Chapter 4 Data description and Empirical result analysis
4.1 Data description
國
立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
20
Chapter 4 Data description and Empirical result analysis
4.1 Data description
Financial instruments like banks and insurance companies hold more financial assets than other different industries, and banks may be greatly affected by the equity volatility caused by the classification of available-for-sale financial assets. But insurance companies focus more on the cash reserve and liquidity so they are more likely than banks to classify the securities as available-for-sale. As a result, the insurance companies provide more opportunities than banks for gains trading. We decide to select the insurance companies as sample to exam whether there exist earning management under the SFAS NO.34 in Taiwan.
We collect the statistic data from Taiwan Insurance Institute database, this institute is founded in 1985. They provide insurance research development and collect actuarial services and statistics. They offer detailed statistic data about the insurance companies and market in Taiwan. Our empirical data is all from their database.
Our sample year period is from 2006 to 2011, and we collect the statistic data from Taiwan insurance institute databases. Before 2006, the insurance companies do not have to disclose their investment securities. And there is no need to classify the financial asset to different categories. As a result, the data before 2006 are not available. This period includes the implementation of SFAS NO.34 in Taiwan. The insurance companies also experienced the financial crisis in 2008. In 2008, because of the financial crisis, many insurance companies had to face many losses from the investment securities. This event also helps us to test whether the insurance companies would engage in the gains trading or some window dressing behavior. This database includes the entire local and foreign insurance companies’ financial data and operating data in Taiwan. There are some missing data in our sample, especially recent years; due to many different cases of mergers and acquisitions in the insurance industry. Besides the enormous change in the insurance companies, there are some new insurance companies whose data are not complete, and we still take these companies in our sample. We try our best to collect the integrated data for the research. For the new insurance companies we deal with the previous data (no statistic
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
21
data) as missing data. We also make some changes about our original data to make it qualify and meet the requirement of our research. The detailed data definition is illustrated in Table 2.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
22 4.2 Data analysis
Table 2 Variable definition
Variable Definition
afs The percentage of available-for-sale financial assets to total financial securities, that is AFS%
ft The percentage of financial assets at fair value through profits and loss (for trading portfolio) to total financial securities. That is FT%
nft This is the percentage of net income divided by the sum offinancial assets at fair value through profits and loss (for trading portfolio) and the available-for-sale financial assets.
roa Return on assets, the percentage of net income to total assets
roe Return on equity, the percentage of net income to total equity
lnassets Natural log of insurance company total assets
la Leverage ratio, the percentage of total liability to total asset currentratio Current ratio, the percentage of total current assets to total
current liabilities
Table 3 Descriptive statistics of each variable
Variable Obs Mean Std. Dev. Min Max
afs 170 60.95615 32.21001 0 100
ft 170 10.93505 16.76 0 88.07954
nft 166 2.400738 5.160776 -18.9014 29.28838
roa 170 -3.16568 11.85272 -97.1659 2.509372
roe 170 -10.624 96.72689 -836.09 552.2149
lnassets 170 18.34015 1.963669 11.28213 21.91031
la 170 0.95702 0.143723 0.088798 1.388885
currentratio 169 16.18775 31.54418 0.657259 328.0623
‧
The insurance industry in Taiwan is changeable, our sample year period cross six year, there are some new insurance companies entering this market and some insurance companies out of this market. So our sample is few than the original expected. The NFT (percentage of net income divided by the sum of financial assets at fair value through profits and loss (for trading portfolio) and the available-for-sale financial assets.) variable data is missing because the new companies do not include in the database of Taiwan insurance institute.
According to Table 3, we find that the percentage of financial assets at fair value through profits and loss (for trading portfolio) to total financial securities, that is, the mean of for-trading(ft)financial assets variable is smaller than the percentage of available-for-sale financial assets(afs). And the standard deviation of for trading financial assets is smaller than the available-for-sale financial assets. In our sample, the insurance companies are inclined to hold more available-for-sale financial assets, and tend not to classify the financial assets into the for-trading portfolio category.
