Unconditional Accounting Conservatism
& Efficient Contracting
Authors: Joel Björklund & Antor Ullah Supervisor: Niuosha Samani
Master Degree Project in Accounting Spring 2018
Graduate School
Abstract
We examine how accounting conservatism affects contracting efficiency. More specifically, we examine the link between unconditional conservatism and accounting sensitivity in executive compensation contracts where we predict a negative relationship. We look at theoretical arguments of unconditional conservatism and how it affects gains recognition, noise in payoffs, and management possibilities for earnings management. We also consider the inverse relationship between conditional and unconditional conservatism. We conduct our study on Swedish listed firms between the years 2006 - 2016 by using an accrual based and a cash-flow to earnings-based proxy to measure unconditional conservatism. We find an inverse relationship suggesting that an increased degree of unconditional conservatism decrease contracting efficiency and accounting sensitivity in executive compensation contracts.
Keywords: Unconditional conservatism, Contracting efficiency, Accounting sensitivity
Table of Contents
1. Introduction ... 1
2. Literature and hypothesis... 3
2.1 Accounting conservatism and contracting ... 3
2.2 Hypothesis ... 6
3. Research design ... 8
3.1 Measuring Unconditional Conservatism ... 8
3.2 Model of Executive Compensation Sensitivity ... 9
4. Results ...11
4.1 Sample ... 11
4.2 Descriptives, Correlations & Regressions ... 12
4.3 Robustness test ... 14
5. Conclusion ...16
5.1 Limitations & Further research ... 17
References ...18
1. Introduction
Prior research has examined the benefits of conditional conservatism on contracting (Ahmed et al, 2002; Iyengar & Zampelli, 2010; O'Connell, 2006; Watts, 2003). However, studies examining the links between unconditional conservatism and contracting is absent. We build on the existing literature by adding the perspective of unconditional conservatism, by examining the financial statement effects of unconditional conservatism and its consequences on contracting parties.
Accounting conservatism is defined by policies and accounting actions taken by management that results in a lower book value of net assets compared to economic value. (Ruch & Taylor, 2015). Furthermore, the majority of research conducted within conservatism defines conservatism as the asymmetry in recognizing gains and losses where losses are recognized in a timely manner whilst gains recognition is deferred (Basu, 1997), this is called conditional conservatism. However, there is another type of conservatism that is called unconditional conservatism which is defined as the systematic way of undervaluing net assets and earnings (Beaver & Ryan, 2005).
The different types of conservatism are materialized differently, an example of conditional conservatism is impairments and lower of cost and market value of inventories and is of a more transitory nature. Examples of unconditional conservatism is the systematic process of immediate expensing of R&D costs and lower asset lifetime (Beaver & Ryan, 2005). Both forms of conservatism result in lower net assets and lower earnings, but in different ways and in different degrees. Unconditional conservatism will value an asset lower initially by either a lower value or a shorter asset lifetime, for instance, expensing more of the development costs instead of capitalizing and accelerated depreciation and amortization. Conditional conservatism on the other hand will not value the asset lower initially but will be more careful when it comes to “bad news” and are thereby more prone to impairing the asset later on (Beaver
& Ryan, 2005).
Watts (2003) argues that conservative accounting is beneficial when it comes to contracting due to asymmetrical recognition in gains and losses. Contracting is a process that aims to align the interests of different parties to the firm. Considering agency costs arises when different parties to the firm have divergent interests, firm value decreases due to different parties maximizing their own wealth (Watts, 2003). More specifically, issues arise when organizations are subject to separation of ownership and control. Jensen & Meckling (1976) states that agency costs arises when owners (principals) and managers (agents) have different incentives. In order to align these interests, executive compensation contracts to managers are based on firm performance, derived from both accounting and market measures (Holmström, 1979).
Additionally, Watts (2003) points out that it is the conditional type of conservatism that is beneficial for contracting. Through conditional conservatism, losses are recognized earlier in the financial statement compared to gains. From bondholder’s standpoint, debt covenant violations are thereby triggered earlier and the risks for bondholders minimizes (Watts, 2003).
Moreover, conditional conservatism lowers net assets and earnings which reduces the possibilities of excessive dividend policies and opportunistic management payouts, resulting in lower possibilities of wealth transferring from bondholders to shareholders and managers.
This has been observed by Zhang (2008) and Ahmed et al (2002) where both detected that conditional conservatism was negatively associated with debt cost of capital and that conditional conservatism mitigates bondholder-shareholder conflicts.
Contracting benefits further applies to shareholders and managers in executive compensation contracts. Given that firm performance dictates executive compensation and information about firm performance is gathered from financial statements, executives have incentives to bias the accounting upwards by overstating earnings and net assets. Conditional accounting conservatism mitigates this by constraining the recognition of gains and enabling a timelier recognition of losses (Watts, 2003). Additionally, conditional conservatism enhances firm performance through the asymmetrical recognition of gains and losses, resulting in earlier dismissal of negative net-present value projects (Francis & Martin, 2010; Ball & Shivakumar, 2005).
Since conditional conservative accounting results in better contracting efficiency, it is expected that shareholders rely more on accounting-based information rather than market-based information in compensation contracts to managers. This results in shareholders emphasising accounting based performance measures over market-based performance measures, and thereby increase the accounting sensitivity in executive compensation contracts (Iyengar &
Zampelli, 2010).
Similar to conditional conservatism, unconditional conservatism lowers net assets and earnings, but it does not increase contracting efficiency (Qiang, 2007). Qiang (2007) argues that unconditional conservatism does not increase contracting efficiency due to noise and the lack of usage of new information. Researchers argue instead that unconditional conservatism will lower contracting efficiency (Ball & Shivakumar, 2005; Beaver & Ryan, 2005; Qiang, 2007). Ball and Shivakumar (2005) implies that unconditional conservatism has a negative effect or no effect at all on contracting efficiency. Their arguments are backed up by Qiang (2007) and Beaver and Ryan (2005) who argues that the application of unconditional conservatism will lower the possibilities of using conditional conservatism later on. The explanation for this is that unconditional conservatism affects conditional conservatism negatively, which is why it should affect contracting efficiency negatively as well.
