The authors gratefully acknowledge the helpful comments from Augustine Duru, Kevin Forbes, Samuel Kotz and Robert Moffie. We also thank participants at North Carolina Central University’s Statistical Modeling Seminar for their comments and suggestions. Research assistance of Julius Bradshaw and Sarika Ramakrishnan are sincerely appreciated. Raghavan Iyengar acknowledges summer research grants received from North Carolina Central University.

About

Sections

PDF

Tools

Share a link

Email
Wechat
Bluesky

Abstract

We investigate whether or not there is a link between conservative accounting practices and the sensitivity of executive pay to accounting performance. Using several accrual-based measures of accounting conservatism as well as alternative measures of accounting performance, we estimate an econometric model of CEO compensation that incorporates the interaction of accounting conservatism and accounting performance. Consistent with optimal contracting theory, we find that the sensitivity of executive pay to accounting performance is higher for firms that report conservative accounting earnings. These results support the hypothesis that accounting conservatism, by limiting earnings management opportunities and improving the reliability of accounting performance measures, allows firms to formulate contracts that tie executive compensation more closely to accounting performance.

1. Introduction

According to Watts (2003a), conservative accounting is the ‘asymmetrical verification requirements for gains and losses’ that leads to the ‘persistent understatement of net asset values’. This definition is consistent with the notion of accounting conservatism used by a number of researchers including Feltham and Ohlson (1995), Basu (1997), Gjesdal (1999), Zhang (2000), Beaver and Ryan (2000), Ahmed et al. (2002), Penman and Zhang (2002) and Ryan (2006). It follows from such a definition that conservative accounting practices are accounting choices that tend to impart a downward bias to earnings and the book value of net assets. Relatively speaking, therefore, last-in first-out (LIFO) is a more conservative inventory accounting method than first-in first-out (FIFO), the expensing of R&D is more conservative than capitalising and amortising R&D, and accelerated depreciation methods are more conservative than straight line depreciation.

One explanation of why firms might choose to implement conservative accounting practices lies in efficient contracting theory, i.e. conservative accounting can be used as part of a firm’s strategy to mitigate the conflicts that arise among the many claimants of a firm’s net assets. This is due to the fact that conservative accounting methods place constraints on the distribution of those net assets thereby limiting the scope for self-serving opportunistic behaviour. For example, Ahmed et al. (2002) look at the role of accounting conservatism in mitigating the conflict between bondholders and shareholders. Specifically, because conservative accounting lowers earnings and retained earnings, dividend policy is necessarily restricted, reducing the probability of excessive dividend payouts. Consequently, one hypothesis they put forth is that firms using more conservative accounting practices incur lower costs of debt.

Similarly, conservative accounting may help in aligning the interests of managers and shareholders through its impact on accounting earnings measures that are frequently used in management compensation contracts. As Watts (2003a) points out, by reducing current earnings and understating cumulative earnings and net assets, conservative accounting restricts ‘managements’ opportunistic payments to themselves…’. Similarly, Chen et al. (2007), in a theoretical model which allows for the presence of both conservative accounting and earnings management biases, demonstrates that conservative accounting practices reduce the noise and improve the informativeness of accounting performance measures and reduce the marginal benefits from earnings management. Kwon (2005) also shows that accounting earnings become more useful in controlling the costs of suboptimal managerial decisions when those earnings are measured conservatively rather than neutrally or liberally. Such results suggest that a portion of the cross-firm variation in the parameters of compensation contracts may be explained by the differences among firms in their use of conservative accounting methods. In particular, because conservatism limits the compensation-based incentives for earnings management and improves the reliability and usefulness of accounting performance measures, one would expect to observe higher pay-performance sensitivities in firms that are more conservative in their accounting practices.

An alternative, albeit not mutually exclusive, explanation of the impact of accounting conservatism on managerial compensation relates to the fact that conservatism implies the persistence of earnings increases. If compensation committees reward managers not only for current earnings increases, but also for the persistence of those earnings increases, then one would expect this to be reflected in the relationship between conservatism and managerial compensation. A possible reason for rewarding the persistence of earnings is that it may help to counter the negative impacts of what is known as the ‘horizon problem’, i.e. the problem of aligning the managerial interests with shareholder interests when the expected tenure of the manager is shorter than the optimal investment horizon of the firm (Smith and Watts 1982). In a cross-sectional analysis of chief executive officer (CEO) compensation, Baber et al. (1998) provide evidence of such intervention by compensation committees. In particular, the authors’ results support their hypothesis that the sensitivity of executive cash compensation (salary plus bonuses) to earnings is positively related to the persistence of earnings.¹ In a recent Australian study, Lai and Taylor (2008) document the positive association of conservatism with stock-return volatility, investment cycle length and prior period conservatism. They also report a negative association between conservatism on the one hand and firm age, firm size and leverage. Conservatism thus extends the manager’s investment cycle length and counteracts against the ‘horizon problem’.

In the spirit of Baber et al. (1998), this paper intends to examine the relationship between accounting conservatism and executive compensation. Specifically, it seeks to test whether or not there is a link between conservative accounting practices and the sensitivity of executive pay to accounting performance. The empirical analysis is based on a sample of 4508 firm-year observations over the period of 1994–2003 from the ExecuComp/Compustat databases. Using several accrual-based measures of accounting conservatism as well as alternative measures of accounting performance, we estimate an econometric model of CEO cash compensation that incorporates the interaction of accounting conservatism and accounting performance. The paper’s main result strongly supports the hypothesis that the sensitivity of executive pay to accounting performance is higher for firms that engage in more conservative accounting practices. The result survives a battery of robustness tests.

The remainder of the paper is organised as follows. Section 2 provides an analytical framework and a platform for testable hypothesis. A detailed discussion of the conservatism variables and their construction is undertaken in Section 3. The econometric model is developed in Section 4. A detailed description of the sample and the data is provided in Section 5. Section 6 presents and discusses the major empirical results. A summary and some concluding remarks are offered in Section 7.

