**4. Empirical analysis**

0 1 1


We utilized the BCC input-oriented model to measure phase I and II to find a maximum

**DMUs Country Name DMUs Country Name** D1 Australia D18 Japan

D2 Austria D19 Korea

D3 Belgium D20 Luxembourg

D4 Canada D21 Mexico

D5 Chile D22 Netherlands

D6 Czech Republic D23 New Zealand

D7 Denmark D24 Norway

D8 Estonia D25 Poland

D9 Finland D26 Portugal

D11 Germany D28 Slovenia

D12 Greece D29 Spain

D13 Hungary D30 Sweden

D14 Iceland D31 Switzerland

D16 Israel D33 United Kingdom

D17 Italy D34 United States

D15 Ireland D32 Turkey

D10 France D27 Slovak Republic

å å (1)

0 1 1

1,2,..., ,...

³

*r jR*

*n m*

=

output with certain medial output.

202 New Developments in Renewable Energy

*Note*. Source from this study

**Table 2.** Country Names of each DMU

. . / 1; r

*n jn m jm*

*N M j n jn m jm n m*

å å

= =

U , V 0 m=1,2,...,M; n=1,2,...N;

E/

= -

*Max U Y u V X*

*N M*

*st U Y u V X*

*n m*

= =

First, multi-collinearity analysis was employed to examine the correlation coefficient be‐ tween input and input variables, and then between output and output variables [56]. We used isotonicity diagnosis to examine positive correlation coefficients between input and output variables [57]. We then used sensitivity analysis to sequentially increase or reduce the input or output variables to examine variation of efficiency [8]. The obtained sensitivity analysis result does not consider the operating expenses because of their highly correlation. Additionally, we also test the rule of thumb issued by Golany and Roll [58]. The four tests are all hold.

Tables 3 and 4 report the BCC efficiency scores of OE and DE for the 34 OECD countries from 2007 to 2009. Table 3 shows the comparison of the main goal in phase I to evaluate how efficiently countries use their resources; in other words, to identify any inefficiency result from PTE or SE. The resource inefficiency 2007, 2008, and 2009 is primarily pure technical efficiency (0.729, 0.704, and 0.727, respectively). In other words, the inefficiency is a result of inappropriate input and output configuration, rather than inappropriate scale. Table 4 shows a comparison of the main goal in Phase II to evaluate how efficiency energy is used to identify inefficiency resulting from PTE or SE. The resource inefficiency during 2007, 2008, and 2009 is primarily scale efficiency (0.439, 0.431, and 0.45, respectively). In other words, the inefficiency is result of inappropriate scale.


Efficient DMUs


**Table 3.** BCC-efficiency Scores for operating efficiency for each year

We employed the Mann-Whitney U-Test, a non-parameter statistical method, to test the same mean between two groups. The results in Table 4 show that the OE and DE for all cas‐ es do not achieve a level of significance (*p* >.05) for all compared years. Therefore, these 3 years are suitable for the DEA model using 102 DMUs to determine if there is a significant difference between OE and DE (Table 5).

renewable energy development. This implies that the development of renewable energy is crucial. Furthermore, we want to determine if there is a significant difference between OE

Tobit regression analysis was conducted to determine whether the efficiency scores are re‐ lated to characteristics such as GDP, population, capital, endowment and subsidy (Table 7). Furthermore, a dummy variable was included to evaluate the renewable energy subsidies in OECD countries. The function of a regression model can be expressed as:*Y* =*a* + *bX* , where *Y* represents the dependent variable and *X* represents regression form to a logistic probability function because the efficiency ranges from zero to one. The transformed regression func‐

> ln( ) , <sup>1</sup> *<sup>Y</sup> a bX Y*

<sup>1</sup> ( ) . 1 exp( )

The Tobit regression analysis result shows that endowment, population, and capital all have high positive significance with OE and DE. Thus, H1a, H1b, H2a, H2b, H3a, and H3b are supported. However, GDP has a non-significant negative correlation with OE and DE. Thus, H4a and H4B are rejected. GDP, capital, trade balance, household consumption, and energy imports are critical factors for measuring renewable energy indicators [5]. Our finding in H4a and H4b is that the GDP and OE are negatively correlated, and GDP and DE are also negative correlated. This is potentially because countries did not allocate the use of renewa‐ ble energy in accordance with GDP degree. For example, compared to poorer countries, wealthy countries must improve the relatively large number of renewable energy use to ach‐