Furthermore, Penman and Zhang (2002) and Jackson and Liu (2010) illustrates that unconditional conservatism creates reserves in the balance sheet which can be released into earnings. This increases accounting flexibility which subsequently increases the possibilities of earnings management which lowers contracting efficiency.
Prior research has mainly focused on the beneficial properties of conditional conservatism and we see a lack of studies regarding contracting and unconditional conservatism. Studies that have examined unconditional conservatism and contracting have only discussed the indirect relationship between unconditional conservatism and contracting through its inverse relationship with conditional conservatism. We therefore aim to investigate the direct links
between unconditional conservatism and contracting. Because of the negative relationship between unconditional and conditional conservatism and the effects of unconditional conservatism on the financial statement. We hypothesize that unconditional conservatism has a negative relationship with contracting efficiency. More precisely, we will look at accounting sensitivity in executive compensation contracts when examining contracting efficiency.
Our sample consists of Swedish listed companies from the years 2006-2016 with a total of 1613 firm years. We use panel data and the data is collected from annual reports and WRDS Compustat database. Furthermore, we proxy unconditional conservatism with two measures developed by Givoly and Hayn (2000). The first measure accounts for the non-operative accruals measured by total accruals less operational accruals. The second measure is a time- series measure of skewness between cash flows and earnings. Moreover, we measure executive compensation sensitivity by a model developed by Iyengar and Zampelli (2010) that accounts for total compensation as a function of accounting and market-based performance. The unconditional conservatism measures are integrated into one model in order to examine the effects of unconditional conservatism on accounting and market performance and its subsequent effect on total compensation.
In accordance with our hypothesis, we find a negative association between unconditional conservatism and accounting sensitivity in executive compensation contracts. We highlight the financial statement effects of unconditional conservatism through discussing gains recognition, possibilities of earnings management, and noise in payoffs. We thereby contribute to the existing literature by illustrating a direct relationship between unconditional conservatism and contracting.
The remainder of this paper is organized as follows. Section 2 reviews earlier research and theories whereby we build our hypothesis and predict our results. Section 3 presents the chosen research design. Section 4 describes the sample, our main results, and the robustness of our tests. Finally, in section 5, we conclude our study, discuss limitations and present suggestions for future research.
2. Literature and hypothesis
2.1 Accounting conservatism and contracting
By studying organizations, it is observed that problems arise when managers and owners are not the same actor. In such cases, the shareholders (principal) does not manage the firm, instead they hire a manager (agent) (Jensen & Meckling, 1976). The reason why this is problematic is because rational actors aims to maximize their own wealth instead of the firms. Managers would want to maximize their wealth, shareholder theirs, and thereby there is a misalignment of interests which will decrease firm value (Watts, 2003).
This decrease in wealth is called agency costs and firms apply different methods to decrease agency costs. Efficient contracting is one solution where the aim is to align the interests of
bondholders, shareholders, and managers (Watts, 2003). Additionally, efficient contracting formulates compensation contracts to managers based on performance, higher wealth to the firm results in higher wealth to managers. According to Sloan (1993), this provides managers a proper compensation based on what they actually perform, however this also requires that shareholders are able to evaluate the performance of managers. Holmström (1979) states that managers are evaluated by the payoffs from the actions they take. He further argues that payoffs are a function of the managers actions and a random error that the manager cannot control. For example, stock prices are partly dictated by the actions that the CEO takes, and partly dictated by outcomes that the CEO cannot control such as investor speculations, market factors, and other macro-circumstances. This is called the random error and is illustrated as noise. Payoffs is evaluated from both market based and accounting based measures, and it is the random error that dictates the sensitivity and emphasis between accounting respectively market-based measures in the compensation contracts (Holmström, 1979)
Holmströms (1979) theorizes that a better accounting measure with higher reliability and verifiability results in less noise and thereby, a higher focus on accounting measures in compensation contracts. From the arguments made by Watts (2003) about the beneficial properties of conservative accounting, it is both expected and illustrated that conservative accounting results in a higher focus on accounting measures in executive compensation contracts. Furthermore, the role of accounting serves the stewardship role. As demonstrated by Gjesdahl (1981), the stewardship role of accounting aim to aid shareholders in controlling managers through performance evaluation and is thereby a mechanism within contracting.
Qualitative and informative accounting is thereby beneficial in contracting purposes.
Iyengar and Zampelli (2010) studied conditional conservatism and accounting sensitivity in compensation contracts and found that executive compensation is more closely tied to accounting performance in firms that use conditional conservative accounting methods.
Further, they argue for the importance of conditional conservatism in aligning management and shareholder incentives and constraining managerial opportunistic behaviour. In addition, O'Connell (2006) finds similar results where he finds evidence that accounting sensitivity in executive compensation contracts is higher in “good news firm years” compared to “bad news firm years” where “bad” or “good” news was proxied by positive or negative adjusted market returns. Moreover, several researchers have studied corporate governance factors and found evidence for that conditional conservatism lowers agency costs (Ahmed & Duellman, 2007;
Lara et al, 2009).
Additionally, problems do not only arise between shareholders and managers but also between shareholders and bondholders. Ahmed et al (2002) examined bondholder-shareholder conflicts and concluded that conditional conservatism mitigates these conflicts through limiting earnings and retained earnings, resulting in limited possibilities for excessive dividend payouts. The consequence of this was decreased cost of debt capital. As excessive dividend payouts transfers wealth from bondholders to shareholders, bondholders are exposed to higher risk of losing their lent capital, especially in highly leveraged firms. In addition, due to the timely loss recognition other positive aspects of conditional conservatism is the earlier recognition of debt covenant
violations and the decreased probability of engaging in negative net present value projects (Ball
& Shivakumar, 2005; Watts, 2003).
