Determinants of commercial mortgage-backed securities credit ratings: Australian evidence.

Newell, G. (2006) The significance of property in industry based superannuation funds in Australia. 12th Annual Conference of the Pacific Rim Real Estate Society, January 22 to 25, 2006, Auckland, New Zealand.

Nguyen, N. and Cripps, A. (2001) Predicting housing value: a comparison of multiple regression analysis and artificial neural networks, Journal of Real Estate Research, 22(3), pp. 313-336.

Ovnerud-Potter, P. (2003) CMBS: Moody's Approach to Rating Australian CMBS. Moody's Investors Service, Sydney.

Roche, T. (2002) Commercial mortgage-backed securities: A homogeneous asset class? Australian Property Journal, August, pp. 170-174.

Rowland, P.J. (1993) Property Investments and Their Financing. The Law Book Company Limited, Sydney.

Salchenberger, L.M., Cinar, E.M. and Lash, N.A. (1992) Neural networks: A new tool for predicting thrift failures, Decision Sciences, 23(4), pp. 899-915.

Shin, K. and Han, I. (2001) A case-based approach using inductive indexing for corporate bond rating, Decision Support Systems, 32(1), pp. 41-52.

SPSS Inc. (1968) SPSS 13.0. Chicago, IL.

Standard & Poor's (2001) Australian Commercial Mortgage-Backed Securitization--The Rating Process. Standard & Poor's, Melbourne.

Standard & Poor's (2004) CMBS Property Evaluation Criteria. Standard & Poor's, New York.

Standard & Poor's (2007) Fourth-Quarter and Year-End 2006 Australia and New Zealand Structured Finance Performance Trends. Standard & Poor's, Melbourne.

Surkan, A.J. and Singleton, J.C. (1990) Neural networks for bond rating improved by multiple hidden layers. Proceedings of the IEEE Inter national Conference on Neural Networks, Revised edn, San Diego, pp. 163-168.

Tam, K.Y. and Kiang, M.Y. (1992) Managerial applications of neural networks: The case of bank failure predictions, Management Science, 38(7), pp. 926-947.

Tay, D.P.H. and Ho, D.K.H. (1991) Artificial intelligence and the mass appraisal of residential apartments, Journal of Property Valuation and Investment, 10(2), pp. 525-540.

Weeraprajak, I. (2007) Faster Adaptive Network Based on Fuzzy Inference System. University of Canterbury, New Zealand.

Worzala, E.M., Lenk, M.M. and Silva, A. (1995) An exploration of neural networks and its application to real estate valuation, Journal of Real Estate Research, 10(2), pp. 185-202.

Yesilyaprak, A. (2004) Bond ratings with artificial neural networks and econometric models, American Business Review, 22(1), pp. 113-123.

Zhang, G.B., Patuwo, E. and Hu, M.Y. (1998) Forecasting with artificial neural networks: The state of the art, International Journal of Forecasting, 14(1), pp. 35-62.

Bwembya CHIKOLWA (1) (email) and Felix CHAN (2)

(1) School of Urban Development, Queensland University of Technology, Brisbane 4001, Queensland, Australia E-mail: bwembya.chikolwa@qut.edu.au; tel.: +61 7 3138 4072; fax: +61 7 3138 1170

(2) School of Economics and Finance, Curtin University of Technology, Perth WA 6845, Western Australia

