Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
Skip to main content

A generalized flexible functional form for α-returns to scale

  • Published:
Journal of Productivity Analysis Aims and scope Submit manuscript

Abstract

This study proposes a new flexible functional form for distance functions. This generalizes the existing functional form of Diewert (1992b) by allowing for α-returns to scale technology and employing the quadratic mean of order r aggregator function. Because we allow parameter α to be any strictly positive real number and r to be any non-zero real number, it includes a variety of functional forms as special cases, and all of them can now be considered as flexible functional forms.

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

Access this article

Subscribe and save

Springer+ Basic
$34.99 /Month
  • Get 10 units per month
  • Download Article/Chapter or eBook
  • 1 Unit = 1 Article or 1 Chapter
  • Cancel anytime
Subscribe now

Buy Now

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

Instant access to the full article PDF.

Similar content being viewed by others

Notes

  1. See Lau (1986) and Chapter 6 of Sickles and Zelenyuk (2019) for details.

  2. Diewert (1992b) showed that the square-root quadratic functional form is flexible only at the production frontier, while Balk (1998) showed that it is also flexible at any point of technology set.

  3. See Färe and Primont (1995) and Sickles and Zelenyuk (2019) for the regularity conditions and their implications.

  4. This definition comes from Boussemart et al. (2009). The concept of α-returns to scale goes back to Aczél (1966), Lau (1978) and Färe and Mitchell (1993). While α itself has been often predetermined in applications, Boussemart et al. (2019) recently proposed an estimation procedure for α, which is applicable to various specifications of underlying technology.

  5. As Diewert (1993) mentions, the weighted means of order r is widely regarded as a family of means, because it includes various types of means as special cases under different specification of r. On the same reasoning, the proposed functional form (10) is considered as a family of flexible functional forms. This functional form allows for any rate of returns α > 0. Moreover, once α is determined, it can include various flexible functional forms as special cases by using different choices (for non-zero real values) of r.

  6. See Diewert (1980), Färe and Sung (1986), Chambers (1988) and Hill (2006) for earlier uses of a variant of the quadratic-mean-of-order-r aggregator function.

  7. Diewert (1980) shows that for a single output case, gα,r converges to the translog functional form, which in turn includes the Cobb-Douglas form as a special case. This can potentially be generalized to a multi-output case, which we leave for future research.

  8. If α = 1, aj,k = 0 for all j ≠ k, bu,v = 0 for all u ≠ v, am = 0 for all m, bn = 0 for all n, and cm,n = 0 for all m and n, g1,r coincides with the CES functional form. Since these restrictions severely limit the interactions between inputs and outputs, this case is not regarded as a flexible functional form.

  9. Note that gα,r defined by (10)–(12) is a flexible functional form without imposing the restrictions (13)–(16). Thus, they are considered as harmless restrictions. Superlative index number results often rely on this type of restrictions. See Diewert (1992a, b), Mizobuchi and Zelenyuk (2021) and Sickles and Zelenyuk (2019).

  10. When α = 1 and r = 2, Proposition 1 coincides with Theorem 7 of Diewert (1992b).

  11. When α = 1 and r = 2, Corollary 1 coincides with Theorem 5 of Diewert (1992b).

  12. When r changes, the proposed functional form (10) changes in the third and higher order. As Hill (2006) empirically compares the price indexes defined by the quadratic-mean-of-order r, under different values of r, the empirical comparison of the functional form (10) under different values of r would be useful for understanding this functional form. This idea owes much to the suggestion by the referee. On the other hand, since it is beyond the scope and goal of this paper, we leave it for future research.

  13. When a technology satisfies α-returns to scale, the revenue function is homogenous of degree α in input quantities and the cost function is homogeneous of degree 1/α in output quantities. Thus, by replacing α in (10) and (18) by −α, (10) and (18) become a flexible function form for the revenue and cost functions under the α-returns to scale technology.

  14. See Mizobuchi and Zelenyuk (2021) for related discussions.

  15. Euler’s theorem directly implies (26), (27), (29), and (30). Furthermore, (28) and (31) are obtained by differentiating both sides of (26) and (29). See Mas-Colell et al. (1995) for Euler’s theorem.