Watts (2003) argues for three attributes of accounting measures in contracting purposes, namely; timeliness, verifiability, and asymmetrical verifiability. Contracting parties demand timeliness since timely accounting information is more effective in contracting purposes. For instance, it is preferred that managers actions are reflected in the financial statement at the same period as the action was taken. Problem arises when we include the second attribute, verifiability. Measures that are timely are less verifiable. For instance, data of future cash flows from product development is timely information, however that information is not verifiable because it is based on assumptions. The importance of verifiability is due to legal contracting liabilities where a legal liability ties the contracting parties to their actions (Watts, 2003).
In addition, verifiability needs to be asymmetrical because different contracting parties have different payoffs from the contracts. Debt holders can only claim what they have lent to the firm, meaning that an increase in net assets does not increase the amount debt holders can claim. This does not apply the other way around, debt holders risk to lose their lent capital when net assets decrease which is why it is important that losses are timely recognized compared to gains. (Watts, 2003)
In executive compensation contracts, conditional accounting conservatism and asymmetrical verifiability is important due to the fact that managers have more information than shareholders, auditors, and other parties to the firm. With the information advantage managers have, they can bias the accounting in their favour, resulting in excessive payments to managers under performance-based payment plans. In addition, managers tend to achieve short term goals rather than long term goals due to limited tenure. Deferred gains recognition is from a manager’s standpoint, not beneficial since it will understate managers short term performance.
This in turn increases the managers unwillingness to invest in projects with long term payoffs (Watts, 2003). A large role in the biased incentives is the managers limited liability and tenure, Watts (2003) argues that due to the short-term mindset, managers have incentives to accounting in an opportunistic fashion, which conditional conservatism mitigates.
The theoretical link between contracting and conditional conservatism is clear where researchers mainly argue for the benefits of this association (Watts, 2003; Qiang, 2007; Ball &
Shivakumar, 2005). Unconditional conservatism on the other hand is efficient in other aspects.
Firms often apply unconditional conservatism to reduce taxation costs (Basu, 2005).
Shackelford and Shevlin (2001) argues that tax lowering strategies requires lowering of book earnings since firms are concerned with book-tax conformity, i.e large differences in accounting earnings and taxable earnings. Qiang (2007) further mentions that losses from unconditional conservatism is generally more deductible compared to losses from conditional conservatism. Losses from conditional conservatism is characterized by unrealized losses due to decrease in market values rather than losses from transactions. In addition, unconditional conservatism is beneficial in several other aspects. Unconditional conservatism is easier to control because it occurs prior to asset recognition, costs less to apply since it does not require impairment testing and results in smoother earnings. Furthermore, unconditional conservatism
is also favourable in litigation cost situations. The probability of legal disputes is higher for firms with overstated assets compared to firms with understated assets. Initially, unconditional conservatism results in lower value in assets1. It is therefore more favorable than conditional conservatism in reducing litigation costs (Qiang, 2007).
2.2 Hypothesis
As demonstrated, Watts (2003) argues for the benefits of conditional conservatism on contracting through the beneficial properties of timeliness, verifiability and, asymmetrical verifiability. Unconditional conservatism does not have the characteristics of timely loss recognition and deferred gains recognition. As a result, it does not contain the beneficial properties of timeliness, verifiability and asymmetrical verifiability (Beaver & Ryan, 2005).
Furthermore, Beaver and Ryan (2005) outline that the two different types of conservatism are studied in different situations, conditional conservatism being studied in contracting situations whilst unconditional conservatism being studied in situations regarding difficulties valuing certain assets and liabilities and its effects on future earnings.
By comparing the definitions of unconditional and conditional conservatism (Basu, 1997;
Beaver & Ryan, 2005; Watts, 2003) we see that both results in higher costs due to timely recognition of losses, immediate expensing, and accelerated depreciation. However, conditional conservatism results in lower revenues compared to unconditional conservatism.
Revenues are recognized in a deferred matter in conditional conservatism, which does not have to be the case in unconditional conservatism (Beaver & Ryan, 2005). This implies that the different types of conservatism affect earnings differently where unconditional conservatism creates higher earnings compared to conditional conservatism. By assuming that both kinds of conservatism lower assets to the same degree, financial measures such as return on assets will be higher when unconditional conservatism is used. Financial performance is thereby overvalued, this further affects compensation contracts due to the fact that managers are paid for the performance of the firm, and in this case, they will be paid for a performance that is better than their actual performance. By concluding our first argument, unconditional conservatism can actually increase opportunistic accounting and behavior from manager's standpoint and is thereby the complete opposite of conditional conservatism and can thereby decrease contracting efficiency.
Regarding additional financial statement effects of unconditional conservatism. Our second argument is based on the increased possibilities of earnings management and decreased earnings quality from unconditional conservatism. Penman and Zhang (2002) studied how investments and conservative accounting affects earnings and concludes that a higher degree of unconditional conservatism leads to lower earnings quality. They argue that the use of unconditional conservatism in combination with investments lowers earnings and thereby creates unrecorded reserves which can be released onto earnings in later periods, giving management flexibility to manage earnings. A similar study was done by Jackson and Liu
1 By expensing more and thereby understating assets from the beginning rather than capitalizing the whole amount and impairing later on.
