarrow 0, \end{aligned}$$

and by Eq. (13),

$$\begin{aligned}&\int _{-\infty }^{0}\!\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!u^{4}\Big |p_{Z}(\pm u)-C_{\pm }u^{-1-Y}\Big |du\bigg )\phi (x)\,dx\le \widetilde{K}\int _{-\infty }^{0}\!\bigg (\int _{\frac{-\sigma \sqrt{h}x\mp \widetilde{\gamma }h}{h^{1/Y}}}^{\frac{\varepsilon \mp \widetilde{\gamma }h-\sigma \sqrt{h}x}{h^{1/Y}}}\!u^{3-2Y}du\bigg )\phi (x)\,dx\\&\quad =\frac{\widetilde{K}\varepsilon ^{4-2Y}}{2(2\!-\!Y)h^{4/Y-2}}\int _{-\infty }^{0}\!\left( \!\bigg (\!1\!-\!\frac{\sigma \sqrt{h}x\!\pm \!\widetilde{\gamma }h}{\varepsilon }\bigg )^{4-2Y}\!\!\!-\!\bigg (\!\!-\!\frac{\sigma \sqrt{h}x\!\pm \!\widetilde{\gamma }h}{\varepsilon }\bigg )^{4-2Y}\right) \ \phi (x)\,dx=O\bigg (\frac{\varepsilon ^{4-2Y}}{h^{4/Y-2}}\bigg ). \end{aligned}$$

Hence, we obtain that

$$\begin{aligned} F_{4}^{\pm }(h)=\frac{C_{\pm }}{2(4-Y)}h^{1-4/Y}\varepsilon ^{4-Y}+O\big (h^{2-4/Y}\varepsilon ^{4-2Y}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(79)

Combining Eqs. (76), (77), (78), and (79), leads to

$$\begin{aligned}&\widetilde{\mathbb {E}}\Big (Z_{1}^{4}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )=\sum _{i=1}^{4}\big (F_{i}^{+}(h)+F_{i}^{-}(h)\big ) \nonumber \\&\quad =\frac{{\big (C_{+}+C_{-}\big )}h^{1-4/Y}\varepsilon ^{4-Y}}{4-Y}+O\big (h^{2-4/Y}\varepsilon ^{4-2Y}\big )+O\big (h^{3-4/Y-Y/2}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(80)

Therefore, by combining Eqs. (74), (75), and (80), we have

$$\begin{aligned} I_{32}(h)=O\big (h\,\varepsilon ^{2-Y/2}\big )+O\big (h^{3/2}\varepsilon ^{2-Y}\big )+O\big (h^{2-Y/4}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(81)

Finally, by combining Eqs. (62), (73), and (81), we obtain that

$$\begin{aligned} I_{3}(h)=\frac{C_{+}+C_{-}}{2-Y}\,h\varepsilon ^{2-Y}+O\big (h\varepsilon ^{2-Y/2}\big )+O\big (h^{2-Y/2}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(82)

Step 2.2

In this step, we will investigate the asymptotic behavior of \(I_{4}(h)\), as \(h\rightarrow 0\). Note that

$$\begin{aligned} I_{4}(h)&=h^{1/Y}\widetilde{\mathbb {E}}\Big (Z_{1}{} \mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )+\widetilde{\mathbb {E}}\bigg (\Big (e^{-\widetilde{U}_{h}}\!-1\Big )Z_{h}{} \mathbf{1}_{\{|\sigma W_{h}+Z_{h}+\widetilde{\gamma }h|\le \varepsilon \}}\bigg )\\&\,\,\le h^{1/Y}\bigg (\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )\bigg )^{1/2}\left( 1+\left( \widetilde{\mathbb {E}}\bigg (\Big (e^{-\widetilde{U}_{h}}-1\Big )^{2}\bigg )\right) ^{1/2}\right) \\&\,\,=O\left( h^{1/Y}\bigg (\widetilde{\mathbb {E}}\Big (Z_{1}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+h^{1/Y}\!Z_{1}+\widetilde{\gamma }h|\le \varepsilon \}}\Big )\bigg )^{1/2}\right) ,\quad h\rightarrow 0, \end{aligned}$$

where the second inequality above follows from Cauchy-Schwarz inequality. Therefore, by Eqs. (62) and (73), we obtain that

$$\begin{aligned} I_{4}(h)=O\big (\sqrt{h}\,\varepsilon ^{1-Y/2}\big )+O\big (h^{1-Y/4}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(83)

Finally, by combining Eqs. (60), (61), (82), and (83), we conclude that

$$\begin{aligned} \mathbb {E}\Big (J_{h}^{2}\,\mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )=\frac{C_{+}+C_{-}}{2-Y}\,h\varepsilon ^{2-Y}+O\big (h\varepsilon ^{2-Y/2}\big )+O\big (h^{2-Y/2}\big ),\quad \text {as }\,h\rightarrow 0, \end{aligned}$$
(84)

which completes the analysis of Step 2.