  16. For simplicity, we adopt the notation \(\frac{{\partial g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}} = \left. {\frac{{\partial g_{\alpha ,r}\left( {x,y} \right)}}{{\partial x_n}}} \right|_{x = x^ \ast ,\,y = y^ \ast}\) throughout this paper.

  17. This specification of parameters is just an example. There might exist an alternative specification which guarantees the second-order approximation property.

References

  • Aczél J (1966) Lectures on functional equations and their applications. Academic Press, New York

    Google Scholar 

  • Balk BM (1998) Industrial price, quantity, and productivity indices: the micro-economic theory and an application. Springer US, Boston, MA

    Book  Google Scholar 

  • Boussemart J-P, Briec W, Peypoch N, Tavéra C (2009) α-Returns to scale and multi-output production technologies. Eur J Oper Res 197(1):332–339

    Article  Google Scholar 

  • Boussemart J-P, Briec W, Leleu H, Ravelojaona P (2019) On estimating optimal α-returns to scale. J Oper Res Soc 70(1):1–11

    Article  Google Scholar 

  • Chambers RG (1988) Applied production analysis: a dual approach. Cambridge University Press, New York, NY

    Google Scholar 

  • Christensen LR, Jorgenson DW, Lau L (1975) Transcendental logarithmic utility functions. Am Econ Rev 65(3):367–383

    Google Scholar 

  • Christensen LR, Jorgenson DW, Lau LJ (1973) Transcendental logarithmic production frontiers. Rev Econ Stat 55(1):28–45

    Article  Google Scholar 

  • Diewert WE (1971) An application of the Shephard duality theorem: a generalized leontief proudction function. J Political Econ 79(3):481–507

    Article  Google Scholar 

  • Diewert WE (1980) Aggregation problems in the measurement of capital. In: Usher D (ed) The measurement of capital. University of Chicago Press, Chicago and London, p 433–538

    Google Scholar 

  • Diewert WE (1992a) Exact and superlative welfare change indicators. Econ Inq 30(4):565–582

    Article  Google Scholar 

  • Diewert WE (1992b) Fisher ideal output, input, and productivity indexes revisited. J Prod Anal 3(3):211–248

    Article  Google Scholar 

  • Diewert WE (1993) Symmetric means and choice under uncertainty. In: Diewert WE, Nakamura AO (eds) Essays in index number theory, vol. 1. North-Holland, Amsterdam, p 357–441

  • Färe R, Mitchell T (1993) Multiple outputs and “Homotheticity”. South Econ J 60(2):287–296

    Article  Google Scholar 

  • Färe R, Primont D (1995) Multi-output production and duality: theory and applications. Kluwer Academic Publishers, Boston, MA

    Book  Google Scholar 

  • Färe R, Sung KJ (1986) On second-order Taylor’s-series approximation and linear homogeneity. Aequ Math 30:180–186

    Article  Google Scholar 

  • Hill RJ (2006) Superlative index numbers: not all of them are super. J Econom 130(1):25–43

    Article  Google Scholar 

  • Lau LJ (1978) Applications of profit functions. In: Fuss MA, McFadden DL (eds) Production economics: a dual approach to theory and applications, vol. I: the theory of production. North-Holland, Amsterdam

  • Lau LJ (1986) Functional forms in econometric model building. In: Griliches Z, Intriligator MD (eds) Handbook of econometrics. Elsevier, Amsterdam, p 1515–1566

    Google Scholar 

  • Mas-Colell A, Whinston MD, Green JR (1995) Microeconomic theory. Oxford University Press, New York, NY

    Google Scholar 

  • Mizobuchi H (2017) A superlative index number formula for the Hicks-Moorsteen productivity index. J Prod Anal 48(2–3):167–178

    Article  Google Scholar 

  • Mizobuchi H, Zelenyuk V (2021) Quadratic-mean-of-order-r indexes of output, input and productivity. J Prod Anal 56(2–3):133–138