(2010) where they examined the interrelation between earnings management and unconditional conservatism. Similar to Penman and Zhang (2002), Jackson and Liu (2010) found evidence that unconditional conservatism facilitates earnings management through accruals and in this case, by accruing bad debt expenses. They further argue from a more holistic standpoint that unconditional conservatism results in lower earnings and net assets which results in increased reserves. This results in accounting flexibility where the reserves can be released into earnings in later periods (Jackson & Liu, 2010). Consequently, both Penman and Zhang (2002) and Jackson and Liu (2010) have found evidence for higher probability of earnings management and lower earnings quality through unconditional conservatism which affects contracting negatively through decreased accounting quality and increased noise in accounting measures.
Thirdly, we argue that noise in payoffs will alter the sensitivity in compensation contracts where shareholders will focus more on market performance rather than accounting performance. Ball and Shivakumar (2005) argues that if it is known that the firm applies unconditional conservatism, related parties to the firm would simply reverse the accounting bias that comes with unconditional conservative accounting. This was hypothesized and tested by Chen et al (2013) who found that pricing multiples was higher for firms applying conservative accounting measures where the pricing multiple was higher for firms using unconditional conservatism compared to firms using conditional conservatism. Ball and Shivakumar (2005) continues by stating that the reversal of the biased accounting can lead to randomness of decisions based on accounting information if the amount of said bias is not known to all parties. This will according to Ball and Shivakumar (2005) bring noise into payoffs which results in lower contracting efficiency. Similarly, Holmström (1979) states that a higher degree of noise in accounting measure will decrease the accounting sensitivity in executive compensation contracts. Shareholders will instead put more focus on market performance when formulating executive compensation contracts which results in decreased contracting efficiency.
Researchers argue that there is an inverse relationship between conditional and unconditional conservatism (Ball & Shivakumar, 2005; Beaver & Ryan, 2005; Qiang, 2007). By applying unconditional conservatism, you preempt the possibilities of applying conditional conservatism. Because assets are already understated when accounting in an unconditional conservative fashion through low valuation or lower asset lifetime, the possibilities of writing down that subsequent asset later on decreases for the reason that it is already understated (Ball
& Shivakumar, 2005; Beaver & Ryan, 2005). We have mentioned that conditional conservatism is beneficial in contracting purposes (Ball & Shivakumar, 2005; Watts, 2003), and since unconditional conservatism has an inverse relationship with conditional conservatism, it is believed that unconditional conservatism will have the opposite effect. We thereby conclude our last argument by stating that unconditional conservatism is disadvantageous in contracting purposes through limiting conditional conservatism.
In addition, Roychowdhury and Watts (2007) stand in opposition to the argument about an inverse relationship between conditional and unconditional conservatism. They found that when proxying conditional conservatism with Basu’s (1997) measures of asymmetric timeliness and unconditional conservatism with market-to-book value the relationship was
positive over a cumulative period of time and negative when examining the effect on an annual basis. Roychowdhury and Watts (2007) argued that the beginning of period equity value was responsible for the negative effect when measuring conservatism on shorter periods of time and that conservatism is measured more accurately when measured over long periods. How well market-to-book value measures unconditional conservatism is debatable. As mentioned earlier, Chen et al (2013) found that pricing multiples for firms using unconditional conservatism was higher than firms using conditional conservatism. This is evidence for that market-to-book value works fairly well in proxying for unconditional conservatism. However, Chen et al (2013) further found that pricing multiples were higher for conservative firms compared to non-conservative firms, suggesting that market-to-book value fails in measuring unconditional conservatism exclusively. Thereby, unconditional conservatism is still believed to have an inverse relationship with conditional conservatism.
In conclusion, we see that unconditional conservatism should have negative effects on contracting and accounting sensitivity in compensation contracts. We have four main arguments. The first one is the definition of unconditional conservatism where gains are recognized timely instead of deferred as in conditional conservatism. This results in overstated return on assets which affects performance evaluation of managers. The second argument states that unconditional conservatism increases the possibilities of earnings management and decreases earnings quality (Jackson & Liu, 2010; Penman & Zhang, 2002). Thirdly, we argue for increased noise in payoffs through randomness of decisions based on financial information.
This happens when related parties reverse the biased accounting from unconditional accounting conservatism when the biased amount is not known (Ball & Shivakumar, 2005; Holmström, 1979). Lastly, our fourth argument is based on the relationship between unconditional and conditional conservatism where the inverse relationship results in a negative association between unconditional conservatism and contracting (Ball & Shivakumar, 2005; Beaver &
Ryan, 2005; Qiang, 2007). Based on these four arguments, we hypothesize that we will find a negative relationship between unconditional conservatism and accounting sensitivity in executive compensation contracts.
H1: Unconditional conservatism has a negative relationship with accounting sensitivity in executive compensation contracts.
3. Research design
3.1 Measuring Unconditional Conservatism
We proxy unconditional conservatism in accordance with the measures developed by Givoly and Hayn (2000). Givoly and Hayn (2000) observed a cumulative, increasing difference between accounting performance (proxied as return on assets) and real economic performance (proxied as cash flows from operations). They argued that this is due to conservatism where they observed an increase in total accruals.
By breaking down accruals, Givoly and Hayn (2000) observed that there were major differences in the time-series properties in the different accrual measures. Over the period, operating accruals stemming from working capital increased while non-operating accruals decreased. Total accruals also decreased within this period, but not to the same extent as non- operating accruals. Givoly and Hayn (2000) argue that the accumulation of negative non- operating accruals is consistent with an increase in accounting conservatism, which is why, based on Givoly and Hayn (2000) and Beatty et al (2008), we proxy unconditional conservatism as negative non-operating accruals.
UNCON1= (Net Income + Depreciation + Amortization - Operating Cash Flow + ∆Accounts Receivables + ∆Inventories + ∆Accounts Payables + ∆Tax Payables) * -1
Furthermore, we add the negative 1 multiplier for interpretation purposes, with increasing non- operating accruals, we have increasing conservatism as well. We further deflate UNCON1 with total assets in accordance with Beatty et al (2008) and Chen et al (2013).