2. The analytical model

Our theoretical model is a simplified version and an extension of the model developed by Feltham and Xie (1994). We assume a risk averse manager who performs two activities that generate a payoff to shareholders. Shareholders do not observe the manager’s effort levels, e₁ and e₂, and thus face an incentive problem to be addressed through an incentive compensation contract. Because e₁ and e₂ are not observed, the contract is based on a reported performance measure, y. The first activity is productive as it is aimed entirely at improving y for the sake of increasing the shareholders’ payoff. The second activity is partially productive, but also partly ‘window dressing’ aimed at increasing the measure of performance to the benefit of the manager with no additional benefit to shareholders. The relationship between y and the activity levels is assumed to be linear and given by:

()

where μ_i is the marginal productivity of e_i and inline image

represents the stochastic component of the performance measure. The manager’s wage, w, is assumed to be a linear function of y represented by:

()

where w₀ is a fixed payment and ν is the pay-performance sensitivity. The direct cost to the agent of exerting effort is assumed to be quadratic and given by:

()

The manager realises, however, that he will be penalised by the shareholders if the window dressing activity is uncovered. The penalty is assumed to be proportional to μ₂e₂ and equal to τμ₂e₂, where 0 < τ < 1. The manager estimates a non-zero probability of detection at 0 < ρ < 1. Augmenting equation (3) by the expected penalty, we have:

()

The manager is assumed to choose e₁ and e₂ to maximise the objective function w − C (e₁, e₂). The solutions to the first-order conditions are:

()

From equation (5), we see that the comparative statics for inline image

are sensible. Optimal effort is increasing with the pay-performance sensitivity, i.e. inline image

and with the marginal productivity of effort, i.e. inline image

, as long as (ν − ρτ) > 0. Predictably, the optimal level of inline image

declines with the probability of detection and the penalty proportion, i.e. inline image

Shareholders (the ‘principal’) are assumed to maximise expected surplus, defined as the difference between the gross payoff, B(e₁, e₂), and expected costs, where expected costs include expected compensation costs plus the costs the principal expects to bear due to managerial manipulation of the performance measure y, subject to the incentive constraints given by equation (5). The expected compensation costs include a premium that must be paid to compensate for the risk associated with the incentive programme equal to inline image

. The expected costs of manipulation to the principal are proportional to μ₂e₂ and equal to ημ₂e₂. The principal assigns a non-zero probability of manipulation given by θ. The expected surplus can be written as:

()

where the b_i are the marginal contributions of the manager’s efforts to the gross payoff of the principal. Substituting equation (5) into equation (6), the expected surplus can be written as:

()

Choosing ν to maximise S yields the first-order condition:

()

Note that the optimal pay-performance sensitivity is increasing in the manager’s assessment of the probability of detection, decreasing in the principal’s assessment of the probability of manipulation and decreasing in the variance (noise) associated with the performance measure y, i.e. inline image

The linkages between conservative accounting practices as defined by Watts and managerial compensation can be established by the comparative statics reported above. In particular, conservatism in accounting may: (i) reduce the noise associated with accounting performance measures; (ii) reduce the probability of manipulation, thereby lowering the expected marginal cost of manipulation to the principal; and/or (iii) increase the probability that manipulation is detected, thereby increasing the expected marginal penalty cost of manipulation to the manager-agent.² The reported comparative statics suggest that the optimal pay-performance sensitivity will be lower for all three of these possibilities and hence, generate the paper’s main testable hypothesis stated in its alternative form:

H_a: Conservative accounting practices lead to an increase in the optimal pay-performance sensitivity or equivalently, to an increase in the optimal weight on accounting performance measures used for incentive compensation.

3. Measuring accounting conservatism

Givoly and Hayn (2000) offer a descriptive definition of accounting conservatism that provides a basis for our choice of conservatism measures. Specifically, the authors define conservatism as ‘a selection criterion between accounting principles that leads to the minimisation of cumulative reported earnings by slower revenue recognition, faster expense recognition, lower asset valuation and higher liability valuation.’³ This characterisation, in highlighting the multi-period aspect of conservatism, suggests that accounting accruals can be used to determine the extent to which conservative accounting practices are chosen over more aggressive ones. In particular, the definition suggests that firms choosing more conservative accounting practices should, over time, exhibit a pattern of negative accruals, i.e. a predominance of periods when reported earnings are less than operating cash flows. This is reinforced by Watts (2003b) who writes:

Conservatism’s asymmetrical treatment of gains and losses produces an asymmetry in accruals. Because losses tend to be fully accrued while gains do not, periodic accruals tend to be negative and cumulative accruals to be understated. (p. 289)

Moreover, because our interest lies in the degree to which firms choose conservatism, it is the discretionary component of accruals that becomes the important metric.

To decompose total accruals into their non-discretionary and discretionary components, we begin with the standard Jones (1991) model given by⁴:

()

where TA_it is total accruals for firm i in year t, ΔREV_it is the change in i’s total revenue from t − 1 to t, GPPE_it is the gross acquisition cost of property, plant and equipment for firm i in year t, and A_it−1 is the value of total assets for firm i in year t − 1. The fundamental notion is that working capital accruals are expected to increase with sales and long-term accruals are expected to increase with GPPE. The residuals from the firm-by-firm ordinary least squares (OLS) estimation of equation (9) are taken as measures of discretionary accruals. The model was later modified by Dechow et al. (1995) who point out that the standard Jones model implicitly assumes that managers exercise no discretion over revenues. The authors argue that such an assumption is highly questionable since managers can exercise considerable discretion in accruing revenues especially those related to credit sales. Use of the standard Jones model will therefore yield downwardly biased estimates of discretionary accruals. To mitigate this bias, Dechow et al. (1995) suggest a modified Jones model given by:

()

where ΔREC_it is the change in accounts receivable for firm i from the previous year. The authors provide empirical evidence that the OLS residuals from equation (10) are relatively better estimates of discretionary accruals than the residuals from equation (9).

Subsequently, Ball and Shivakumar (2005) argued that the above Jones-type models are inherently misspecified because they fail to address the asymmetry in the timeliness of (unrealized) gains and losses recognition. The authors contend that because of such asymmetry, accruals cannot be linear in cash flows and instead are better modelled in a piecewise linear fashion. They go on to show that compared with the typical linear accruals models like equations (9) and (10), piecewise linear constructs of these same models are better at predicting normal (expected) accruals so that the associated residuals are better estimates of discretionary accruals.