    Article  Google Scholar 

  • Sickles RC, Zelenyuk V (2019) Measurement of Productivity and Efficiency: Theory and Practice. Cambridge University Press, New York, NY

Download references

Acknowledgements

We thank the editors, two anonymous referees, Bert Balk, Erwin Diewert, Knox Lovell, Antonio Peyrache, Prasada Rao, Christopher O’Donnell, and the seminar participants at the 2017 CEPA International Workshop on Performance Analysis: Theory and Practice, November 2017, Brisbane, Australia. We also thank Bao Hoang Nguyen, Zhichao Wang, Arhan Boyd, and Evelyn Smart for their feedback on proofreading this paper. HM acknowledges the financial support from the Japan Society for the Promotion of Science: Grant-in-Aid for Scientific Research (C) (18K01552, 21K01432). VZ acknowledges the financial support from the Australian Research Council (ARC FT170100401) and The University of Queensland. All remaining errors are the authors’ responsibility.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Hideyuki Mizobuchi.

Ethics declarations

Conflict of interest

The authors declare no competing interests.

Additional information

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

Appendix

Appendix

In this appendix, we provide the proof of Propositions 1. For this purpose, we introduce the following lemma. They are the restrictions on the derivatives of the output distance function, implied by homogeneity conditions (5) and (8).

Lemma 1: Suppose that the technology satisfies α-returns to scale and the output distance function Do is twice differentiable at (x*, y*). Then, Do satisfies the following equationsFootnote 15:

$$\mathop {\sum}\limits_{n = 1}^N {\frac{{\partial D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n}}} x_n^ \ast = - \alpha D_o\left( {x^ \ast ,y^ \ast } \right)$$
(26)
$$\mathop {\sum}\limits_{v = 1}^N {\frac{{\partial ^2D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_u\,\partial x_v}}} x_v^ \ast = - \left( {1 + \alpha } \right)\frac{{\partial D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_u}},\,\forall u \in \{ {1, \ldots ,N} \}$$
(27)
$$\mathop {\sum}\limits_{n = 1}^N {\frac{{\partial ^2D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n\,\partial y_m}}x_n^ \ast = - \alpha \frac{{\partial D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_m}},\,\forall m \in \{ {1, \ldots ,M} \}}$$
(28)
$$\mathop {\sum}\limits_{m = 1}^M {\frac{{\partial D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_m}}y_m^ \ast = D_o\left( {x^ \ast ,y^ \ast } \right)}$$
(29)
$$\mathop {\sum}\limits_{j = 1}^M {\frac{{\partial ^2D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_j\,\partial y_k}}y_j^ \ast = 0,\,\forall k \in \{ {1, \ldots ,M} \}}$$
(30)
$$\mathop {\sum}\limits_{m = 1}^M {\frac{{\partial ^2D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n\,\partial y_m}}y_m^ \ast = \frac{{\partial D_o\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n}},\,\forall n \in \{ {1, \ldots ,N} \}}$$
(31)

Proof of Proposition 1: This proposition claims that even under the restrictions of (13)–(17), gα,r can approximate any arbitrary output distance function \(D_o^ \ast\) to the second-order at (x*, y*). Stated differently, the following equations are satisfied.Footnote 16

$$g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right) = D_o^ \ast \left( {x^ \ast ,y^ \ast } \right)$$
(32)
$$\frac{{\partial g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n}} = \frac{{\partial D_o^ \ast \left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n}},\,\forall n \in \{ {1, \ldots ,N} \}$$
(33)
$$\frac{{\partial g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_m}} = \frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}},\,m \in \{ {1, \ldots ,M} \}$$
(34)
$$\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_u\,\partial x_v}} = \frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\,\partial x_v}},\,\forall u,v \in \{ {1, \ldots ,N} \}$$
(35)
$$\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_j\,\partial y_k}} = \frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\,\partial y_k}},\,\forall j,\,k \in \{ {1, \ldots ,M} \}$$
(36)
$$\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_n\,\partial y_m}} = \frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n\,\partial y_m}},\,\forall n \in \{ {1, \ldots ,N} \},\,\forall m \in \{ {1, \ldots ,M} \}$$
(37)

