An empirical study of the attributes impacting the decisions of accountants to accept a pay cut in exchange for reduced work hours.

Ward, Terry J. ; Albright, Meredith T.

INTRODUCTION

Accounting is considered a well-respected and highly beneficial career choice for most college graduates. If a new graduate puts in his/her time to gain experience and knowledge, he/she can expect a good income and benefits and the prestige that goes with the "accountant" title. However, do the benefits exceed the costs of working in accounting? A career in accounting typically involves much more than a forty-hour a week job. Many accountants work sixty or more hours per week during tax season, and the workload does not always lessen during non-peak times (Solnik, 2005).

These increased workloads and extended hours can lead to increased levels of stress. A common term used to describe a situation where someone reaches his/her stress limit at work is job burnout. One answer to the problems of job burnout and work-family conflict is for CPA firms to offer flexible work arrangements. The idea of flexible work schedules has been around for some time and has been documented in the work place as a major incentive in recruiting by Almer and Kaplan (2002). Almer and Kaplan documented higher job satisfaction and lower turnover among firms offering flexible hours.

However, little empirical research has attempted to identify the variables that determine which employees are willing to accept or desire flexible schedules. Not all employees desire flexible schedules and some employers may not have the resources to implement flexible schedules. Being able to identify, at the time of recruitment, employees that may desire flexible schedules later in their careers would be very useful to employers.

The purpose of this paper is to identify the attributes associated with employee desires for flexible work schedules. We specifically look at how gender and certification impact attitudes regarding flexible work schedules.

PRIOR RESEARCH

Gaertner and Ruhe (1981) mailed out a questionnaire similar to the one used for this study. It was sent out to 398 professionals at regional and national CPA firms in the Midwestern USA, and garnered 193 usable responses. According to Gaertner and Ruhe, many accountants reported heavy smoking and alcohol consumption, ulcers, chronic back pain and headaches as responses to stress (Gaertner & Ruhe, 1981, p. 68). Their results suggest that senior and junior staff accountants exhibit the highest levels of work-related stress. Senior accountants' stress was caused by uncertainty about future employment prospects, compensation inequity, unwanted overtime, and boredom. Junior staff accountants' stress resulted mostly from not being involved in decisions affecting their work assignments (Gaertner & Ruhe, 1981). Results did not differ greatly between regional and national firms.

Greenhaus and Beutell (1985) extended Gaetner and Ruhe's research by looking at work-family conflict. Greenhaus and Beutell found that conflict at work impacted behavior at home. Greenhaus and Beutell determined that work conflict is associated with the number of hours worked per week, the time spend commuting, the amount of overtime, and an inflexible work schedule. Family conflict comes about from specific family role characteristics that require time spent in family activities. When the job requirements are incompatible with family requirements, it creates a time-based conflict. Conflict-based pressures occur when the stressors of one role affects the performance in another role. Thus, time-based conflict consists of both work and family related conflicts, with job stress affecting how employees behave at home. Similarly, family related conflict that results when there is excessive strain at home and can affect how employees perform at work.

Subsequent research across various areas of accounting seems to validate the findings of Gaertner and Ruhe (1981) and Grennhaus and Beutell (1985); stress and the work environment in accounting often create situations that lead to employee dissatisfaction and turnover (e.g., Choo, 1986; Collins and Killough, 1992; Dalton et al., 1997; Fisher, 2001; Fogarty et al., 2000; Kelly and Margheim, 1990; Rebele et al., 1996; Sanders et al., 1995; Snead and Harrell, 1991; Viator, 2001; Weick, 1983). It appears that the phenomenon known as "job burnout" may be at the core of such employee dissatisfaction and turnover

Sweeney and Summers (2002) distinguish job burnout by three components: emotional exhaustion, depersonalization and reduced personal accomplishment. Sweeney and Summers (2002) conducted a study based on this three-component model to measure role stressors and job burnout. In their study, they explain that emotional exhaustion is the first sign of job burnout, and is created by higher levels of work demand. Emotional exhaustion is followed by depersonalization, which occurs when the employee starts becoming emotional detached from clients and co-workers. The last of the components is a reduced sense of personal accomplishment, caused by a negative attitude about the job.

Sweeney and Summers (2002) results showed no direct relationship between burnout and the workload prior to an accountants' busy season when hours were just beginning to increase. The average workload during that time was forty-nine hours per week. They did, however, show a direct increase in burnout among public accountants during the busy season when the workload increased to approximately sixty-three hours per week. Sweeney and Summers also discovered a high level of emotional exhaustion at the end of the busy season, which is considered a key component in job burnout.

