lih-ru chen national chengchi university gene c. lai washington state university jennifer l. wang

20
1 Nonspecialized Strategy versus Specialized Strategy Evidence from the Property-Liability Insurance Industry Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang National Chengchi University ARIA Meeting, August, 2007

Upload: helen-powers

Post on 30-Dec-2015

15 views

Category:

Documents


0 download

DESCRIPTION

Nonspecialized Strategy versus Specialized Strategy : Evidence from the Property-Liability Insurance Industry. Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang National Chengchi University ARIA Meeting, August, 2007. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

1

Nonspecialized Strategy versus Specialized Strategy :Evidence from the Property-Liability Insurance Industry

Lih-Ru ChenNational Chengchi University

Gene C. LaiWashington State University

Jennifer L. WangNational Chengchi University

ARIA Meeting, August, 2007

Page 2: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

2

Why do we observe the long-lived coexistence of joint producers and specialist in the U.S. property-liability insurance industry? Whether joint producers or specialist is more efficient ?

1.whether nonspecialized strategy or specialized strategy is more efficient for property-liability insurers. 2.what types of insurers are likely to realize economies of scope.

Research Question and Research Purpose

Page 3: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

3

Conglomeration hypothesis Teece(1980)

Argued that Multiproduct enterprise is an efficient way of organizing economic activity.

Panzar and Willing (1981) First to introduce the concept of economies of scope.

Strategic focus hypothesis Comment and Jarrell (1995)

The greater corporate focus is consistent with shareholder wealth maximization.

Kanatas and Qi (2003) Integrated financial services market is less innovative than one

with specialized intermediaries. Chen (JFQA,2006)

focused firms exhibit significantly better post-investment operating performance than diversified firms.

The Literature I

Page 4: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

4 The Literature II

Insurance RelatedKellner and Mathewson (1983); Meador, Ryan, and Shellhorm (1998), Cummins, Weiss, and Zi (2003), and Hirao and Inoue (2004) found results consistent with diversification strategy.Grace and Timme (1992) and Yuengert (1993)Berger et al. (2000) found mixed results.Jeng and Lai (2005) find that Keiretsu firms seem to be more cost-efficient than nonspecialized independent firms.

Page 5: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

5 The main contribution of this study

1. This paper examines diversification/focus strategy from the perspective of a property-liability insurer rather than two sectors of insurance industry.

2. We examines whether the technology of nonspecialists is the same as that of the specialists using the cross-frontier approach.

3. Focusing on a more recent time period enables us to determine whether the most recent wave of financial consolidation will change the focus strategy.

Page 6: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

6Data and Methodology

Sample & Data Period: 1997-2004

Insurer’s Annual Statement from NAIC Herfindahl index of net premium is used to

identify the specialists and nonspecialists.

Methodology Data Envelopment Analysis (DEA) Stochastic Frontier Approach (SFA)

Page 7: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

7 Hypotheses Development

H1: The Nonspecialized Hypothesis H2: The Specialized Hypothesis H3: The technology of nonspecialists is the same

as that of the specialists. H4: It is infeasible to replicate nonspecialists

input-output bundle using the specialists technology.

H5: There are no scope economies for either nonspecialists or specialists.

Page 8: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

8 Inputs/Outputs using in value-added approach

Cummins and Weiss (1993), Cummins, Weiss, and Zi (1999), and Cummins et al. (2004)Output

Y1= Losses incurred in short-tailed personal lines

Y2= Losses incurred in long-tailed personal lines

Y3= Losses incurred in short-tailed commercial lines

Y4= Losses incurred in long-tailed commercial lines

Y5= Total Invested Assets

Input Input Price

X1=Labor cost P1=Price of Labor

X2=Business Service P2= Price of Business Service

X3=Equity P3= Price of Equity

X4=Debt P4= Price of Debt

Page 9: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

9 Table 1 Summary Statistics for non-specialists and specialists Non-specialists Specialists

Mean Standard deviation

Test stat

Mean Standard deviation

Output

Y1 11,257,223 24,468,928 *** 14,374 151,373

Y2 23,298,699 50,688,162 *** 123,717 831,187

Y3 9,123,543 26,047,217 *** 4,143,038 20,745,509

Y4 46,590,136 110,748,946 *** 6,617,191 15,993,637

Y5 430,559,048 991,784,924 *** 78,075,255 193,335,845

Input

X1 42,615 92,319 *** 5,478 12,837

X2 123,143 288,458 *** 10,451 24,917

X3 165,473,202 427,431,018 *** 27,754,645 53,933,908

X4 285,914,578 663,981,251 *** 36,906,586 77,479,769

Input prices

P1 435 13 435 13

P2 281 21 281 21

P3 1.94 1.91 ** 1.47 7.07

P4 0.038 0.0178 0.038 0.0178

Number of firms 133 144

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 10: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