Based on the above articles, we compute firm-specific residuals from the estimation of the following four versions of the piecewise linear construct:

()

where TA_it is total accruals for firm i in year t, ΔREV_it is the change in i’s total revenue from t − 1 to t, ΔREC_it is the change in i’s accounts receivable from t − 1 to t, GPPE_it is the gross acquisition cost of property, plant and equipment for firm i in year t, A_it−1 is the value of average total assets for firm i in year t − 1, CF_it is the level of cash flow for firm i, DCF_it is a dummy variable that is equal to 1, if CF_it is negative and 0 otherwise, ΔCF_it is the change in the level of cash flow for firm i and DΔCF_it is a dummy variable that is equal to 1, if ΔCF_it is negative and 0 otherwise. The parameter, α₅, is expected to be positive, indicative of a more timely recognition of unrealized losses than unrealized gains.

Following the approach proposed in DeFond and Jiambalvo (1994), the conservatism measures used in this paper are the residuals (i.e. abnormal accruals) from the OLS estimations of equations (11)–(14)multiplied by−1.⁵ Multiplication by −1 yields a conservatism measure that increases with conservatism. The measures are labelled as CNSV₁, CNSV₂, CNSV₃ and CNSV₄, respectively. As explained in a following section, we recognise that the abnormal accruals models (equations (11)–(14)) measure conservatism with an error. Consequently, we also employ errors-in-variables regression in our tests.

4. A model of executive compensation

It has been long established in the literature that managerial compensation contracts are based both on accounting and stock price measures of performance.⁶ Consistent with this literature, our basic model of executive compensation is given by:

()

where Δln COMP_it is the one-period change in the natural log of executive compensation, ΔAcctPerf_it and ΔMKRET_it are the one-period change in accounting performance (as measured by return on assets (ROA)) and market performance, respectively.⁷ΔROA is the change in ROA defined as the sum of after-tax interest expense and earnings before extraordinary income and discontinued operations, less tax expense divided by total assets.⁸ Equation (15) or variants thereof have been employed in a number of empirical studies examining the sensitivity of executive pay to accounting performance and stock price measures, including Lambert and Larcker (1987), Sloan (1993), Baber et al. (1998), Core et al. (2003) and Leone et al. (2009). The natural log specification for the dependent variable helps in mitigating the adverse impact of any outliers, while the ‘changes’ specification controls for a number of firm-specific factors that have been shown to influence compensation but which exhibit very little if any variation over time.⁹

The coefficients, β₁ and β₂, represent the sensitivities of pay to accounting and stock performance, respectively. To examine whether these parameters are affected by the degree of accounting conservatism, we extend equation (15) to include the interactions of the performance variables with our conservatism measures. Specifically, the primary model is given by:

()

where CNSV_j is one of the four measures of conservatism as defined in the previous section for j = 1–4.

Equation (16) is estimated for both CEO cash (salary plus bonus) and total compensation (in our robustness tests). ΔMKRET represents the change in stockholders’ return, including reinvestment of dividends. A complete description of how the models’ variables are measured is provided in the Appendix.

Consistent with prior research, we expect the signs of β₁ and β₂ to be positive indicating that improvements in a firm’s accounting and market performance are positively related to the changes in executive compensation. A positive value for β₃ is expected suggesting that the responsiveness of executive pay to changes in accounting performance increases with increases in conservatism. This would support the hypothesis that conservatism, by limiting earnings management opportunities and improving the reliability of performance measures, allows firms to formulate contracts that tie executive compensation more closely to accounting performance. Additionally, this would support the hypothesis by Baber et al. (1998) that compensation committees reward managers for the persistence of earnings as a way to mitigate the horizon problem. Finally, the parameter β₄ is expected to be either zero or negative. If stock prices efficiently incorporate earnings persistence, good news and bad news, then the interaction term between the market return and the conservatism measure adds no information not already reflected by the market return itself. Hence, one would expect β₄ to be equal to zero. Alternatively, if increases in accounting conservatism cause greater reliance and weight to be put on accounting earnings in compensation decisions, then β₄ would be expected to be negative.

5. Research design

5.1. Sample

To compile the sample of firms, we began with the population of non-financial, non-utility firms from the ExecuComp database for the period 1994–2003 which consisted of 6642 firm-year observations. From this population, we deleted 1651 observations with insufficient financial data in the Compustat database and eliminated another 202 observations for firms with insufficient compensation data in ExecuComp database. Another 63 observations were deleted because of insufficient time-series data with which to calculate firm-specific residuals.¹⁰ Finally, the winsorizing of extreme outliers in the top and bottom 0.5 per cent of each of the variables’ distributions resulted in the elimination of 218 observations and a final sample of 4508 firm-year observations. Table 1 provides details of the sample selection process.

Table 1.
Selection of sample of firm-years 1994–2003

	Number of firm-years
Number of non-financial, non-utility, firm-year observations from ExecuComp database for the sample period	6642
Less: firm-years
With insufficient financial data in the Compustat database	(1651)
With insufficient CEO compensation data in ExecuComp database	(202)
With insufficient observations to obtain firm-specific residuals (minimum eight observations required for each firm to run firm-specific regression)	(63)
With extreme outliers (i.e. observations in the top and bottom 0.5 per cent of each of the variables)	(218)
Number of firm-year observations in the final sample	4508

5.2. Descriptive statistics

Table 2 presents the distribution of our sample by industry and mean values of the study’s main variables. The industry distribution of our sample is similar to prior studies using comparable data, e.g. Frankel et al. (2002) and Whisenant et al. (2003). Table 2 also reports the mean of the four conservatism measures namely, CNSV₁, CNSV₂, CNSV₃ and CNSV₄ for all firms by industry. The data indicate that companies in the computer industry, on average, have the largest mean decrease in accounting returns (0.89 per cent decrease in ROA) and market returns (22.21 per cent); they also report the most conservative measure of accounting income as evidenced by CNSV₁ (0.0169), CNSV₂ (0.0159), CNSV₃ (0.0178), CNSV₄ (0.0167). On average, companies in the chemical and pharmaceutical industries have the largest discretionary accruals and concomitantly the least conservative accounting earnings measure.