We also find that the variable, nft (percentage of net income on investment securities to the summarization of available-for-sale financial asset and for-trading financial assets) is with low standard deviation, and the range of this variable is also wider, from a minimum of -18.9%, to a maximum of 29.29%, showing very different gains or losses between different companies and time period. Take a look at the ROE ratio. We find that the ROE standard deviation is larger than the ROA standard deviation, and this variable is with wider range. The volatility of ROE is obviously larger than ROA. We also find that the mean of ROA and ROE are also negative. This implies the insurance industry faces a difficult time in Taiwan for past few years. We could also find most insurance companies in Taiwan do not make profit in recent years. Even worse, they still have to face the loss from operating. This may also explain why so many mergers and acquisitions of insurance companies happen in Taiwan recently. Furthermore, although the insurance market in Taiwan is growing and becomes mature and more Taiwan people accept and buy insurance, but it seems that the insurance companies do not benefit from the growing market in Taiwan. This may also give the manager of insurance companies the motivation to engage in the gain trading or window dressing financial report. Finally, the current ratio range in our sample is also wide, from the smallest one 0.657 to the largest one, 328.062. We note the demand of liquidity is very different within different insurance companies and time period.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
24 Table 4 The correlation matrix for each variable
afs ft nft roa roe lnassets la currentratio
afs 1
ft -0.4258* 1
nft -0.4089* -0.0367 1
roa -0.1121 -0.093 0.1895* 1
roe -0.1922 0.1951* 0.1137 0.2258* 1
lnassets -0.3688* -0.0774 0.2409* 0.3979* 0.1128* 1
la -0.3847 0.5235* -0.1206 0.1495* 0.1754* 0.1706* 1
currentratio -0.3334* 0.2452* 0.2092* 0.0017 0.0992 0.1232 0.2833* 1
*Represent significant under p <0.05
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
25
Through Table 4, correlation matrix, we could have a view about the relationship between dependent and independent variables. Because there is no correlation coefficient over 0.5 between each variable, we could conclude there is no highly collinearity problem between different control variables. We only find the natural log of asset is correlated with ROA, but it is normally, because in the calculation of ROA we use the asset as denominator. But it seems does not cause a serious collinearity problem in our model. We still have a good explanation and significant power in all of our models. There are some variables, which have different coefficient sign from our original expectation. We will do further examination in our following analysis.
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
26
Table 5 Regression analysis of the available-for-sale and for-trading classification, and the ratio of net securities investment income to financial assets at fair value through profits and loss (for trading portfolio) and available-for-sale financial assets.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
afs afs afs ft ft ft ft nft nft nft
roa -0.060 -0.172 0.129 0.220** 0.078 0.106* 0.142***
(-0.300) (-1.060) (1.200) (2.000) (0.770) (1.890) (2.690)
roe -0.012 -0.012 -0.013 0.016 0.003 0.015* 0.018* 0.004 0.004 0.006 (-0.720) (-0.750) (-0.840) (1.580) (0.380) (1.710) (1.830) (1.010) (0.990) (1.470)
lnassets -3.864 -4.540 -1.393 -0.446 -0.087 -1.213 0.576* 0.777***
(-0.990) (-1.420) (-1.390) (-0.220) (-0.080) (-1.210) (1.840) (2.720) la 14.831 6.546 15.972 24.193** 11.586 22.321** -16.596*** -16.033*** -14.743***
(0.720) (0.350) (0.790) (2.480) (1.150) (2.390) (-3.610) (-3.430) (-3.280) currentratio -0.087* -0.085* -0.087* 0.050* 0.035 0.043 0.037*** 0.038*** 0.036***
(-1.940) (-1.900) (-1.950) (1.790) (1.390) (1.640) (3.080) (3.170) (2.920) _cons 118.595* 55.213** 130.082** 12.837 19.270 0.629 12.064 7.478 17.624*** 1.819
(1.780) (3.050) (2.380) (0.750) (0.520) (0.040) (0.700) (1.080) (3.870) (0.290) R
20.054 0.039 0.058 0.206 0.020 0.210 0.212 0.160 0.128 0.149
N 169 169 169 169 169 169 170 165 165 165
The values in the parentheses denote the values of t-statistic
*p<0.1, **p<0.05, ***p<0.01
‧
In models (1)~(3), we try two different methods initially. The fixed effect model and random effect model, and then we use the Hausman test as an indicator to decide which model is the better one. After considering the Hausmantest, models (1)~(3) are all suitable for fixed effect model. We use available-for-sale financial asset as dependent variable. Model (1) includes all the independent variables; we find that the current ratio is negatively relative to the available-for-sale financial assets, and it is significant. Note for one thing that we try to identify different variables in our models (1)~(3), and we find the current ratio plays an important role in our model. It implies that the insurance company with lower current ratio will be inclined to classify their investment securities into the available-for-sale category, and this will reduce the volatility of net income, and thus reduces the demand of liquidity and releases the intense concern from investors and shareholders. On the other hand, they have lower demand for the current assets, so they are not afraid that it will reduce the liquidity if they classify the investment securities to the available-for-sale financial asset category. It also could be explained by the fact that the insurance companies have more tolerance on the net income volatility. They have more ability to meet the short-term liability, and don't have to engage in window dressing for their financial report.