One answer to the problems of job burnout and work-family conflict is for CPA firms to offer flexible work arrangements. Flexible schedules are already becoming a major incentive in recruiting. Studies show a higher job satisfaction and lower turnover among those firms offering flexible hours (Almer & Kaplan, 2002). Flexible work hours may also help alleviate work-family conflicts. Padgett et al. (2005) found in a survey of 118 professionals who had worked in public accounting that a flexible schedule may be more important than the number of hours worked. They found that flexible schedules reduced stress by allowing employees to manage their own schedules, thus improving job satisfaction (Padgett et al., 2005).

HYPOTHESES

Stress, job burnout, and work-family conflict in accounting likely impacts female accountants differently than male accountants. Scheuermann et al. (1998) found that women often leave public accounting because of work stress and burnout. Scheuermann et al. conclude in their analysis, "It appears that many of these problems are gender specific and are the cause of these women's stress" (p. 38). They based this conclusion on written comments from the respondents. Similar studies also suggested that women leave public accounting sooner than men (Padgett et al., 2005).

The works of Meyers-Levy (1988) and Meyers-Levy and Sternthal (1991) on gender differences suggest that gender roles impact employee satisfaction. Their research suggests that males are more agentic (masterful) and forcefully follow goals having immediate personal consequences. Females, however, are more communal and focus on harmonious relationships. Females are also more likely to consider family harmony as the most important goal while males are more sensitive to self-relevant information.

Thus, based on prior gender research, we hypothesize that:

H1: Female accountants are more likely than males to have considered accepting compensation reductions in exchange for reduced work hours.

Stress also plays an important role in employee satisfaction. Prior research indicates that higher stress levels result in higher employee dissatisfaction and turnover (e.g., Choo, 1986; Collins and Killough, 1992; Fisher, 2001; Fogarty et al., 2000; Viator, 2001). Although these studies often looked at hours worked, salary levels, and job environments, they did not consider how certification can impact desires for flexible programs. Certified accountants are normally assigned greater responsibilities in their work environments and are thus more likely to encounter stressful situations than noncertified accountants. To investigate this certification-satisfaction association, we hypothesize that:

H2: Certified accountants are more likely than noncertified accountants to have considered accepting compensation reductions in exchange for reduced work hours.

SUBJECTS

Members of the Nashville, Tennessee chapter of the Institute of Management Accountants and employees of the Nashville branch of a major CPA firm agreed to participate in our study. Two hundred questionnaires were sent to potential subjects, and of those, 114 were returned (a response rate of fifty-seven percent). Four surveys were removed from the sample because the subjects were either retired or not in the accounting profession. One additional response was deleted because of missing data. These deletions left 109 usable survey responses.

Our survey was similar to that of Gaertner & Ruhe (1981) and contained thirteen questions. The survey asked basic demographic and job information such as: (1) job title, gender, years experience, salary level, education level, average hours subject works per week, marital status, number of children, age, and spouse's occupation. In addition to the above information, the survey also asked if the subjects had ever considered taking a pay cut in order to decrease the number of hours worked per week, and if they had considered taking a pay cut, what percentage reduction would they accept?

Table 1 provides demographic information on the respondents. It shows that there is a good representation among the various demographics. Most of the respondents were certified, with forty-eight percent possessing a CPA license, fourteen percent possessing a CMA license, and seven percent possessing both. Most of the respondents had over fifteen years of experience, while the salary levels were somewhat concentrated in the higher levels (fifty percent of the subjects had incomes of $70,000 or higher). The respondents were a highly education group of professional, with ninety-six percent possessing a Bachelors degree or higher. The gender demographics of the sample were as expected, with sixty-four percent of the sample being male. These gender numbers should be sufficient for running the models to test the hypotheses in this study.

METHODOLOGY

Dependent (Response) and Independent Variables

We developed three log-linear models to test our hypotheses. The dependent variable in this study was a dichotomous indicator variable coded as:

CUT = 0, if the subject has not considered accepting a reduction in compensation in order to decrease the number of hours worked per week, or 1, if the subject has considered accepting a reduction in compensation in order to decrease the number of hours worked per week.

Seven control variables and two variables of interest made up the independent variables. The main two independent variables of interest are subject gender and certification coded as:

GENDER = 0, if subject was male, or 1, if subject was female;

CERTI = 0, if subject was not certified, or 1, if subject was certified as either a CMA or CPA.