Givoly and Hayn (2000) argue that since conservatism constitutes that unfavorable events such as losses are recognized fully and favorable events such as gains is recognized gradually due to the asymmetric verifiability criterion, earnings distribution should be skewed. We therefore conduct a second measure of conservatism, accordingly with Beatty et al (2008) and Givoly and Hayn (2000) where we divide cash flow from operations and earnings with total assets, respectively, and then subtract earnings from cash flows. As examined by Givoly and Hayn (2000), a major evidence for conservatism was the difference between reported earnings and real economic performance proxied by cash flows. Cash flows are not affected by accruals, which is why this skewness measure is used as a proxy for unconditional conservatism.
UNCON2 = Earnings skewness = (Cash flow from operations/ Total assets) - (Earnings/Total assets)
From this measure a larger skewness is consistent with more unconditional accounting conservatism.
There are difficulties in measuring unconditional and conditional conservatism separately.
Givoly and Hayn (2000) states that it is difficult to proxy for immediate expensing and accelerated depreciation rates. Furthermore, the cash flow to earnings skewness measures captures parts of conditional conservatism as well. Other examples of similar measures are market-to-book value. As said before, market-to-book value captures both conditional and unconditional conservatism (Chen et al, 2013). Market-to-book value and cash flow to earnings skewness measures are based on the idea that there is a difference between accounting value and real economic value which causes the difficulties in fully excluding conditional conservatism in the proxies.
3.2 Model of Executive Compensation Sensitivity
The model for executive compensation that we have applied is based upon the model used by Iyengar and Zampelli (2010). As stated above by Holmström (1979), executive compensation
contracts are a function of payoffs from measures using accounting and market performance where the emphasis between the two performance measures are dictated by the amount of noise. Higher noise in one variable will alter the focus to the other measure. This provides us the following model:
lnCOMPit = b0+b1ROAit+b2ARit+eit
where lnCOMPit is the natural log of total compensation where total compensation is the sum of salary and bonus. Iyengar and Zampelli (2010) argues that the logged version of executive compensation helps in minimizing the effect of outliers and thereby aids in a more normally distributed data. ROAit is return on assets, the accounting performance calculated by net earnings divided by total assets. ARit is annual return, the variable representing market performance. ARit is calculated by end of year share price less beginning of year share price divided by beginning of year share price.
In order to incorporate unconditional conservatism into our model, we need to add two variables in which we multiply one of our conservatism measure with accounting and market performance respectively. This will enable us to observe the effect that conservatism has on our performance variables. This provides us the following model
lnCOMPit = b0+b1ROAit+b2ARit+b3UNCONj+b4(UNCONj×ROAit) + b5(UNCONj×∆ARit) + b6SIZE+eit
where UNCONj is either UNCON1 or UNCON2.
Prior research reveals that executive compensation is correlated with firm size where Tosi et al (2000) concluded in their meta study that firm size explained 40% of the variation in executive compensation. We therefore include firm size as a control variable measured by natural log of total assets in our analysis. In addition, we control for industry measured by a one-digit dummy variable based on SIC-codes. Furthermore, we include the unconditional conservatism variable separately to examine its association with compensation directly.
By Examining the model, b1 and b2 represents accounting and market performance respectively. The executive compensation is dictated by both accounting and market performance where both affects total compensation positively (Holmström, 1979) We thereby expect positive association between b1, b2 and total compensation.
We hypothesized that unconditional conservatism should have a negative relationship with accounting sensitivity in executive compensation contracts and thereby lowering contracting efficiency. Moreover, we argue that executive compensation will be less sensitive to accounting performance with increasing unconditional conservatism. We can thereby expect that b4 will be negatively associated with our dependent variable. Our fifth measure, b5, is expected to have zero or a positive association to executive compensation. We do not expect to see any association between compensation and market performance when conservatism increases or decrease, however considering Holmström (1979), we could expect a positive
association as a result of lower accounting sensitivity which results in more emphasis on market performance in executive compensation contracts.
4. Results
4.1 Sample
The sample collection begins with a list of executive compensation data for all Swedish firms listed on Nasdaq OMX Stockholm during the years 2006-2014. We expand this list to include executive compensation for the years 2015 and 2016 by manually extracting data from annual reports. From this point we download financial and market data from Compustat, at this point the sample is 2701 firm years for the years 2006-2016. From this sample we delete 862 firm years due to insufficient financial data to calculate our unconditional conservatism and accounting performance measures. Furthermore, we eliminate 76 firm years due to missing market data for our annual return measures and lastly, we exclude 150 firm years due to lacking executive compensation data. Our final sample is 1613 firm years for 206 different firms and the whole sample process is shown in table 2. Moreover, most studies within accounting conservatism has been focused on US companies with only a few studies within the European
Table 1. Definition of variables
lnComp Natural log of total compensation which is the sum of salary and bonus for CEOs
ROA Return on assets, calculated as net income divided by total assets
AR
Annual stock return, calculated by ending year stock price - beginning of year stock price divided by beginning of year value
UNCON1
First measure of conservatism which is cumulative negative non-operating accruals calculated by (total accruals - operating accruals)*-1 deflated by total assets
UNCON2
Second measure of conservatism, calculated as the difference between operational cash-flow divided by total assets and earnings divided by total assets
lnTA Natural log of total assets as a proxy for SIZE SIC One digit industry dummy based on SIC codes
context and none with an exclusively Swedish sample. We thereby add further insight to the European and Swedish literature on accounting conservatism with our study.
4.2 Descriptives, Correlations & Regressions
Table 3 illustrates the descriptive statistics for our main variables. We present statistics for mean, standard deviation, min, max, and values for the quartiles. The means on our two unconditional conservatism measures are similar to those that have been found in earlier studies such as Beatty et al (2008) and Chen et al (2013). One difference from our data is that both the mean and median are positive which was not the case for Chen et al (2013). Furthermore, ROA and AR have been winsored at the 99th and 1th percentile to reduce the adverse effect of outliers. This has been done by several researchers in this field before (Ahmed & Duellman, 2007; Chen et al, 2013; Iyengar & Zampelli, 2010; Lara et al, 2016).