Table 2.
Descriptive statistics: means of variables

Industry^a	No.	Cash compensation	Total compensation	ΔROA	ΔMKRET	CNSV₁	CNSV₂	CNSV₃	CNSV₄
Mining and construction	41	962	2101	−0.28	−18.42	0.0135	0.0017	0.0150	0.0028
Food	293	1887	5542	−0.06	−1.99	−0.0106	0.0007	−0.0111	0.0004
Textiles, print/publishing	686	1101	2675	−0.19	−0.66	−0.0037	−0.0026	−0.0043	−0.0032
Chemicals	365	1302	3692	0.17	1.21	−0.0161	−0.0153	−0.0171	−0.0164
Pharmaceuticals	363	1211	5127	−0.17	−4.07	−0.0208	−0.0103	−0.0212	−0.0105
Extractive	130	958	2047	0.76	4.48	0.0045	−0.0041	0.0074	−0.0013
Durable	2062	1090	3110	−0.50	−15.04	−0.0083	−0.0106	−0.0084	−0.0110
Computers	568	986	4935	−0.89	−22.21	0.0169	0.0159	0.0178	0.0167
Overall	4508	1152	3601	−0.36	−10.17	−0.0056	−0.0054	−0.0057	−0.0056

^aIndustry membership is determined by SIC code as follows: mining and construction (1000–1999, excluding 1300–1399), food (2000–2111), textiles & printing/publishing (2200–2799), chemicals (2800–2824, 2840–2899), pharmaceuticals (2830–2836), extractive (2900–2999, 1300–1399), durable manufacturers (3000–3999, excluding 3570–3579 and 3670–3679), computers (7370–7379, 3570–3579, 3670–3679). Transportation (4000–4799), retail (5000–5999), services (7000–8999), excluding 7370–7379) and financial services (6000–6999) firms and utilities (4900–4999) are excluded from the sample. This classification is followed in the prior accounting literature (see Whisenant et al. 2003). ΔROA, changes in return on asset; ΔMKRET, changes in total return to shareholders, including reinvestment of dividends.

Other descriptive statistics are also interesting. For example, companies in the food industry had the highest CEO cash and total compensation of $1 887 000 and $5 542 000, respectively. It is also worth noting that, firms in the extractive industry, on average had the lowest CEO cash compensation ($958 000) and lowest CEO total compensation ($2 047 000) while reporting the largest mean increase in accounting and market returns of 0.76 and 4.48 per cent, respectively.

6. Results

6.1. Ordinary least squares results

We begin the analysis by presenting OLS results ignoring potential measurement errors in the conservatism variable. Table 3 presents the results of the OLS estimation of equation (16) where the conservatism measure is based on residuals obtained from equations (11) through (14), respectively. Parameter estimates are given along with the corresponding p-values. In all four models, we use the change in cash compensation (Δln COMP) as the dependent variable.

Table 3.
Results of pooled OLS regressions with Δ in CEO cash compensation as the dependent variable^a (n = 4508)

ΔlnCOMP = β₀ + β₁ CNSV_j + β₂ΔROA + β₂ΔMKRET + β₃ (CNSV_j × ΔROA) + β₄ (CNSV_j ×ΔMKRET) + ε₁
Details	Model 1	Model 2	Model 3	Model 4
Intercept	0.1073 (0.000)***	0.1079 (0.000)***	0.1080 (0.000)***	0.1088 (0.000)***
CNSV₁	0.0975 (0.264)
CNSV₂		0.1033 (0.245)
CNSV₃			0.2050 (0.024)**
CNSV₄				0.2191 (0.016)**
ΔROA	0.0006 (0.000)***	0.0010 (0.000)***	0.0007 (0.000)***	0.0011 (0.000)***
ΔMKRET	0.0001 (0.013)**	0.0001 (0.041)**	0.0001 (0.016)**	0.0001 (0.045)**
CNSV₁ × ΔROA	0.0150 (0.000)***
CNSV₂ × ΔROA		0.0156 (0.000)***
CNSV₃ × ΔROA			0.0165 (0.000)***
CNSV₄ × ΔROA				0.0175 (0.000)***
CNSV₁ × ΔMKRET	−0.0000 (0.772)
CNSV₂ × ΔMKRET		0.0001 (0.352)
CNSV₃ × ΔMKRET			−0.0000 (0.922)
CNSV₄ × ΔMKRET				0.0002 (0.160)
Adjusted R²	0.0126	0.0140	0.0132	0.0150
F-statistic	7.46 (0.000)***	9.08 (0.000)***	7.60 (0.000)***	9.63 (0.000)***

^a p-Values in parentheses; **, *** indicates two-tailed significance at the 0.10, 0.05 levels, respectively. CNSV₁, CNSV₂, CNSV₃ and CNSV₄ are residuals obtained by running firm-specific regressions from equations (11) through (14)multiplied by−1. ΔlnCOMP, changes in natural log of CEO cash compensation, where cash compensation is the sum of CEO salary and bonus; ΔROA, changes in return on asset; ΔMKRET, changes in total return to shareholders, including reinvestment of dividends. All changes are from 1 year to the next.

Table 3, Model 1, shows that the changes in cash compensation are significantly associated with ΔMKRET and ΔROA indicating that improvements in a firm’s market and accounting performance are positively related to the changes in executive compensation. The positive and significant estimated coefficient on the interaction term CNSV₁ × ΔROA (i.e. β₃) provides support for the hypothesis that executive compensation is more sensitive to the changes in accounting performance with increases in conservatism. A similar result holds for the remaining three models where we employ the alternative measures of conservatism; the estimated coefficients of CNSV_j × ΔROA are positive and significant at standard levels for all of the models.

Table 2 illustrates that changes in accounting performance and market performance vary widely across industries. Hence, we perform the analysis again using industry-adjusted accounting and market performance variables.¹¹ Table 4 reports the results. We find that coefficients of ADJMKRET and ADJROA are positively associated with CEO compensation as predicted with p-values of <0.05. The coefficient on CNSV_j is statistically insignificant, suggesting that CEOs are not penalised for using a conservative measure. Once again, as expected, the coefficients of CNSV_j × ADJROA are positive and significant, indicating that the responsiveness of executive pay to the changes in accounting performance increases with accounting conservatism.