10

Year Csp(Xsp,Ysp) Cnsp(YnspXnsp) Cnsp(Xsp,Ysp) Csp(YnspXnsp) Mean 0.4895 0.6798 0.9826 1.4439 (0.2542) (0.2051) (1.7989) (1.4254) 1997 0.4909 0.7509 1.0541 1.3417 (0.1719) (0.1950) (2.3301) (0.8190) 1998 0.4210 0.7333 1.1153 1.3192 (0.1880) (0.1679) (2.1386) (0.8612) 1999 0.7001 0.7788 1.0830 1.1194 (0.1904) (0.1204) (1.5154) (0.5212) 2000 0.4322 0.7298 0.9422 2.8919 (0.2229) (0.1426) (1.5715) (2.1899) 2001 0.2530 0.5821 0.8858 2.5850 (0.2154) (0.1585) (1.9626) (2.0548) 2002 0.4328 0.4606 0.9890 0.5915 (0.2359) (0.2378) (2.3438) (0.3386) 2003 0.7809 0.7649 0.9118 1.0040 (0.1930) (0.1706) (0.9855) (0.5304) 2004 0.4116 0.6380 0.8732 0.6771 (0.1602) (0.1981) (0.8809) (0.3699)

Table 5 Cost efficiency results: 1997–2004

Page 11: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

11

Year

Csp(Xsp,Ysp) vs. Cnsp(Xnsp,Ynsp)

Csp(Xsp,Ysp) vs. Cnsp(Xsp,Ysp)

Cnsp(YnspXnsp) vs. Csp(YnspXnsp)

Mean *** *** *** 1997 ** *** *** 1998 *** *** *** 1999 *** *** 2000 *** *** *** 2001 ** *** *** 2002 *** *** 2003 *** 2004 *** *** ***

Table 5 Cost efficiency results: 1997–2004 (continuous)

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 12: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

12

( , )( , ) 1

( , )np np np

t np npsp np np

T Y XF Y X

T Y X

Cummins et al., (1999)F-scores measure the dominance with respect to the frontiers.

Page 13: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

13Table 7 Dominance Statistics by Size Quartile

Panel C: Cost Frontiers Fc(yi,xi) Nonspecialist Specialist Quartile 1 Mean 0.054 *** –1.184 *** (0.557) (2.262) Quartile 2 Mean 0.268 *** –0.808 *** (0.495) (1.132) Quartile 3 Mean 0.330 *** –1.342 *** (0.383) (3.135) Quartile 4 Mean 0.257 *** –1.161 *** (0.471) (2.237)

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 14: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

14

Nonspecialists Specialists

Variables Mean

Standard

deviation

Test

stat. Mean Standard

deviation

Total Assets (Thousands) $736,912 $1,653,818 *** $137,014 $314,393

Commercial Insurance

Output/Total Output 61.78% 21.85% *** 93.79% 23.34%

Cost Efficiency 68.27% 19.92% *** 48.39% 24.86%

ROE 5.34% 6.56% *** 11.04% 30.48%

Capital to Asset Ratio 35.44% 13.58% *** 48.44% 25.76%

WCONC 0.05% 0.21% *** 0.02% 0.09%

Reinsurance Ratio 43.90% 27.55% *** 17.60% 76.70%

Agent Balance

(Thousands) $45,196 $100,297 *** $6,536 $32,540

Mutual Dummy 20.69% 40.53% *** 30.36% 46.01%

Stock Dummy 79.31% 40.53% *** 69.64% 46.01%

Vertical Integration

Dummy 8.91% 28.50% *** 31.96% 46.66%

N 1,010 1,023

Table 8 Summary Statistics for regression

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 15: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