Moreover, from table 3 we can see the mean annual increase in annual return is 12,3 percent which is similar to the figure that O'Connell (2006) found in their study. lnTA is the natural logarithm of total assets which is our control variable for size.
Table 2. Sample selection process
Selection of sample Firm years
Nmber of firm year observations form
compustat for the given sample period 2701 Less: Firm years
Due to insufficient financial from
Compustat 862
Due to insufficient market data from
securities daily 76
Due to insufficient executive compensation
data 150
Final sample 1613
Table 3. Descriptives
Variable n Mean Std. Dev Min 25% Median 75% Max
lnComp 1613 15,158 0,852 11,396 14,587 15,096 15,706 17,683
ROA 1613 0,026 0,146 -0,664 0,009 0,051 0,092 0,303
AR 1613 0,123 0,710 -0,873 -0,274 0,027 0,341 4,306
UNCON1 1613 0,017 0,124 -2,472 -0,016 0,006 0,036 1,961
UNCON2 1613 0,046 0,137 -2,670 -0,001 0,033 0,072 1,618
lnTA 1613 7,594 1,947 3,336 6,146 7,269 8,782 12,897
Table 4 is a table of Pearson and Spearman correlation tests on all our main variables, the bottom left section of the table is the Pearson correlations and the upper right section is the spearman correlations. Firstly, we can see that our compensation variable is positively correlated with both of our performance measures with the market measure being the strongest.
Furthermore, ROA is significantly correlated with all of the other variables at the 5 percent level. The control variable for size, lnTA demonstrates a significance with all the variables except AR. Our two unconditional conservatism measures are correlated to each other and the both are negatively correlated to firm size.
We present our main regressions in table 5. We have two models where we test the UNCONj
variables separately. On both our models, we have positive significance on a 1%- level for AR.
On model 2 ROA is significant on the 1%- level while in model 1 we have significance for ROA on a 5%- level with a p-value of 0,012. The result demonstrate that we have evidence for that both accounting and market performance have a significant positive association with total compensation. We do not draw any further conclusions from model 1 in the difference in significance level between ROA and AR due to the minor differences in the p-value.
Both ROA and UNCONj exclusively have positive relationships with compensation in both our models where UNCONj is significant on a 5%-level in model 1 and 10%-level in model 2.
Suggesting that compensation increases with increasing ROA and UNCON, separately.
In both our models we see a significant negative association on a 5%- significance level between the effect of unconditional conservatism on ROA and compensation. Executive pay decreases with increasing ROA when conservatism increases. This suggests that accounting sensitivity in executive compensation contracts decreases with increasing unconditional accounting conservatism which is in line with what we predicted. We do not have any significant association between the effect of unconditional accounting conservatism on AR and compensation which is what we expected because we predicted a positive or zero effect from this variable.
Table 4. Correlations
lnCOMP ROA AR UNCON1 UNCON2 lnTA
lnCOMP 1 0,1923 0,114 -0,0178 -0,0074 0,7484
ROA 0,2045 1 0,1965 -0,1166 -0,3498 0,1045
AR 0,0576 0,1052 1 0,0106 0,0386 0,0454
UNCON1 -0,0278 -0,3942 -0,0279 1 0,4206 -0,0952
UNCON2 -0,0417 -0,4983 -0,0209 0,7559 1 -0,0207
lnTA 0,7356 0,2342 -0,0084 -0,0698 -0,0735 1
Pearson (Spearman) correlation are below (above) diagonal. Bold numbers represent significance at the 5% level from a two- tailed test
4.3 Robustness test
To test the robustness of our analysis, we conduct further analysis where we incorporate additional variables. In our first test, we change our accounting performance measure from ROA to return on equity (ROE). Subramaniam (2000) states that the choice of performance measure in executive compensation situations may differ, mainly depending on noise and managerial opportunism. Subramaniam (2000) finds evidence for that ROE is more frequent in firms with higher leverage, smaller size, lower growth and lower stock-ownership by managers.
Table 5. Linear regression analysis with lnComp as the dependent variable
Variable Model 1 Model 2
Intercept 12,6148(0,000)*** 12,5969(0,000)***
UNCON1 0,2716(0,030)**
UNCON2 0,2313(0,061)*
ROA 0,2840(0,012)** 0,3139(0,010)***
AR 0,0744(0,000)*** 0,0774(0,000)***
UNCON1ROA -0,1513(0,045)**
UNCON1AR 0,0095(0,826)
UNCON2ROA -0,1764(0,020)**
UNCON2AR -0,0255(0,582)
lnTA 0,3275(0,000)*** 0,3277(0,000)***
F-statistic 167,80(0,000)*** 167,70(0,000)***
Adjusted R-square 0,5539 0,5538
The regression in Model 1 is used with the first measurement for unconditional conservatism UNCON1, Model 2 is based on UNCON2. Both Models are controlled for industry with a one- digit dummy variable based on SIC codes. The p-values are shown within the parenthesis which are based on a two-tailed test. The asterix indicate significance with * significant at 10%, ** significant at 5%, *** significant at 1%
When using ROE as an accounting performance measure instead of ROA, we get positive significant coefficients in ROE and AR where ROE is significant on a 5%-level and AR is significant on a 1%-level. Additionally, UNCON1 is significant and positive at a 1%-level and UNCON2 is significant and positive on a 5%-level.
As we can see, our base variables have the same orientation when conducting the analysis with ROA and ROE. However, our interaction variables change. When using ROE as performance measure, we get a positive significant effect from UNCONjROE instead of negative which was the case when using ROA. In addition, we conducted an untabulated pearson correlation analysis between UNCONjROA and UNCONjROE where we found a negative significant relationship at the 1% level. The opposite applies when correlating ROA and ROE where we instead find a positive significant relationship.