Table 4.
Results of pooled OLS regressions with Δ in CEO cash compensation as the dependent variable^a (n = 4508)

ΔlnCOMP = β₀ + β₁ CNSV_j + β₂ ADJROA + β₂ ADJMKRET + β₃ (CNSV_j × ADJROA) + β₄ (CNSV_j × ADJMKRET) + ε₁
Details	Model 1	Model 2	Model 3	Model 4
Intercept	0.1000 (0.000)***	0.0977 (0.000)***	0.0996 (0.000)***	0.0974 (0.000)***
CNSV₁	−0.0591 (0.458)
CNSV₂		−0.1036 (0.208)
CNSV₃			0.0024 (0.976)
CNSV₄				−0.0388 (0.639)
ADJROA	0.0003 (0.002)***	0.0004 (0.001)***	0.0004 (0.001)***	0.0004 (0.000)***
ADJMKRET	0.0014 (0.000)***	0.0014 (0.000)***	0.0014 (0.000)***	0.0014 (0.000)***
CNSV₁ × ADJROA	0.0080 (0.003)***
CNSV₂ × ADJROA		0.0087 (0.000)***
CNSV₃ × ADJROA			0.0093 (0.001)***
CNSV₄ × ADJROA				0.0096 (0.000)***
CNSV₁ × ADJMKRET	0.0029 (0.002)***
CNSV₂ × ADJMKRET		0.0013 (0.175)
CNSV₃ × ADJMKRET			0.0027 (0.005)***
CNSV₄ × ADJMKRET				0.0010 (0.321)
Adjusted R²	0.0496	0.0486	0.0500	0.0490
F-statistic	44.35 (0.000)***	44.28 (0.000)***	44.64 (0.000)***	44.78 (0.000)***

^a p-Values in parentheses; *** indicates two-tailed significance at 0.01 level, respectively. CNSV₁, CNSV₂, CNSV₃ and CNSV₄ are residuals obtained by running firm-specific regressions from equations (11) through (14) multiplied by−1. ΔlnCOMP, changes in natural log of CEO cash compensation, where cash compensation is the sum of CEO salary and bonus; ADJROA, industry-adjusted (three-digit SIC) return on asset; ADJMKRET, industry-adjusted (three-digit SIC) 1 year total return to shareholders, including reinvestment of dividends. All changes are from 1 year to the next.

6.2. Errors-in-variables regression

It is well-known that measurement error in one or more of the independent variables in a linear regression can result in biased OLS slope estimators, even asymptotically. Because accounting conservatism is proxied by abnormal accruals from different models, constrained as they are by the availability of data over an 8 year period to compute firm-specific residuals, it would be rather naïve to ignore measurement errors in the conservatism variable. Specifically, we are interested in ascertaining whether the results reported in Tables 3 and 4 are mere artefacts of measurement error bias, rather than signals of real association. We therefore run all four models by explicitly stipulating the reliability of the conservatism variable, as well as its interactions with other explanatory variables, at 0.90.¹² Table 5 reports the results from the errors-in-variables (EIV) specification. We find that taking into account measurement error in conservatism variable leads to similar inferences about the association between executive pay and accounting conservatism. The estimation of errors-in-variables regression yields statistically significant and positive coefficients for the interaction terms CNSV_j × ΔROA in all four models. Thus, potential measurement error bias does not appear to be the cause of the documented positive sensitivity of executive compensation to accounting performance in the presence of conservatism.

Table 5.
Results of errors-in-variables regressions with Δ in CEO cash compensation as the dependent variable^a (n = 4508)

ΔlnCOMP = β₀ + β₁ CNSV_j + β₂ΔROA + β₂ΔMKRET + β₃ (CNSV_j × ΔROA) + β₄ (CNSV_j ×ΔMKRET) + ε₁
Details	Model 1	Model 2	Model 3	Model 4
Intercept	0.1085 (0.000)***	0.1106 (0.000)***	0.1093 (0.000)***	0.1120 (0.000)***
CNSV₁	0.1560 (0.133)
CNSV₂		0.2172 (0.045)**
CNSV₃			0.2823 (0.007)***
CNSV₄				0.3689 (0.001)*
ΔROA	0.0007 (0.000)***	0.0014 (0.000)***	0.0008 (0.000)***	0.0016 (0.000)***
ΔMKRET	0.0001 (0.000)***	0.0001 (0.000)***	0.0001 (0.000)***	0.0001 (0.000)***
CNSV₁ × ΔROA	0.0187 (0.000)***
CNSV₂ × ΔROA		0.0234 (0.000)***
CNSV₃ × ΔROA			0.0208 (0.000)***
CNSV₄ × ΔROA				0.0273 (0.000)***
CNSV₁ × ΔMKRET	−0.0001 (0.693)
CNSV₂ × ΔMKRET		0.0001 (0.477)
CNSV₃ × ΔMKRET			−0.0000 (0.905)
CNSV₄ × ΔMKRET				0.0002 (0.298)
Adjusted R²	0.0139	0.0172	0.0149	0.0194
F-statistic	11.53 (0.000)***	12.83 (0.000)***	12.09 (0.000)***	13.80 (0.000)***

^a p-Values in parentheses; *, **, *** indicates two-tailed significance at the 0.10, 0.05 and 0.01 levels, respectively. CNSV₁, CNSV₂, CNSV₃ and CNSV₄ are residuals obtained by running firm-specific regressions from equations (11) through (14) multiplied by−1. ΔlnCOMP, changes in natural log of CEO cash compensation, where cash compensation is the sum of CEO salary and bonus; ΔROA, changes in return on asset; ΔMKRET, changes in total return to shareholders, including reinvestment of dividends. All changes are from 1 year to the next.

We also replicate Table 4 by using the EIV specification with the industry-adjusted variables and report the results in Table 6. The estimated coefficients on ADJMKRET and ADJROA are both generally positive and significant, suggesting that improvements in stock performance or industry-adjusted accounting performance are associated with increases in CEO compensation. Crucially, in all four cases, the estimated coefficients on the interaction terms (CNSV_j × ADJROA) are positive and significant at conventional levels. These results also lend support to our hypothesis that the sensitivity of executive pay to industry-adjusted accounting performance is higher for firms that engage in more conservative accounting practices.