15

Panel A: Cost Dominance Regression Random Effect Model

Independent variables Coefficient T stat

Intercept 0.4744 1.73*

Q2* nonspecialist 0.4251 1.90*

Q3* nonspecialist 0.5855 2.50**

Q4* nonspecialist 0.6857 2.85***

Q2* specialist –0.1000 –0.69

Q3* specialist 0.0791 0.43

Q4* specialist –0.6359 –2.08**

% Commercial Output –1.0134 –3.31***

% Commercial Output* nonspecialist –0.3406 –0.97

ROE Standard Deviation –0.5597 –0.68

ROE Standard Deviation* nonspecialist 0.5053 0.61

Capital to Asset Ratio 0.0525 0.12

Capital to Asset Ratio* nonspecialist –0.3549 –0.77

WCONC –10.0635 –0.33

WCONC* nonspecialist –159.8470 –1.59

Stock –0.1483 –0.78

Stock* nonspecialist 0.2493 0.98

R-square 0.0774

Hausman Test for Random Effects (P value) 19.82 (0.2284)

Table 9 Dominance Regression for nonspecialists and specialists

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 16: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

16 SFA variablesVariables Definitions

y1 losses incurred for personal line

y2 losses incurred for nonpersonal lines

y3 invested assets

p1 price of labor=average weekly wages for insurance agent (standard industrial classification [SIC] Class 6411) by using U.S. Department of Labor data.

p2 business services price index =average weekly earnings of workers in SIC 7300.

Costs total expense paid of other underwriting expense

Revenues totals of net premiums written + net investment income-losses incurred current year – total expenses incurred

Profit revenue-cost

Page 17: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

17

Nonspecialists Specialists

Variables MeanStandarddeviation

Test stat.

MeanStandarddeviation

y1 34,397,479 73,221,866 *** 135,658 918,295

y2 62,233,108 144,033,832 *** 10,595,823 25,071,999

y3 428,584,725 989,856,777 *** 77,766,932 192,140,149

p1 435 13 435 13

p2 281 21 281 21

Cost ($ thousand)

$47,819 $102,550 *** $5,063 $9,790

Revenue($ thousand)

$62,983 $146,969 *** $10,908 $35,355

Profit ($ thousand)

$15,164 $64,700 *** $5,845 $27,689

Table 10 Summary Statistics for stochastic frontier approach

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 18: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

18

  Q1 Median Q3 Q4

Cost ScopeEconomy

25.35% *** –9.93% *** –28.30% *** –58.07% ***

T stat (18.85) (–7.39) (–19.67) (–48.18)

Revenue Scope Economy –8.07% *** 19.93% *** 34.24% *** 60.99% ***

T stat (-5.17) (13.61) (22.88) (49.29)

Profit ScopeEconomy

–39.14% *** –5.63% *** 14.98% *** 43.84% ***

T stat (–34.60) (–4.28) (12.46) (38.7)1

Table 11 Scope Economy by Quartile

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Page 19: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

19

Independent variables Random Effect Fixed Effect

Coefficient T stat Coefficient T stat

Intercept –2.3521 –9.34*** –1.8529 –3.37***

Log Asset 0.1172 8.75*** 0.0923 3.09***

Commercial Output % –0.2046 –6.41*** –0.4028 –4.65***

Cost Efficiency 0.2913 6.99*** 0.3854 7.93***

ROE Standard Deviation 0.0555 1.96** 0.0482 1.57

Capital to Asset Ratio 0.1944 2.82*** 0.1469 1.43

WCONC 16.5840 1.03 30.5046 0.97

Stock –9.38E–02 –3.04*** –9.77E–02 –0.86

Vertical Integration Dummy 2.70E–03 0.05

Average value of dependent variable

–12.43% –12.43%

R2 9.74% 59.79%

Note:*** Statistically significant difference at 1% level or better; ** Statistically significant difference at 5% level; * Statistically significant difference at 10% level.

Table 12 Profit Scope Economy Regression for simulated non-specialists

Page 20: Lih-Ru Chen National Chengchi University Gene C. Lai Washington State University Jennifer L. Wang

20 Concluding remarks

1. Supports for Hypothesis1,2,4 are found.2. The nonspecialists and specialists operate on

different efficient frontiers in P/L insurance industry.

3. Nonspecialists (specialists) dominate specialists (nonspecialists) in producing nonspecialists (specialists) input–output vectors.

4. The regression results generally support the coexistence of the nonspecialized strategy and specialized strategy.