Step 3

In this last step, we will study the asymptotic behavior of the third term in Eq. (33), as \(h\rightarrow 0\). By Eqs. (7) and (8), we first decompose it as

$$\begin{aligned}&\mathbb {E}\Big (W_{h}J_{h}{} \mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )=\widetilde{\mathbb {E}}\Big (e^{-\widetilde{U}_{h}-\eta h}\,W_{h}J_{h}{} \mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big ) \nonumber \\&\quad =\sqrt{h}\,e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (W_{1}J_{h}{} \mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|\le \varepsilon \}}\Big )+\sqrt{h}\,e^{-\eta h}\,\widetilde{\mathbb {E}}\Big (\Big (e^{-\widetilde{U}_{h}}-1\Big )W_{1}J_{h}{} \mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|\le \varepsilon \}}\Big )\nonumber \\&\quad =:e^{-\eta h}\sqrt{h}\,I_{5}(h)+e^{-\eta h}\sqrt{h}\,I_{6}(h). \end{aligned}$$
(85)

For \(I_{5}(h)\), by conditioning on \(J_{h}\), and using the fact that, for any \(x_{1},x_{2}\in \mathbb {R}\) with \(x_{1}<x_{2}\),

$$\begin{aligned} \widetilde{\mathbb {E}}\Big (W_{1}{} \mathbf{1}_{\{W_{1}\in [x_{1},x_{2}]\}}\Big )=\phi (x_{1})-\phi (x_{2}), \end{aligned}$$

we obtain from Eq. (50) that, as \(h\rightarrow 0\),

$$\begin{aligned} I_{5}(h)=\widetilde{\mathbb {E}}\left( J_{h}\bigg (\phi \bigg (\frac{\varepsilon +J_{h}}{\sigma \sqrt{h}}\bigg )-\phi \bigg (\frac{\varepsilon -J_{h}}{\sigma \sqrt{h}}\bigg )\bigg )\right) =O\Big (h\,e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h^{3/2}\varepsilon ^{-Y}\big ). \end{aligned}$$
(86)

As for \(I_{6}(h)\), by Cauchy-Schwarz inequality, Eq. (9), (62), (73), and (75), we obtain that

$$\begin{aligned} \big |I_{6}(h)\big |&\le \bigg (\widetilde{\mathbb {E}}\bigg (\Big (e^{-\widetilde{U}_{h}}-1\Big )^{2}\bigg )\bigg )^{1/2}\bigg (\widetilde{\mathbb {E}}\Big (J_{h}^{2}\,\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|\le \varepsilon \}}\Big )\bigg )^{1/2} \nonumber \\&\le \bigg (\widetilde{\mathbb {E}}\bigg (\Big (e^{-\widetilde{U}_{h}}-1\Big )^{2}\bigg )\bigg )^{1/2}\bigg (\widetilde{\mathbb {E}}\Big (2\big (Z_{h}^{2}+\widetilde{\gamma }^{2}h^{2}\big )\mathbf{1}_{\{|\sigma \sqrt{h}W_{1}+J_{h}|\le \varepsilon \}}\Big )\bigg )^{1/2} \nonumber \\&=O\big (h\varepsilon ^{1-Y/2}\big ),\quad \text {as }\,h\rightarrow 0. \end{aligned}$$
(87)

Therefore, by combining Eqs. (85), (86), and (87), we obtain that, as \(h\rightarrow 0\),

$$\begin{aligned} \mathbb {E}\Big (W_{h}J_{h}{} \mathbf{1}_{\{|\sigma W_{h}+J_{h}|\le \varepsilon \}}\Big )&=O\Big (h^{3/2}e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )+O\big (h^{2}\varepsilon ^{-Y}\big )+O\big (h^{3/2}\varepsilon ^{1-Y/2}\big ), \end{aligned}$$
(88)

which completes the analysis in Step 3.

Finally, by combining Eqs. (33), (59), (84), and (88), we conclude that, as \(h\rightarrow 0\),

$$\begin{aligned} \mathbb {E}\big (b_{1}(\varepsilon )\big )\!=\!\sigma ^{2}h\!-\!{\frac{\sigma \varepsilon \sqrt{2h}}{\sqrt{\pi }}e^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\!+\!\frac{C_{+}\!\!+\!C_{-}}{2-Y}h\varepsilon ^{2-Y}}\!\!+\!O\Big (he^{-\varepsilon ^{2}/(2\sigma ^{2}h)}\Big )\!+\!O\big (h\varepsilon ^{2-Y/2}\big )\!+\!O\big (h^{2-Y/2}\big ), \end{aligned}$$
(89)

which completes the proof of the theorem.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Figueroa-López, J.E., Gong, R. & Han, Y. Estimation of Tempered Stable Lévy Models of Infinite Variation. Methodol Comput Appl Probab 24, 713–747 (2022). https://doi.org/10.1007/s11009-022-09940-7

Download citation

  • Received:

  • Revised:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11009-022-09940-7

Keywords

MSC 2000 Subject Classifications

Navigation