Table 6.
Results of errors-in-variables regressions with Δ in CEO cash compensation as the dependent variable^a (n = 4508)

ΔlnCOMP = β₀ + β₁ CNSV_j + β₂ ADJROA + β₂ ADJMKRET + β₃ (CNSV_j × ADJROA) + β₄ (CNSV_j × ADJMKRET) + ε₁
Details	Model 1	Model 2	Model 3	Model 4
Intercept	0.0998 (0.000)***	0.0971 (0.000)***	0.0994 (0.000)***	0.0967 (0.000)***
CNSV₁	−0.0750 (0.417)
CNSV₂		−0.1325 (0.162)
CNSV₃			−0.0082 (0.930)
CNSV₄				−0.0621 (0.517)
ADJROA	0.003 (0.016)**	0.0004 (0.003)***	0.0004 (0.007)***	0.0005 (0.001)***
ADJMKRET	0.0014 (0.000)***	0.0014 (0.000)***	0.0014 (0.000)***	0.0014 (0.000)***
CNSV₁ × ADJROA	0.0092 (0.004)***
CNSV₂ × ADJROA		0.0103 (0.001)***
CNSV₃ × ADJROA			0.0106 (0.001)***
CNSV₄ × ADJROA				0.0114 (0.000)***
CNSV₁ × ADJMKRET	0.0032 (0.009)
CNSV₂ × ADJMKRET		0.0013 (0.294)
CNSV₃ × ADJMKRET			0.0028 (0.020)**
CNSV₄ × ADJMKRET				0.0009 (0.468)
Adjusted R²	0.0501	0.0491	0.0505	0.0495
F-statistic	47.04 (0.000)***	46.01 (0.000)***	47.44 (0.000)***	46.38 (0.000)***

^a p-Values in parentheses; **, *** indicates two-tailed significance at the 0.05 and 0.01 levels, respectively. CNSV₁, CNSV₂, CNSV₃ and CNSV₄ are residuals obtained by running firm-specific regressions from equations (11) through (14) multiplied by−1. ΔlnCOMP, changes in natural log of CEO cash compensation, where cash compensation is the sum of CEO salary and bonus; ADJROA, industry-adjusted (three-digit SIC) return on asset; ADJMKRET, industry-adjusted (three-digit SIC) 1 year total return to shareholders, including reinvestment of dividends. All changes are from 1 year to the next.

6.3. Robustness tests

In this section, we explore whether the main findings are robust to alternative specifications.¹³ Prior research suggests that CEO compensation is sensitive to firm size. For instance, Jensen and Meckling (1976) contend that the largest firms hire the better performing executives to maximise firm productivity. Using meta-analysis, Tosi et al. (2000) conclude that firm size accounts for over 40 per cent of the variance in executive pay. Conyon and Murphy (2000) using USA and UK data and Kaplan (1994) using Japanese data provide strong evidence in favour of the size hypothesis. First, since firms are not of a uniform size in our sample, we perform our analysis by considering the size-adjusted ROA, SZROA. We find that our main result of a pay-performance sensitivity that increases with conservatism holds in all four cases. Specifically, we find that the estimated coefficient on CNSV₄ × SZROA is positive and highly significant (coefficient estimate = 0.0103; p-value = 0.006).¹⁴ We find qualitatively similar evidence using OLS and EIV estimations.

Second, we consider an alternative accounting performance measure to ROA. Subramaniam (2000) investigates the choice of accounting performance measure (specifically ROA versus ROE) for executive incentive pay. He argues that while managerial opportunism may play a part in choosing ROE for incentive pay, an efficient contracting rationale for choosing ROE method cannot be ruled out. We therefore specifically replace ROA by ROE to investigate whether the sensitivity of executive pay to accounting conservatism is affected by the choice of accounting performance measure. We find that when firms use ROE, the results are no different to those reported in Tables 5 and 6. The estimated coefficient on CNSV₄ × ΔROE is positive and significant (coefficient estimate = 0.0041; p-value = 0.005).

Finally, we replace changes in CEO cash compensation by changes in CEO total compensation. We find markedly similar evidence using OLS and EIV estimations. The estimated coefficient on CNSV₄ × ΔROA is positive and significant (coefficient = 0.0371; p-value < 0.001).

The hypothesis parameters remain positive and significant throughout all of these robustness tests thereby supporting our main hypothesis that the sensitivity of executive pay to accounting performance increases in accounting conservatism.

7. Summary and discussion

Prior literature has extensively examined the design of executive bonus and performance plans. Given the importance of conservatism in accounting theory and literature, it seems critical to examine the relationship between accounting conservatism and executive compensation. Specifically, we test whether or not there is a link between conservative accounting practices and the sensitivity of executive pay to accounting performance. Our study is based on a large sample of 4508 observations during the period from 1994 to 2003. Using several accrual-based measures of accounting conservatism as well as alternative measures of accounting performance, we estimate econometric models of both cash and total CEO compensation that incorporate the interaction of accounting conservatism and accounting performance. We report two important findings. First, our results indicate that the sensitivity of executive pay to accounting performance is higher for firms that engage in more conservative accounting practices. Second, the results are robust to a number of alternative specifications and estimation techniques. These findings strongly support the hypothesis that conservatism, by limiting earnings management opportunities and improving the reliability of accounting performance measures, allows firms to formulate contracts that tie executive compensation more closely to accounting performance. Alternatively, these results may be construed as reinforcing the argument that compensation committees reward managers for reporting conservative earnings as a way to mitigate the horizon problem.

Accounting conservatism, from our perspective, is critical to the alignment of the interests of managers and shareholders. Conservative accounting practices can afford firms an opportunity to use incentive pay effectively to increase firm value while lowering the probability of managerial opportunism. In its absence, the use of performance incentives in managerial and executive compensation contracts is fraught with the risks of managerial opportunism, aggressive accounting, the overstatement of the financial performance of firms, and the diminishment of the integrity and information content of financial reporting. The results of this paper together with recent research on the information role of conservatism, e.g. Khan and Watts (2007) and LaFond and Watts (2008), holds an important lesson for the accounting profession especially in light of the current financial crisis that will no doubt engender an intense scrutiny of accounting methods and practices. The message, especially to the Financial Accounting Standards Board (FASB), is that conservative accounting is an effective response to the information asymmetry between managers and shareholders, and the subsequent agency costs. Recent FASB proposal (FAS 157-e) to relax mark-to-market rules for banks sitting on billions of dollars in toxic assets would allow banks to set their own values for certain troubled mortgages, corporate loans and consumer loans. This rule change will afford banks greater leeway in determining the market value of loans and provide them with a powerful incentive to keep the assets directly on their books at values far in excess of the fair values. The FASB proposal would thus forestall efforts by the United States Department of the Treasury to rid the banks of those toxic assets and record the loss in the income statement. Such attempts by the FASB to eliminate accounting conservatism from financial reporting may very well be misguided and in need of re-evaluation.