SANTRAUKA

KOMERCINE HIPOTEKA UZTIKRINTU VERTYBINIU POPIERIU KREDITO REITINGU EMPIRINE ANALIZE: AUSTRALIJOS PAVYZDYS

Bwembya CHIKOLWA, Felix CHAN

Sisteminant komercine hipoteka uztikrintu vertybiniu popieriu prekybos sandorius, svarbiausias tikslas--gauti auksta kredito reitinga, nes tai daro poveiki pelningumui ir emitento sekmei. Kredito reitingu agenturos teigia, kad ju vertinimai isreiskia kiekvienos agenturos nuomone apie potencialia emitento nemokumo rizika ir daugiausia remiasi emitento gebejimo bei noro grazinti savo skola analize, kuria atlieka komitetas, taigi tyrinetojams ju reitingu kiekybiskai replikuoti nepavyktu. Taciau tyrinetojai replikavo obligaciju reitingus, remdamiesi prielaida, kad finansiniai koeficientai turi daug informacijos apie imones kredito rizika. Prognozuodami komercine hipoteka uztikrintul vertybiniu popieriu reitingus, kaip alternatyvius metodus naudojame dirbtinius neuroninius tinklus ir rangine regresija,. Rangines regresijos rezultatai rodo, kad reitingu agenturos naudoja tik ta kintamuju poaibi, kuriuos jos apibudina arba nurodo kaip svarbius komercine hipoteka uztikrintu vertybiniu popieriu reitingui, nes kai kurie is naudojamu kintamuju statistiskai nereiksmingi. Apskritai dirbtiniu neuroniniu tinklu rezultatai, prognozuojant komercine hipoteka uztikrintu vertybiniu popieriu reitingus, geresni nei rangines regresijos. Table 1. Number of Australian CMBS issues by sector (2000-2005) Sector 2000 2001 2002 2003 Diversi fied 1 2 11 7 Industrial 4 3 6 12 Office 0 3 4 5 Retail 0 0 15 9 Total 5 8 36 33 Sector 2004 2005 2000-2005 Diversified 7 14 42 Industrial 4 3 32 Office 9 10 31 Retail 0 8 32 Total 20 35 137 Source: Author's compilation from Standard and Poor's presale reports Table 2. Observations per CMBS rating Rating Training sample Test sample

Count Proportion Count Proportion A 17 14% 4 23% AA 25 21% 3 18% AAA 62 53% 3 18% BBB 14 12% 7 41% Total 118 100% 17 100% Table 3. Descriptive statistics Training sample

Issued Bond [DSCR.sup.**]

amount tenure

(A$m) (years) Mean 79.87 3.97 2.14 Standard error 7.36 0.12 0.05 Standard deviation 79.9 1.31 0.51 Minimum 1 1 1.28 Maximum 350 7 3.5 Test sample

Issued Bond [DSCR.sup.**]

amount tenure

(A$m) (years) Mean 47.59 4.94 1.81 Standard Error 13.33 0.06 0.09 Standard Deviation 54.96 0.24 0.36 Minimum 3 4 1.2 Maximum 190 5 2.7 Training sample

[LTV.sup.**] Property Geographical

diversity diversity Mean 0.46 0.29 0.48 Standard error 0.01 0.02 0.01 Standard deviation 0.1 0.18 0.15 Minimum 0.31 0.08 0.2 Maximum 0.76 1 1 Test sample

[LTV.sup.**] Property Geographical

diversity diversity Mean 0.48 0.32 0.51 Standard Error 0.02 0.04 0.06 Standard Deviation 0.07 0.18 0.26 Minimum 0.36 0.11 0.21 Maximum 0.61 0.55 0.78 Table 4. OR results Variable Model 1 (Expected sign) A 1.980 (0.310) [1.031] AA 3.053 (0.118) [1.952] AAA 5.515 (0.006) [2.006] DSCR (+) 0.471 (0.321) [0.983] LTV (-) 6.268 (0.011) [6.548] SIZELN (+) TENURE (-) PD (-) GD (+) Chi-Square 7.036 (0.030) * Pseudo 0.018 R-Square Variable Model 2 (Expected sign) A 3.861 (0.100) [2.700] AA 4.959 (0.035) [4.428] AAA 7.481 (0.002) [9.545] DSCR (+) 0.622 (0.207) [1.593] LTV (-) 8.307 (0.003) [9.004] SIZELN (+) 0.590 (0.122) [0.331] TENURE (-) -0.079 (0.565) [2.394] PD (-) GD (+) Chi-Square 9.778 (0.044) * Pseudo 0.033 R-Square Variable Model 3 (Expected sign) A 4.115 (0.088) [2.914] AA 5.221 (0.031) [4.664] AAA 7.757 (0.002) [9.768] DSCR (+) 0.801 (0.122) [2.393] LTV (-) 9.512 (0.001) [10.401] SIZELN (+) 0.693 (0.077) [3.130] TENURE (-) -0.087 (0.553) [0.353] PD (-) -1.255 (0.230) [1.438] GD (+) -0.949 (0.446) [0.580] Chi-Square 11.495 (0.074) * Pseudo 0.039 R-Square * We utilise McFadden's pseudo R-Square based on Ederington (1985) well as theoretically of all others. Regression coefficients provided with significance levels (in parenthesis) and Wald chi-square [in brackets]. Table 5. OR classifi cation accuracy of models 1-3 Model 1 Actual Predicted CMBS rating CMBS rating