*a bx*

*Y F a bX*

=+ =

**Case Test Value OE DE OEvsDE**

Comparative Analysis of Endowments Effect Renewable Energy Efficiency Among OECD Countries

Z Test -0.006 -0.98 -2.812 p-vale 0.995 0.327 0.005\*\*

http://dx.doi.org/10.5772/52020

205

Z Test -1.258 -0.5.433 -2.819 p-vale 0.208 0.665 0.005\*\*

= + - (2)

+ -- (3)

and DE.

Between 2008 and 2007

Between 20.50.58 and 20.50.57

**Table 6.** Results of Mann – Whitney U Test of OE and DE

*Note*. Source from this study

tion is expressed as:

derived from:

ieve the target.


**Table 4.** Results of Mann – Whitney U Test


**Table 5.** Bcc-efficiency score for OE and DE for each year

The Mann-Whitney U-Test is also used to determine if there is a significant difference be‐ tween OE and DE before and after 2008 (Table 6). The results show that the global financial crisis did not influence OE and DE. Because OE and DE are non-significant, we can assert that the data are consistent and that renewable energy capital investments in each country have a certain proportion; thus, 2008 financial crisis did not have s significant influence on renewable energy development. This implies that the development of renewable energy is crucial. Furthermore, we want to determine if there is a significant difference between OE and DE.


#### **Table 6.** Results of Mann – Whitney U Test of OE and DE

Tobit regression analysis was conducted to determine whether the efficiency scores are re‐ lated to characteristics such as GDP, population, capital, endowment and subsidy (Table 7). Furthermore, a dummy variable was included to evaluate the renewable energy subsidies in OECD countries. The function of a regression model can be expressed as:*Y* =*a* + *bX* , where *Y* represents the dependent variable and *X* represents regression form to a logistic probability function because the efficiency ranges from zero to one. The transformed regression func‐ tion is expressed as:

$$a\ln(\frac{Y}{1-Y}) = a + bX,\tag{2}$$

derived from:

We employed the Mann-Whitney U-Test, a non-parameter statistical method, to test the same mean between two groups. The results in Table 4 show that the OE and DE for all cas‐ es do not achieve a level of significance (*p* >.05) for all compared years. Therefore, these 3 years are suitable for the DEA model using 102 DMUs to determine if there is a significant

**Phase 1**

**Phase 2**

**Case Test Value TE PTE SE**

**Case Test Value TE PTE SE**

**2007 2008 2009**

Between 2008 and 2007 Z Test -0.98 -0.234 -0.5.057

Between 20.50.58 and 20.50.57 Z Test -0.433 -0.179 -0.345

**DMUs 0E DE OE DE OE DE** Ave. 0.666 0.387 0.5 0.373 0.597 0.367 SD 0.26 0.445 0.266 0.441 0.274 0.407

No. of Efficient DMUs 9 9 6 8 7 7

The Mann-Whitney U-Test is also used to determine if there is a significant difference be‐ tween OE and DE before and after 2008 (Table 6). The results show that the global financial crisis did not influence OE and DE. Because OE and DE are non-significant, we can assert that the data are consistent and that renewable energy capital investments in each country have a certain proportion; thus, 2008 financial crisis did not have s significant influence on

p-vale 0.834 0.71 0.617

p-vale 0.177 0.546 0.077

p-vale 0.327 0.815 0.29

p-vale 0.665 0.858 0.73

Between 2008 and 2007 Z Test -0.21 -0.372 -0.501

Between 20.50.58 and 20.50.57 Z Test -1.349 -0.604 -1.571

difference between OE and DE (Table 5).