Footnotes

1 This result is consistent with a number of other studies, including Abdel-khalik (1985), Abdel-khalik et al. (1987), Dechow et al. (1994), Gaver and Gaver (1998), and Duru et al. (2002), which offer evidence that executive compensation is effectively shielded by compensation committees, whereby GAAP earnings are adjusted to account for the income-decreasing effects of otherwise valuable activities, such as investing in R&D, restructuring profitable operations, and discontinuing unprofitable operations. In contrast, research by Healy et al. (1987) examining the compensation effects of changes in inventory valuation and depreciation methods and Defeo et al. (1989) examining the compensation effects of equity-for-debt swaps provide evidence that executive cash compensation (salary plus bonus) is largely determined by reported earnings with no evidence of active intervention by compensation committees to account for possible managerial manipulation of earnings.

2 These are consistent with the work of LaFond and Watts (2008) and Khan and Watts (2007) who argue and provide evidence that accounting conservatism is a governance response to the information asymmetry between investors and managers that reduces managers’ abilities to manipulate and overstate financial performance.

3 Givoly and Hayn (2000), p. 292.

4 The original Jones model does not contain an intercept term, since the intercept was also deflated by lagged assets. It is by no means clear that total accruals would be zero if the independent variables are also zero. We have therefore corrected this, by including an intercept in each of the models.

5 We multiply the value of abnormal accruals by negative one so that larger values represent greater conservatism and vice versa.

6 The theoretical basis for this can be found in papers by Holmstrom (1979) and Banker and Datar (1989).

7 In robustness tests, we use an alternate accounting performance measure, namely, return on equity (ROE).

8 This definition of ROA was originally proposed by Fama.

9 Such factors include firm size, board composition and a number of other governance and stock ownership related characteristics.

10 A minimum of eight time-series observations were required for each firm to run firm-specific regressions.

11 We compute this by taking the difference between firms’ accounting (market) returns and the mean accounting (market) returns for all companies in the firm’s three-digit standard industrial code (SIC) classification.

12 Using a lower reliability of 0.85 produced markedly similar results. These results are available upon request.

13 Although we estimated coefficients for all four models in Tables 5 and 6, we only report the untabulated results of our sensitivity tests corresponding to model 4 of Table 6 due to space considerations. All of our sensitivity results are available upon request.

14 We calculate SZROA as the difference between firms’ accounting (market) returns and the mean accounting (market) returns for all companies within ±20 per cent of the experimental firm’s sales for the corresponding year.

Appendix

Definition of variables

CNSV₁, CNSV₂, CNSV₃, CNSV₄ are residuals obtained from the following (respective) firm-specific regressions multiplied by−1.

()

ΔlnCOMP is the changes in natural log of CEO cash compensation, where cash compensation is the sum of CEO salary and bonus. Changes are from year to year. ΔROA is the changes in return on assets where changes are from year to year. ROA = 100 (IB + XINT (1 − TR/100) − TXT)/AT, where IB is Income before extraordinary items and discontinued operations, XINT is interest expense, TR is the tax rate, TXT is total income tax and AT is total assets. ΔMKRET is the change in total return to shareholders, including reinvestment of dividends. ADJROA is the firm’s ROA – industry (three-digit SIC) average ROA. ADJMKRET is the firm’s MKRET – industry (three-digit SIC) average MKRET.