We define the parameters in gα,r as follows:Footnote 17

$$\sigma = D_o^ \ast \left( {x^ \ast ,y^ \ast } \right)$$
(38)
$$\begin{array}{l}a_{j,k} = \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\partial y_k}}} \right) - \left( {1 - r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\right.\\ \qquad \quad \left. \times\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_k}}} \right) \right)y_j^{ \ast 1 - r/2}y_k^{ \ast 1 - r/2}\\ \quad \qquad \qquad \qquad \qquad \qquad \qquad \qquad \forall j \ne k \in \{ {1, \ldots ,M} \}\end{array}$$
(39)
$$\begin{array}{l}a_{m,m} = \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left(\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m^2}}} \right) - \left( {1 - r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\right.\\ \quad \quad \quad \left. \times \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)^2 + \left( {1 - \frac{r}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)\left( {\frac{1}{{y_m^ \ast }}} \right)\right)y_m^{ \ast 2 - r}\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \forall m \in \{ {1, \ldots ,M} \}\end{array}$$
(40)
$$\begin{array}{l}b_{u,v} = \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\partial x_v}}} \right) + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \right.\\ \quad \quad \left. \times \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_v}}} \right) \right)x_u^{ \ast 1 - \alpha r/2}x_v^{^\ast 1 - \alpha r/2}\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \qquad \qquad \forall u \ne v \in \{ {1, \ldots ,N} \}\end{array}$$
(41)
$$\begin{array}{l}b_{n,n} = \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( - \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n^2}}} \right) + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\right.\\ \quad \quad \quad \left. \times \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)^2 - \left( {1 - \frac{{\alpha r}}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)\left( {\frac{1}{{x_n^ \ast }}} \right)\right)x_n^{ \ast 2 - \alpha r}\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \qquad \quad \forall n \in \{ {1, \ldots ,N} \}\end{array}$$
(42)
$$\begin{array}{l}c_{m,n} = \left( {\frac{4}{{\alpha r}}} \right) \left( {\frac{{y_m^{ \ast 1 - r/2}x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left(\left( { - \frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}}} \right)\vphantom{\left( {\frac{{y_m^{ \ast 1 - r/2}x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)}\right.\\ \quad \quad \quad \left.\vphantom{\left( {\frac{{y_m^{ \ast 1 - r/2}x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)}+ \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)\right)\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \qquad \forall n \in \{ {1, \ldots ,N} \},\,\forall m \in \{{1, \ldots ,M} \}\end{array}$$
(43)

First, we show that these parameters specified by (38)–(43) satisfy the restrictions (14)–(17). By summing \(a_{j,k}y_k^{ \ast r/2}\) over k, we can derive the following equation:

$$\begin{array}{l}\mathop {\sum}\limits_{k = 1}^M {a_{j,k}y_k^{ \ast r/2}} = \mathop {\sum}\limits_{k \ne j} {a_{j,k}y_k^{ \ast r/2} + a_{j,j}y_j^{ \ast r/2}}\\= \mathop {\sum}\limits_{k \ne j} \left( {\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left(\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\partial y_k}}} \right) - \left( {1 - r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_k}}} \right)\right)y_j^{ \ast 1 - r/2}y_k^ \ast } \right)\\\quad \quad+ \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left(\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j^2}}} \right) - \left( {1 - r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)^2 \right. \\ \quad \quad \left. + \left( {1 - \frac{r}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)\left( {\frac{1}{{y_j^ \ast }}} \right)\right)y_j^{ \ast 2 - r/2}\end{array}$$