204 New Developments in Renewable Energy

*Note*. Source from this study

*Note*. Source from this study

**Table 5.** Bcc-efficiency score for OE and DE for each year

**Table 4.** Results of Mann – Whitney U Test

$$Y = F(a + bX) = \frac{1}{1 + \exp(-a - bx)}.\tag{3}$$

The Tobit regression analysis result shows that endowment, population, and capital all have high positive significance with OE and DE. Thus, H1a, H1b, H2a, H2b, H3a, and H3b are supported. However, GDP has a non-significant negative correlation with OE and DE. Thus, H4a and H4B are rejected. GDP, capital, trade balance, household consumption, and energy imports are critical factors for measuring renewable energy indicators [5]. Our finding in H4a and H4b is that the GDP and OE are negatively correlated, and GDP and DE are also negative correlated. This is potentially because countries did not allocate the use of renewa‐ ble energy in accordance with GDP degree. For example, compared to poorer countries, wealthy countries must improve the relatively large number of renewable energy use to ach‐ ieve the target.

Specially, subsidies are significant with OE but not with DE. In other words, the subsidy is positively correlated with OE but negatively correlated with DE. Thus, H5a is supported and H5b is rejected. Some researchers have reached an opposing conclusion that subsidies and OE are positively correlated and that energy subsidy reform would produce positive re‐ sults. Promoting "subsidy" policies can reduce industrial production costs. However, if they are implemented inefficiently without carefully assessing the cost-efficiency and associated financial risks, a "free rider" phenomenon is created with consequent disadvantages; thus, the subsidies do not have a positive benefit. Therefore, renewable energy subsidies and DE may be negatively correlated [59].

OECD countries in response to United Nations climate Change Framework Convention and the relevant provisions of the Kyoto Protocol, In addition to adjusting the energy supply and demand side policies, and with the greenhouse gas performance of fiscal policy (subsi‐ dy) to promote energy conservation and reduce dioxide emissions [63]. Our second finding is that subsidies are positively correlated with OE and negatively correlated with DE. Pro‐ moting subsidy policies can reduce industrial production cost. However, if they are imple‐ mented inefficiently without carefully assessing the cost-efficiency and associated financial risks, a "free rider" phenomenon is created with consequent disadvantages; thus, the subsi‐ dies do not have a positive benefit. Therefore, renewable energy subsidies and DE may be

Comparative Analysis of Endowments Effect Renewable Energy Efficiency Among OECD Countries

http://dx.doi.org/10.5772/52020

207

In our study, we attempted to measure OECD countries' renewable OE and DE simultane‐ ously, to examine the OECD renewable energy's promote, employ, and the relevance to the development of research, and provide feasible suggestions for a renewable energy develop‐ ment strategy in Taiwan. For example, because Taiwan is an island country, and resources are difficult to obtain, efforts should be made to actively develop renewable energy technol‐ ogy to replace traditional energy sources. In addition, policy-makers should assess renewa‐ ble energy subsidy programs, promote renewable energy industry research and development, assist the industry in developing cost-efficient production technologies, and develop a new energy market. Furthermore, strengthen the use of renewable energy demon‐

Finally, this study has several limitations that require discussion. First, only 34 OECD sam‐ ples were selected that could provide the data required to conduct this study. Future re‐ search could include more countries, especially developing countries, such as Taiwan, China, and India to achieve more precise results. Second, non-financial data such as output quality and investment of renewable land were not included in our model. These variables are also critical factors for the evaluation of energy industry performance. Future research could include this as an additional evaluation variable. Finally, in this study, we independ‐ ently tested and verified the two phases of efficiency. However, future research could use a supply chain model that assumes that the two phases of efficiency are dependent and fur‐

stration and propaganda work, and to enhance the efficiency of the client to use.

and Yi Hsuan Ko1

1 Graduate Institute of International Business, National Taipei University, New Taipei City,

2 Department of International Business, Ming Chuan University, Taipei, Taiwan, R.O.C.

ther evaluate the real scores of management efficiency.

\*Address all correspondence to: chenty@mail.ntpu.edu.tw

negatively correlated [59].

**Author details**

Taiwan, R.O.C.

Tser-Yieth Chen1\*, Tsai-Lien Yeh2


*Note*. Here is the efficiency scores derived from operating efficiency (OE) and density efficiency (DE). The observation is 102. \*\* represents significant at 0.05 level and \* represents significant at 0.1 level.

**Table 7.** Estimated Results of the Tobit Regression Analysis