Our results from our main analysis are thereby not robust when conducting the analysis with ROE. However, as Subramaniam (2000) stated, ROE is more widely chosen among smaller firms whilst we have larger listed companies. Additional explanations can be that ROE is not affected by unconditional conservatism to the same degree as ROA. As stated in our first argument, unconditional conservatism lowers assets, but due to timely gains recognition earnings are not understated to the same degree as in conditional conservatism. ROA captures the understatement of assets while ROE does not do that to the same degree.
Table 6. Linear regression analysis with lnComp as the dependent variable
Variable Model 1 Model 2
Intercept 12,5798(0,000)*** 12,5860(0,000)***
UNCON1 0,4154(0,004)***
UNCON2 0,2998(0,029)**
ROE 0,0886(0,020)** 0,0895(0,024)**
AR 0,0782(0,000)*** 0,0808(0,000)***
UNCON1ROE 0,0121(0,008)***
UNCON1AR 0,0017(0,969)
UNCON2ROE 0,0104(0,015)**
UNCON2AR -0,0321(0,500)
lnTA 0,3295(0,000)*** 0,3292(0,000)***
F-statistic 154,51(0,000)*** 153,97(0,000)***
Adjusted R-square 0,5755 0,5438
Model 1 is based on UNCON1 and Model 2 is based on UNCON2. This model uses ROE as a measurment for accounting performance which is calculated as Income before extraordinary items divided by equity. Both Models are controlled for industry with a one-digit dummy variable based on SIC codes. The p-values are shown within the parenthesis which are based on a two-tailed test.
The asterix indicate significance with * significant at 10%, ** significant at 5%, *** significant at 1%
In addition to our sensitivity analysis above, we conduct further tests (untabulated) where we divide our dependent variable total compensation into salary and bonus compensation. When conducting the analysis with salary compensation, we get significant negative results with UNCONJ*ROA while AR and ROA is insignificant. The opposite applies when we use bonus compensation as dependent variable where we get positive significant coefficients with ROA and AR while UNCONJ*ROA is insignificant. We can see from this analysis that our performance measures have an association with bonus compensation, which is in accordance with Murphy (1999) who states that annual bonuses are tied to accounting performance. This does not however explain why we have such a strong relationship between bonus compensation and AR. Further, we see that UNCONJ*ROA affects the salary part of total compensation.
5. Conclusion
Our purpose with this study is to investigate the direct relationship between unconditional conservatism and contracting. Prior research has mainly focused on the benefits of conditional conservatism in contracting situations where the majority of studies have focused on the asymmetrical verifiability properties (Qiang, 2007; Watts, 2003; Beaver and Ryan, 2005). The direct association between unconditional conservatism and contracting is unexplored, however researchers argue that unconditional conservatism should have a negative relationship with contracting due to its inverse relationship with conditional conservatism (Ball & Shivakumar, 2005; Beaver & Ryan, 2005; Qiang, 2007). We continue to build on this were we argue for a negative association between unconditional conservatism and contracting. We base this on the definition of unconditional conservatism which does not specify deferred gains recognition but rather, provides managers the opportunity to recognize gains as timely as losses. Additionally, we include the higher possibilities of earnings management and lower earnings quality into our hypothesis development along with the notion of a higher degree of noise into payoffs from unconditional conservatism (Ball & Shivakumar, 2005; Holmström, 1979; Jackson & Liu, 2010; Penman & Zhang, 2002). All these arguments are believed to decrease contracting efficiency and thereby also decrease accounting sensitivity in executive compensation.
More precisely, we examine accounting sensitivity in executive compensations. We proxy unconditional accounting conservatism with two measures, negative non-operating accruals and cash-flow to earnings skewness (Beatty et al, 2008; Chen et al, 2013; Givoly & Hayn, 2000). We find a significant negative association between executive compensation and accounting sensitivity with both our conservatism measures, suggesting that we have evidence for a negative relationship between unconditional conservatism and contracting. However, our results are not robust when we change our accounting performance variable from ROA to ROE.
We contribute to the existing literature by drawing direct linkages between unconditional conservatism and contracting. By highlighting the financial statement effects of unconditional conservatism, we demonstrate that a timelier gains recognition, earnings management possibilities, and noisy payoffs can affect contracting negatively. In practice, these findings are important for standard setters, financial statement preparers, and users. We highlight the importance of separating conditional and unconditional conservatism in contracting situations
since unspecificity can result in the opposite effect rather than the intended. For instance, by applying unconditional rather than conditional conservatism in contracting purposes would have negative instead of positive effects. It is thereby important for standard setters to have the knowledge of how different accounting standards affect contracting depending on the type of conservatism applied. The same applies to prepares and users of financial statements where different types of conservatism will affect the financial statement and thereby also affect contracting parties differently. Through this and in accordance with Qiang (2007), we highlight the importance of trade-off between the two types of conservatism when applying them in practice.
5.1 Limitations & Further research
We believe that future challenges within accounting conservatism research, as well as the limitations with our study, lies within measuring conservatism. We conducted two measures of unconditional conservatism based on negative non-operating accruals and cash flow to earnings skewness. The first measure is based on the fact that negative non-operating accruals increased with increased conservatism and is widely used as a measure in research (Beatty et al, 2008;
Chen et al, 2013). However, proxying unconditional conservatism in measuring accrued liabilities and loss provisions and decreased tendency in capitalizing costs is difficult (Givoly
& Hayn, 2000). Furthermore, cash flow to earnings skewness is based on the idea that conservatism explains the cumulative difference between cash flows and earnings over time, but this is not completely exclusive to unconditional conservatism since impairments can lower earnings as well. We thereby suggest that future research should focus on developing methodological tools to measure conservatism more accurately and more exclusively when it comes to the distinction between these two. We see this as important in order to advance in the research field of accounting conservatism.