from (39) and (40),

$$\begin{array}{l} = \left( \mathop {\sum}\limits_{k = 1}^M \left(\left( \frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\partial y_k}} \right)y_k^ \ast\right.\right.\\\left.\left. - \left( {1 - r} \right)\left(\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_k}}} \right)y_k^ \ast\right) \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)y_j^{ \ast 1 - r/2}\\+ \left( {1 - \frac{r}{2}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^{ \ast 1 - r/2}\\ = \left( {\mathop {\sum}\limits_{k = 1}^M {\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\partial y_k}}y_k^ \ast } \right)} } \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)y_j^{ \ast 1 - r/2}+ \left( - \left( {1 - r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\mathop {\sum}\limits_{k = 1}^M {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_k}}y_k^ \ast } \right)} } \right)\right.\\\left.+ \left( {1 - \frac{r}{2}} \right) \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^{ \ast 1 - r/2}\\ = \left( { - \left( {1 - r} \right)\left( {\frac{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) + \left( {1 - \frac{r}{2}} \right)} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^{ \ast 1 - r/2}\end{array}$$

from (29) and (30),

$$= \left( {\frac{r}{2}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^{ \ast 1 - r/2}$$
(44)

Then, it implies (14) as follows:

$$\begin{array}{l}\mathop {\sum}\limits_{j = 1}^M {\mathop {\sum}\limits_{k = 1}^M {a_{j,k}y_j^{ \ast r/2}y_k^{ \ast r/2}} } \\= \mathop {\sum}\limits_{j = 1}^M {\left( {\left( {\frac{r}{2}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^{ \ast 1 - r/2}} \right)y_j^{^\ast r/2}} \end{array}$$
$$\begin{array}{l} = \left( {\frac{r}{2}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\mathop {\sum}\limits_{j = 1}^M {\left( {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j}}} \right)y_j^ \ast } \right)} \\= \left( {\frac{r}{2}} \right)\left( {\frac{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) = \frac{r}{2}\end{array}$$

from (29)

By summing \(b_{u,v}x_v^{ \ast \alpha r/2}\) over v, we can derive the following equation.

$$\begin{array}{l}\mathop {\sum}\limits_{v = 1}^N {b_{u,v}x_v^{ \ast \alpha r/2}} = \mathop {\sum}\limits_{v \ne u} {b_{u,v}x_v^{ \ast \alpha r/2} + b_{u,u}x_u^{ \ast \alpha r/2}} \\ = \mathop {\sum}\limits_{v \ne u} {\left(\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\partial x_v}}} \right) + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_v}}} \right)\right)x_u^{ \ast 1 - \alpha r/2}x_v^ \ast \right)} + \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\ \qquad\times \left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u^2}}} \right) + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)^2 - \left( {1 - \frac{{\alpha r}}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\left( {\frac{1}{{x_u^ \ast }}} \right)\right)x_u^{ \ast 2 - \alpha r/2}\end{array}$$

from (41) and (42),

$$\begin{array}{l} = \left( {\mathop {\sum}\limits_{v = 1}^N {\left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\partial x_v}}} \right)x_v^ \ast + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_v}}} \right)x_v^ \ast \right)} } \right)\\ \quad\times\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)x_u^{ \ast 1 - \alpha r/2} - \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {1 - \frac{{\alpha r}}{2}} \right)\\ \times\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^{ \ast 1 - \alpha r/2}\end{array}$$
$$\begin{array}{l} = \left(- \mathop {\sum}\limits_{v = 1}^N {\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\partial x_v}}x_v^ \ast } \right)} + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\mathop {\sum}\limits_{v = 1}^N {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_v}}x_v^ \ast } \right)}\right)\\ \quad\times\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)x_u^{ \ast 1 - \alpha r/2} - \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \\\quad\times\left( {1 - \frac{{\alpha r}}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^{ \ast 1 - \alpha r/2}\end{array}$$
$$\begin{array}{l} = \left(- \left( { - \left( {1 + \alpha } \right)} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\right.\\ \left. + \left( {1 + r} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)\left( { - \alpha } \right)D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)\right)\\ \quad\times\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)x_u^{ \ast 1 - \alpha r/2} - \left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\\times\left( {1 - \frac{{\alpha r}}{2}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^{ \ast 1 - \alpha r/2}\end{array}$$

from (26) and (27),

$$\begin{array}{l} = \left( {\left( {1 + \alpha } \right) - \left( {1 + r} \right)\alpha - \left( {1 - \frac{{\alpha r}}{2}} \right)} \right)\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\ \qquad\times \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^{ \ast 1 - \alpha r/2}\end{array}$$
$$= - \left( {\frac{{\alpha r}}{2}} \right)\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^{ \ast 1 - \alpha r/2}$$
(45)