In models (4)~(7), we use the financial assets at fair value through profits and loss (for trading portfolio) as dependent variable. We try to take away different variables in the model. We use random effect model for these models except for the model (5) according to the Hausman test and R-square. In model (4), we put all the independent variables into the model, and we find that leverage and current ratio is significantly positively related to for-trading financial assets. The insurance companies withhigher leverage will tend to classify the investment securities to for-trading category. This may be caused by the demand of liquidity for these insurance companies. The insurance companies with higher leverage will need more liquidity to decrease financial risk. Additionally, we could also find the insurance companies with higher current ratio also have higher for-trading financial assets. And it represents the insurance companies with higher liquidity demand will be inclined to hold more for-trading financial assets. In model (5), we find that the insurance companies with higher ROA will tend to classify the financial assets into the for-trading financial assets category. The ROA is positively related with the for-trading financial assets significantly. It matches our previous expectation. We expect that an insurance
‧
company with lower ROA will classify more investment securities to the available-for-sale category and less to the for-trading financial asset category. We conclude that when the companies are with higher ROA, they may classify more investment securities into the for-trading financial asset category in order to have more profits and take more risk. They also have higher tolerance for the volatility of gains or loss of investment securities. We also find that the companies with higher ROE will tend to classify the investment securities to the for-trading financial assets in model (6) and (7). This result shows the ROA is positively relative to the for-trading financial assets, which meets our first expectation. Although the coefficient of ROA and ROE is not significant in all models (1)~(7), but they fit our previous expectation as well.
In model (7), we also find that leverage is significantly and positively related to the for-trading financial assets. This is because the higher leverage insurance companies want to maintain their liquidity; the companies with higher leverage will have more financial risk and demand of liquidity than the companies with lower leverage. The classification of for-trading financial assets offers more liquidity for the high leverage insurance companies. As a result, they will have more incentives to classify the financial assets to for-trading category. Furthermore, if they classify the investment securities to the available-for-sale category, they need to face the intense concerns from the financial reporting auditors and the shareholders. They have to offer strong evidence if they want to reclassify the investment securities, so as we talk previously, this may make the companies inclined to classify the investment securities to the for-trading category instead of available-for-sale category.
In contrast with the models (1)~(3), from models (4)~(7), we only find in model (4) the current ratio is significant, and the insurance companies with higher current ratio will tend to classify the investment securities to the for-trading financial asset category. That is, the current ratio is positively related to the for-trading financial assets. The companies with higher current ratio represent they have higher demand for the current assets to meet the need of short-term operation funding and to pay the current liability. As a result, they are inclined to classify the investment securities to the for-trading category instead of available-for-sale financial assets category. In models (5)~(7), the current ratio is not a significant variable for the classification of for-trading category.
In models (5)~(7), The asset size of insurance companies dose not have significant effect on the for-trading financial assets. The leverage ratio and ROE and
‧
ROA will indeed affect the classification of financial assets for insurance companies.
But note that in model (5), the R-square is obviously lower than the other three models.
For models (8)~(10), we consider different dependent variables. We use the percentage of net income to financial assets at fair value through profits and loss (for trading portfolio) plus the available-for-sale financial assets. We try to test how the ROA and ROA will influence the insurance companies to engage in the gains trading.