Prior research has discussed conditional conservatism and contracting in terms of asymmetrical verifiability and reliability (Watts, 2003). We try to broaden the discussion by focusing on the financial statement effects of unconditional conservatism and with the stewardship role of accounting, try to argue for how the financial statement consequences of unconditional conservatism affect contracting. We see further possibilities in exploring the linkage between unconditional conservatism, financial statement effects of unconditional conservatism, and contracting. More precisely, it would be interesting to examine the effects of earnings and its difference in conditional and unconditional conservatism considering the difference in gains recognition, the possibilities of earnings management and the difference in earnings quality.
Lastly, we see further research possibilities in examining the difference in ROA and ROE. As stated earlier, we find a negative correlation between UNCONJROA and UNCONJROE. We suggest that future research can focus on this relationship and how unconditional conservatism affects different performance measures and how it affects noise in payoffs when relating them to executive compensation contracts.
References
Ahmed, A. S., Billings, B. K., Morton, R. M., & Stanford-Harris, M. (2002). The role of accounting conservatism in mitigating bondholder-shareholder conflicts over dividend policy and in reducing debt costs. The Accounting Review, 77(4), 867-890.
Ahmed, A. S., & Duellman, S. (2007). Accounting conservatism and board of director characteristics: An empirical analysis. Journal of accounting and economics, 43(2-3), 411- 437.
Ball, R., & Shivakumar, L. (2005). Earnings quality in UK private firms: comparative loss recognition timeliness. Journal of accounting and economics, 39(1), 83-128.
Basu, S. (1997). The conservatism principle and the asymmetric timeliness of earnings.
Journal of accounting and economics, 24(1), 3-37.
Basu, S. (2005). Discussion of “Conditional and Unconditional Conservatism: Concepts and modeling”. Review of Accounting Studies, 10(2-3), 311-321.
Beaver, W. H., & Ryan, S. G. (2005). Conditional and unconditional conservatism: Concepts and modeling. Review of accounting studies, 10(2-3), 269-309.
Beatty, A., Weber, J., & Yu, J. J. (2008). Conservatism and debt. Journal of accounting and economics, 45(2-3), 154-174.
Chen, L. H., Folsom, D. M., Paek, W., & Sami, H. (2013). Accounting conservatism, earnings persistence, and pricing multiples on earnings. Accounting Horizons, 28(2), 233- 260.
Francis, J. R., & Martin, X. (2010). Acquisition profitability and timely loss recognition.
Journal of accounting and economics, 49(1-2), 161-178.
Givoly, D., & Hayn, C. (2000). The changing time-series properties of earnings, cash flows and accruals: Has financial reporting become more conservative?. Journal of accounting and economics, 29(3), 287-320.
Gjesdal, F. (1981). Accounting for stewardship. Journal of Accounting Research, 19(1), 208- 231.
Holmström, B. (1979). Moral hazard and observability. The Bell journal of economics, 74-91.
Lara, J. M. G., Osma, B. G., & Penalva, F. (2016). Accounting conservatism and firm investment efficiency. Journal of Accounting and Economics, 61(1), 221-238.
Iyengar, R. J., & Zampelli, E. M. (2010). Does accounting conservatism pay?. Accounting &
Finance, 50(1), 121-142.
Jackson, S. B., & Liu, X. K. (2010). The allowance for uncollectible accounts, conservatism, and earnings management. Journal of Accounting Research, 48(3), 565-601.
Jensen, M. C., & Meckling, W. H. (1976). Theory of the firm: Managerial behavior, agency costs and ownership structure. Journal of financial economics, 3(4), 305-360.
Lara, J. M. G., Osma, B. G., & Penalva, F. (2009). Accounting conservatism and corporate governance. Review of accounting studies, 14(1), 161-201.
Murphy, K. J. (1999). Executive compensation. Handbook of labor economics, 3, 2485-2563.
O'connell, V. (2006). The impact of accounting conservatism on the compensation relevance of UK earnings. European Accounting Review, 15(4), 627-649.
Penman, S. H., & Zhang, X. J. (2002). Accounting conservatism, the quality of earnings, and stock returns. The accounting review, 77(2), 237-264.
Qiang, X. (2007). The effects of contracting, litigation, regulation, and tax costs on
conditional and unconditional conservatism: Cross-sectional evidence at the firm level. The Accounting Review, 82(3), 759-796.
Roychowdhury, S., & Watts, R. L. (2007). Asymmetric timeliness of earnings, market-to- book and conservatism in financial reporting. Journal of Accounting and Economics, 44(1-2), 2-31.
Ruch, G. W., & Taylor, G. (2015). Accounting conservatism: A review of the literature.
Journal of Accounting Literature, 34, 17-38.
Shackelford, D. A., & Shevlin, T. (2001). Empirical tax research in accounting. Journal of accounting and economics, 31(1-3), 321-387.
Sloan, R. G. (1993). Accounting earnings and top executive compensation. Journal of accounting and Economics, 16(1-3), 55-100.
Subramaniam, C. (2000). Choice of performance measure in a bonus plan: managerial
opportunism or efficient contracting?. Academy of Accounting and Financial Studies Journal, 4(1), 19.
Tosi, H. L., Werner, S., Katz, J. P., & Gomez-Mejia, L. R. (2000). How much does
performance matter? A meta-analysis of CEO pay studies. Journal of Management, 26(2), 301-339.
Watts, R. L. (2003). Conservatism in accounting part I: Explanations and implications.
Accounting horizons, 17(3), 207-221.
Zhang, J. (2008). The contracting benefits of accounting conservatism to lenders and borrowers. Journal of accounting and economics, 45(1), 27-54.