Then, it implies (15) as follows:

$$\begin{array}{l}\mathop {\sum}\limits_{u = 1}^N {\mathop {\sum}\limits_{v = 1}^N {b_{u,v}x_u^{ \ast \alpha r/2}x_v^{ \ast \alpha r/2}} } = \\ \mathop {\sum}\limits_{u = 1}^N {\left( { - \left( {\frac{{\alpha r}}{2}} \right)\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)} \right)x_u^ \ast } \end{array}$$
$$\begin{array}{l} = - \left( {\frac{{\alpha r}}{2}} \right)\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\mathop {\sum}\limits_{u = 1}^N {\left( {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u}}} \right)x_u^ \ast } \right)} \\ = - \left( {\frac{{\alpha r}}{2}} \right)\left( {\frac{1}{{\alpha ^2}}} \right)\left( {\frac{{ - \alpha D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) = \frac{r}{2}\end{array}$$

from (26).

By summing \(c_{m,n}y_m^{ \ast r/2}\) over m, we obtain (16) so as

$$\begin{array}{l}\mathop {\sum}\limits_{m = 1}^M {c_{m,n}y_m^{ \ast r/2}} = \\\left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum }\nolimits_{j = 1}^M a_jy_j^{ \ast r/2}} \right)\left( {\mathop {\sum }\nolimits_{u = 1}^N b_ux_u^{ \ast - \alpha r/2}} \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\\times\left( {\mathop {\sum}\limits_{m = 1}^M {\left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}}} \right)y_m^ \ast + \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)y_m^ \ast \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)\right)} } \right)\end{array}$$

from (43),

$$\begin{array}{l} = \left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum }\nolimits_{j = 1}^M a_jy_j^{ \ast r/2}} \right)\left( {\mathop {\sum }\nolimits_{u = 1}^N b_ux_u^{ \ast - \alpha r/2}} \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\\times\left(- \left( {\mathop {\sum}\limits_{m = 1}^M {\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}}} \right)} y_m^ \ast } \right)\right.\\\left. + \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)\left( {\mathop {\sum}\limits_{m = 1}^M {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)} y_m^ \ast } \right)\right)\end{array}$$
$$\begin{array}{l} = \left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{x_n^{ \ast 1 + \alpha r/2}}}{{\left( {\mathop {\sum }\nolimits_{j = 1}^M a_jy_j^{ \ast r/2}} \right)\left( {\mathop {\sum }\nolimits_{u = 1}^N b_ux_u^{ \ast - \alpha r/2}} \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\ \times \left( { - \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right) + \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)\left( {\frac{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)} \right) = 0\end{array}$$

from (29) and (31).

By summing \(c_{m,n}x_n^{ \ast - \alpha r/2}\) over n, we obtain (17) so as

$$\begin{array}{l}\mathop {\sum}\limits_{n = 1}^N {c_{m,n}x_n^{ \ast - \alpha r/2}} \\= \left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{y_m^{ \ast 1 - r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\\times\left( \mathop {\sum}\limits_{n = 1}^N \left(- \left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}}} \right)x_n^ \ast + \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)x_n^ \ast \right) \right)\end{array}$$

from (43),

$$\begin{array}{l} = \left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{y_m^{ \ast 1 - r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\\\times\left(- \left( {\mathop {\sum}\limits_{n = 1}^N {\left( {\frac{{\partial ^2D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}}} \right)x_n^ \ast } } \right)\right.\\ +\left. \left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right) \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)\left( {\mathop {\sum}\limits_{n = 1}^N {\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}}} \right)x_n^ \ast } } \right)\right)\end{array}$$
$$\begin{array}{l} = \left( {\frac{4}{{\alpha r}}} \right)\left( {\frac{{y_m^{ \ast 1 - r/2}}}{{\left( {\mathop {\sum}\nolimits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\nolimits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)}}} \right)\left( {\frac{1}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\left(- \left( { - \alpha } \right)\left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right)\right.\\ \left. + \left( {\frac{{\partial D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}}} \right) \left( {\frac{{ - \alpha D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}{{D_o^ \ast \left( {x^ \ast ,\,y^ \ast } \right)}}} \right)\right) = 0\end{array}$$

from (26) and (28).