Furthermore, we also want to explore the relationship between other different variables and the behavior of gains trading. We use the Hausman test as an indicator and still consider the R-square of models. We use the random effect model for these three.
Models (8)~(10) show that the entity size significantly affects the insurance companies concerning engaging in the gains trading. The larger insurance companies have more intention to engage in the gains trading significantly. The same result is obtained by 黃劭彥等 (2011). We conclude that the companies with larger asset size have to face more intense pressure to achieve earning level. Therefore, the managers of insurance company with larger asset size have more intention to engage in the gains trading in order to meet the earning level expectations from investors and shareholders than the managers of insurance companies with smaller asset size.
According to these three models, we also find that the current ratio has significant influence on the gains trading behavior for insurance companies. The insurance companies with higher current ratio will be more inclined to engage in gains trading, and note the leverage ratio positively relates to the gains trading behavior for insurance company. This result is the same as 黃劭彥等 (2011). 黃劭彥 (2011) uses the logistic regression and finds that the current ratio is significantly positively related to gains trading. We conclude the insurance companies with higher current ratio will have more demand of liquidity, and they will do more gains trading through manipulating the securities investment. By the gains trading behavior, the insurance companies will have higher current ratio or more liquidity to meet their short-term debt. From these three models, we also find the leverage ratio plays an important role for insurance companies to engage in the gains trading. The leverage ratio is negatively related with the gains trading. It implies the insurance companies with higher leverage will not engage in the gains trading. This result matches the finding of
‧
Jordan et al. (2011a). As the research mentions, most companies earn a higher returns on their asset than they have to pay for finance these assets. Using debt in the presence of positive financial leverage further enhances companies’ earning level.
They will use the earnings from the gains trading to offset the loss of earnings they may experience from the financial leverage which they originally want to generate income. The entities with a large amount of debt may have less need to manage its earnings through gains trading because their earnings are already boosted by the excess return resulting from the positive financial leverage. Our research has proved that in the insurance industry exact exists this phenomenon. Finally, we have a different finding about ROA variable that influences the gains trading behavior.
According to Jordan et al. (2011a), they find ROA significantly negatively related to the gains trading behavior in the insurance industry. They conclude that firms with lower level of pre-gain earnings are more likely to engage in gains trading. In contrast to their finding, our research shows that ROA is positively related with the gains trading behavior. This may be due that the insurance companies use the gains trading action to improve their ROA, and further window dress their financial report. The insurance companies with higher ROA may have to further meet the demand from investors or shareholders of the companies. So they will be more inclined to engage in the gains trading for achieving the target. In our speculation, we find the mean of insurance industry ROA in Taiwan is negative, and we all know that in Taiwan, the insurance companies with a negative net income is more than the insurance companies with positive net income. This also reflects the difficult operating environment for the insurance companies in Taiwan. It may increase the intension for managers of insurance company to engage in gains trading. But in our sample, we don't find the negative relationship between the gains trading and ROA.
Over all, from this research, we find the current ratio plays an important role when the insurance companies make decision about classifying the financial assets significantly. Compared with ROE, it seems that ROA is a factor that will cause insurance companies to classify the financial assets significantly. Insurance companies with higher ROA and ROE will tend to classify the financial assets to for-trading financial asset category. We could conclude that the ROE is not significant in all of our models. But we still could conclude that ROE is a good indicator to measure the gains trading for insurance companies. We also find that the entity size will affect
‧
國立 政 治 大 學
‧
N a tio na
l C h engchi U ni ve rs it y
31
the gains trading behavior. The larger insurance companies are, the more inclined to engage in the gains trading behavior by having more securities investment.
‧
Prior researches (Jordan 2011a, Jordan 2011b) have shown gains trading occurred in the insurance industry. They further explore the influence of implementation of different financial accounting reporting standards and provide the evidence of earning management and gains trading behavior in the insurance industry.
The statement of financial accounting standard in Taiwan still has to face the same problem. The government agency has to prevent the managers of insurance companies from earning management and gains trading behavior. Especially after the
The statement of financial accounting standard in Taiwan still has to face the same problem. The government agency has to prevent the managers of insurance companies from earning management and gains trading behavior. Especially after the