References

Abdel-khalik, A. R., 1985, The effect of LIFO-switching and firm ownership on executives’ pay, Journal of Accounting Research 23(Autumn), 427–447.
10.2307/2490819
Web of Science® Google Scholar
Abdel-khalik, A. R., C. Chi, and D. Ghicas, 1987, Rationality of executive compensation schemes and real accounting changes, Contemporary Accounting Research 4(Fall), 32–60.
10.1111/j.1911-3846.1987.tb00654.x
Google Scholar
Ahmed, A. S., B. S. Billings, R. M. Morton, and M. Stanford-Harris, 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.
10.2308/accr.2002.77.4.867
Web of Science® Google Scholar
Baber, W. R., S. H. Kang, and K. R. Kumar, 1998, Accounting earnings and executive compensation: the role of earnings persistence, Journal of Accounting and Economics 25, 169–193.
10.1016/S0165-4101(98)00021-4
Web of Science® Google Scholar
Ball, R., and L. Shivakumar, 2005, Earnings quality in UK private firms: comparative loss recognition timeliness, Journal of Accounting and Economics 39(1), 83–128.
10.1016/j.jacceco.2004.04.001
Web of Science® Google Scholar
Banker, R. D., and S. M. Datar, 1989, Sensitivity, precision and linear aggregation of signals for performance evaluation, Journal of Accounting Research 27(1), 21–39.
10.2307/2491205
Web of Science® Google Scholar
Basu, S., 1997, The conservatism principle and asymmetric timeliness of earnings, Journal of Accounting and Economics 24, 3–37.
10.1016/S0165-4101(97)00014-1
Web of Science® Google Scholar
Beaver, W. H., and S. G. Ryan, 2000, Biases and lags in book value and their effects on the ability of the book-to-market ratio to predict book return on equity, Journal of Accounting Research 38(1), 127–148.
10.2307/2672925
Web of Science® Google Scholar
Chen, Q., T. Hemmer, and Y. Zhang, 2007, On the relation between conservatism in accounting standards and incentives for earnings management, Journal of Accounting Research 45(3), 541–565.
10.1111/j.1475-679X.2007.00243.x
Web of Science® Google Scholar
Conyon, M. J., and K. J. Murphy, 2000, The prince and the pauper? CEO pay in the United States and United Kingdom, The Economic Journal 110, 640–671.
10.1111/1468-0297.00577
Web of Science® Google Scholar
Core, J. E., W. R. Guay, and R. E. Verrechia, 2003, Price versus non-price performance measures in optimal CEO compensation contracts, The Accounting Review 78(4), 957–981.
10.2308/accr.2003.78.4.957
Web of Science® Google Scholar
Dechow, P., M. R. Huson, and R. G. Sloan, 1994, The effect of restructuring charges on executives’ cash compensation, The Accounting Review 69(1), 138–56.
Web of Science® Google Scholar
Dechow, P., R. G. Sloan, and A. Sweeney, 1995, Detecting earnings management, The Accounting Review 70(2), 193–225.
Web of Science® Google Scholar
Defeo, V. J., R. A. Lambert, and D. F. Larcker, 1989, Executive compensation effects of equity-for-debt swaps, The Accounting Review 64(2), 210–27.
Web of Science® Google Scholar
DeFond, M. I., and J. Jiambalvo, 1994, Debt covenant violation and manipulation of accruals, Journal of Accounting and Economics 17, 145–176.
10.1016/0165-4101(94)90008-6
Web of Science® Google Scholar
Duru, A. I., R. J. Iyengar, and A. Thevaranjan, 2002, The shielding of CEO compensation from the effects of strategic expenditures, Contemporary Accounting Research 19(2), 175–193.
10.1506/UM8Q-NVJ6-JKGT-GH5W
Web of Science® Google Scholar
Feltham, G. E., and J. A. Ohlson, 1995, Valuation of clean surplus accounting for operating and financial activities, Contemporary Accounting Research 11(Spring), 689–731.
10.1111/j.1911-3846.1995.tb00462.x
Google Scholar
Feltham, G. E., and J. Xie, 1994, Performance measure congruity and diversity in multi-task principal/agent relations, The Accounting Review 69(3), 429–453.
Web of Science® Google Scholar
Frankel, R., M. Johnson, and K. Nelson, 2002, The relation between auditors’ fees for nonaudit services and earnings management, The Accounting Review 71, 106.
Google Scholar
Gaver, J. J., and K. M. Gaver, 1998, The relation between nonrecurring accounting transactions and CEO cash compensation, The Accounting Review 73(2), 235–253.
Web of Science® Google Scholar
Givoly, D., and C. Hayn, 2000, The changing time-series properties of earnings, cash flows, and accruals: has financial reporting become more conservative? Journal of Accounting and Economics 29, 287–320.
10.1016/S0165-4101(00)00024-0
Web of Science® Google Scholar
Gjesdal, F., 1999. A steady state growth valuation model: a note on accounting and valuation, working paper (Norwegian School of Economics, Bergen, Norway).
Google Scholar
Healy, P. M., K. Sok-Hyan, and K. Palepu, 1987, The effect of accounting procedure changes on CEOs’ cash salary and bonus, Journal of Accounting and Economics 9, 7–35.
10.1016/0165-4101(87)90015-2
Web of Science® Google Scholar
Holmstrom, B., 1979, Moral Hazard and observability, The Bell Journal of Economics 10, 74–91.
10.2307/3003320
Web of Science® Google Scholar
Jensen, M. C., and W. Meckling, 1976, Theory of the firm: managerial behavior, agency costs, and ownership structure, Journal of Financial Economics 3, 305–360.
10.1016/0304-405X(76)90026-X
CAS Web of Science® Google Scholar
Jones, J., 1991, Earnings management during import relief investigations, Journal of Accounting Research 29, 193–228.
10.2307/2491047
Web of Science® Google Scholar
Kaplan, S. N., 1994, Top executive rewards and firm performance: a comparison of Japan and the United States, The Political Economy 102, 510–546.
10.1086/261944
CAS Web of Science® Google Scholar
Khan, M., and R. L. Watts, 2009. Estimation and empirical properties of a firm-year measure of accounting conservatism, Journal of Accounting and Economics doi: 10.1016/j.jacceco.2009.08.002.
Google Scholar
Kwon, Y. K., 2005, Accounting conservatism and managerial incentives, Management Science 51(11), 1626–1632.
10.1287/mnsc.1050.0417
Web of Science® Google Scholar
LaFond, R., and R. L. Watts, 2008, The information role of conservatism, The Accounting Review 83(2), 447–478.
10.2308/accr.2008.83.2.447
Web of Science® Google Scholar
Lai, C., and S. L. Taylor, 2008, Estimating and validating a firm-year-specific measure of conservatism: Australian evidence, Accounting and Finance 48(4), 673–695.
10.1111/j.1467-629X.2008.00270.x
Web of Science® Google Scholar
Lambert, R., and D. F. Larcker, 1987, An analysis of the use of accounting and market measures of performance in executive compensation contracts, Journal of Accounting Research 25 (Supplement), 85–125.
10.2307/2491081
Web of Science® Google Scholar
Leone, A. J., J. S. Wu, and J. L. Zimmerman, 2009. Asymmetric sensitivity of CEO cash compensation to stock returns, Journal of Accounting and Economics (forthcoming).
Google Scholar
Penman, S. H., and X. Zhang, 2002, Accounting conservatism, the quality of earnings and stock returns, The Accounting Review 77(2), 237–264.
10.2308/accr.2002.77.2.237
Web of Science® Google Scholar
Ryan, S. G., 2006, Identifying conditional conservatism, European Accounting Review 15(4), 511–525.
10.1080/09638180601102099
Web of Science® Google Scholar
Sloan, R. G., 1993, Accounting earnings and top executive compensation, Journal of Accounting and Economics 16, 55–100.
10.1016/0165-4101(93)90005-Z
Web of Science® Google Scholar
Smith, C. W., and R. L. Watts, 1982, Incentive and tax effects of executive compensation plans, Australian Journal of Management 7, 139–157.
10.1177/031289628200700204
Google Scholar
Subramaniam, C., 2000, Choice of performance measure in a bonus plan: managerial opportunism or efficient contracting? Academy of Accounting and Financial Studies Journal 4, 19–37.
Google Scholar
Tosi, H. L., S. Werner, J. P. Katz, and L. R. Gomez-Mejia, 2000, How much does performance matter? A meta-analysis of CEO pay studies, Journal of Management 26, 301–339.
10.1177/014920630002600207
Web of Science® Google Scholar
Watts, R., 2003a, Conservatism in accounting part I: explanations and implications, Accounting Horizons 17, 207–221.
10.2308/acch.2003.17.3.207
Google Scholar
Watts, R., 2003b, Conservatism in accounting part II: evidence and research opportunities, Accounting Horizons 17, 287–301.
10.2308/acch.2003.17.4.287
PubMed Web of Science® Google Scholar
Whisenant, S., S. Sankaraguruswamy, and K. Raghunandan, 2003, Evidence on the joint determination of audit and non-audit fees, Journal of Accounting Research 41(4), 721–744.
10.1111/1475-679X.00121
Web of Science® Google Scholar
Zhang, X., 2000, Conservative accounting and equity valuation, Journal of Accounting and Economics 29, 125–149.
10.1016/S0165-4101(00)00016-1
Web of Science® Google Scholar

Citing Literature

Volume50, Issue1

March 2010

Pages 121-142

Does accounting conservatism pay?

Abstract

1. Introduction

2. The analytical model

3. Measuring accounting conservatism

4. A model of executive compensation