Second, we show that these parameters specified by (38)–(43) satisfy the equations for the second-order approximation (32)–(37).

The value of gα,r evaluated at (x*, y*) is given by the following equation, using (14)–(17):

$$g_{\alpha ,r}\left( {x^ \ast ,y^ \ast } \right) = \sigma$$
(46)

Substituting (38) into (46), we obtain (32).

The first-order derivatives of gα,r with respect to inputs x evaluated at (x*, y*) are given by the following equations using (14)–(17):

$$\begin{array}{l}\frac{{\partial g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n}} = - \alpha \sigma \left( {\frac{r}{2}} \right)^{ - 1}x_n^{ \ast \alpha r/2 - 1}\left( {\mathop {\sum}\limits_{v = 1}^N {b_{n,v}x_v^{ \ast \alpha r/2}} } \right)\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad\forall n \in \{ {1, \ldots ,N} \}\end{array}$$
(47)

Substituting (38) and (45) into (47), we obtain (33):

The first-order derivatives of gα,r with respect to outputs y evaluated at (x*, y*) are given by the following equations using (14)–(17):

$$\begin{array}{l}\frac{{\partial g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m}} = \sigma \left( {\frac{r}{2}} \right)^{ - 1}y_m^{ \ast r/2 - 1}\left( {\mathop {\sum}\limits_{k = 1}^M {a_{m,k}y_k^{ \ast r/2}} } \right)\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad \forall m \in \{ {1, \ldots ,M} \}\end{array}$$
(48)

Substituting (38) and (44) into (48), we obtain (34):

The second-order derivatives of gα,r with respect to inputs x evaluated at (x*, y*) are given by the following equations using (14)–(17):

$$\begin{array}{l}\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_u\partial x_v}} = \alpha \sigma \left( {\frac{r}{2}} \right)^{ - 1}\left(\alpha \left( {1 + r} \right)\left( {\frac{r}{2}} \right)^{ - 1}\left( {\mathop {\sum}\limits_{n = 1}^N {b_{u,n}x_n^{ \ast \alpha r/2}} } \right)\right.\\\left. \times \left( {\mathop {\sum}\limits_{n = 1}^N {b_{v,n}x_n^{ \ast \alpha r/2}} } \right)x_u^{ \ast \alpha r/2 - 1}x_v^{ \ast \alpha r/2 - 1}-\left( {\frac{{\alpha r}}{2}} \right)b_{u,v}x_u^{ \ast \alpha r/2 - 1}x_v^{ \ast \alpha r/2 - 1}\right)\\\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad \forall u \ne v \in \{ {1, \ldots ,N} \}\end{array}$$
(49)
$$\begin{array}{l}\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial x_n^2}} = \alpha \sigma \left( {\frac{r}{2}} \right)^{ - 1}\left(\alpha \left( {1 + r} \right)\left( {\frac{r}{2}} \right)^{ - 1}\left( {\mathop {\sum}\limits_{u = 1}^N {b_{n,u}x_u^{ \ast \alpha r/2}} } \right)^2 x_n^{ \ast \alpha r - 2}\right.\\\qquad\qquad\quad\left.- \left( {\frac{{\alpha r}}{2}} \right)b_{n,n}x_n^{ \ast \alpha r - 2}- \left( {\frac{{\alpha r}}{2} - 1} \right)\left( {\mathop {\sum}\limits_{u = 1}^N {b_{n,u}x_u^{ \ast \alpha r/2}} } \right)x_n^{ \ast \alpha r/2 - 2}\right)\\\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad \forall n \in \{ {1, \ldots ,N} \}\end{array}$$
(50)

Either substituting (38), (41), and (45) into (49) or substituting (38), (42), and (45) into (50), we obtain (35).

The second-order derivatives of gα,r with respect to outputs y evaluated at (x*, y*) are given by the following equations, using (14)–(17):

$$\begin{array}{l}\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_j\partial y_k}} =\\ \sigma \left( {\frac{r}{2}} \right)^{ - 1}\left(\left( {1 - r} \right)\left( {\frac{r}{2}} \right)^{ - 1}\left( {\mathop {\sum}\limits_{m = 1}^M {a_{j,m}y_m^{ \ast r/2}} } \right)\right. \\ \left. \times \left( {\mathop {\sum}\limits_{m = 1}^M {a_{k,m}y_m^{ \ast r/2}} } \right)y_j^{ \ast r/2 - 1}y_k^{ \ast r/2 - 1} + \left( {\frac{r}{2}} \right)a_{j,k}y_j^{ \ast r/2 - 1}y_k^{ \ast r/2 - 1}\right)\\\qquad \qquad\qquad\qquad\qquad\qquad\qquad \forall j \ne k \in \{ {1, \ldots ,M} \}\end{array}$$
(51)
$$\begin{array}{l}\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m^2}} = \sigma \left( {\frac{r}{2}} \right)^{ - 1}\left(\left( {1 - r} \right)\left( {\frac{r}{2}} \right)^{ - 1}\left( {\mathop {\sum}\limits_{j = 1}^M {a_{m,j}y_j^{ \ast r/2}} } \right)^2 y_m^{ \ast r - 2}\right.\\ \left. \qquad \qquad \qquad + \left( {\frac{r}{2}} \right)a_{m,m}y_m^{ \ast r - 2} + \left( {\frac{r}{2} - 1} \right)y_m^{ \ast r/2 - 2}\left( {\mathop {\sum}\limits_{j = 1}^M {a_{m,j}y_j^{ \ast r/2}} } \right) \right)\\\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\qquad\quad\quad \forall m \in \{ {1, \ldots ,M} \}\end{array}$$
(52)

Either substituting (38), (39), and (44) into (51) or substituting (38), (40), and (44) into (52), we obtain (36).

The second-order cross-derivatives of gα,r with respect to inputs x and outputs y evaluated at (x*, y*) are given by the following equation, using (14)–(17):

$$\begin{array}{l}\frac{{\partial ^2g_{\alpha ,r}\left( {x^ \ast ,\,y^ \ast } \right)}}{{\partial y_m\partial x_n}} =\\\alpha \sigma \left(\left( {\frac{r}{2}} \right)^{ - 2}\left( {\mathop {\sum}\limits_{j = 1}^M {a_{m,j}y_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\limits_{u = 1}^N {b_{n,u}x_u^{ \ast \alpha r/2}} } \right)y_m^{ \ast r/2 - 1}x_n^{ \ast \alpha r/2 - 1}\right. \\ \left. - \left( {\frac{r}{4}} \right)\left( {\mathop {\sum}\limits_{j = 1}^M {a_jy_j^{ \ast r/2}} } \right)\left( {\mathop {\sum}\limits_{u = 1}^N {b_ux_u^{ \ast - \alpha r/2}} } \right)c_{m,n}y_m^{ \ast r/2 - 1}x_n^{ \ast - \alpha r/2 - 1}\right)\\ \quad \quad \quad \quad \quad \quad \quad \quad \quad \quad\forall m \in \{ {1, \ldots ,M} \},\,\forall n \in \{ {1, \ldots ,N} \}\end{array}$$
(53)

Substituting (38), (43)–(45) into (53), we obtain (37). QED.

Rights and permissions

Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Mizobuchi, H., Zelenyuk, V. A generalized flexible functional form for α-returns to scale. J Prod Anal 59, 217–224 (2023). https://doi.org/10.1007/s11123-023-00658-3

Download citation

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11123-023-00658-3

Keywords

JEL classification