These two variables were the main variables of interest used to test the hypotheses. We also looked at three partitions of CERTI: not certified, CMA certified, CPA certified, or both CMA and CPA certified. The results (not reported) were similar to that of the dichotomous measure.

Additional independent variables were included in our model to control for work experience, education, average hours worked per week, marital status, number of children in the household, age, and salary. These variables were important to control for information that may explain why an employee would accept a reduction in compensation in order to decrease the number of hours worked per week. The control variables are:

EXPER = years of work experience;

ED = 0, if subject has less than a four year college degree, 1, if subject has a four year college degree, or 2, if subject also has a Masters or higher degree;

HOURS = average number of hours the subject works per week;

MARITAL = 0, if subject is not married, or 1, if subject is married;

NO_CHILD = number of children the subject has in his/her household;

AGE = age;

SALARY = 0, if subject's current annual salary is less than $45,000, 1. if subject's current annual salary is between $45,000 and $60,000,

2, if subject's current annual salary is between $60,000 and $75,000,

3, if subject's current annual salary is between $75,000 and $90,000,

and

4, if subject's current annual salary is over $90,000;

Log-linear Modeling Using Nominal Logistic Regression

The log-linear models were constructed using logistic regression. This study employed proportional odds to generate the models (see, for example, Agresti, 1984 for a discussion of various logistic regression models.). This procedure fits a parallel lines regression model based on transformed cumulative logits and has been used extensively in the past, especially in research regarding financial distress (e.g., Casey and Bartczak, 1985; Gentry et al., 1985; 1987; Aziz et al., 1988; Aziz and Lawson, 1989; Kennedy, 1992; Ward, 1992; 1994; 1999; 2006). Hosmer and Lemeshow (1989) provide a thorough discussion of binary logit regression.

In prior distress studies, the independent variables were normally either continuous or dichotomous variables. For this study, two of the independent control variables (ED and SALARY) have more than two categories or levels. For these variables, nominal logistic regression will create multiple parameter estimates and test statistics to determine the differences between each two levels of the independent variable across the dependent variable. For example, ED will have two parameter estimates and SALARY will have four.

RESULTS

Incremental Value of GENDER and CERTI

We first ran the following base model: CUT = EXPER + ED + HOURS + MARITAL + NO_CHILD + AGE + SALARY. Then, each of the two variables of interest, GENDER and CERTI, were added separately to the base model to determine whether either variable has incremental value beyond that of the control variables. Results of these regressions are reported in Table 2.

The results for the Base Model indicate that only the number of hours worked (HOURS) is a strong predictor of subjects having considered a pay cut in exchange for reduced work hours (at p-value < .05). As expected, the sign of the parameter estimate for HOURS is positive, indicating that the more hours the accountant worked per week, the more likely he/she would consider a pay cut in exchange for reduced hours. The signs of the coefficients on the four comparisons across levels of SALARY reveal an interesting result. The signs on the parameter estimates for SALARY switch from positive to negative across the various levels, indicating that after the accountant earns a certain minimum salary per year ($45,000 for this study), the more money the accountant makes the less likely he/she is willing to take a pay cut for reduced work hours. The parameter estimate for SALARY for the comparisons between the third and fourth categories (between incomes of $75,000 to $90,000 and over $90,000) indicates a limited negative significant difference (at p-value < .10). Thus, this negative relationship becomes significant for incomes over $90,000.

The most likely explanation for this income relationship is a function of necessity and a lessening of stress (or more control over stressful situations) at more senior levels of employment. The authors looked at correlations and frequency totals between pay cut, job rank, and experience. As expected, job rank and experience were highly correlated (p-value < .001). The frequency totals (not reported) showed that employees with less experience making lower salaries did not consider a pay cut. Accountants in the middle ranks, salaries, and experience often considered taking a pay cut for reduced work hours. Thus, once a desired minimum salary is obtained by the accountant, he/she appears to more willing to take a pay cut for less stress.

However, the trend reverses at the highest rank, pay level, and experience. Accountants at the highest ranks seldom considered a pay cut for reduced work loads. Apparently, accountants at the highest ranks either have less stress than those at lower ranks or have greater control over stressful situations. This result is not surprising since high level employees have more input into their work schedules and would tend to have greater abilities to control stressful situations.

Our two variables of interest, GENDER and CERTI, were added separately to the base model to test whether gender and certification significantly impact an employee's decision to accept a pay cut for reduced work hours. The Change in -2Log Likelihood statistic was calculated to measure the incremental contribution of each variable to the log-linear models. The test is a one-tailed test since we anticipate a directional effect; specifically, a positive relationship between the two variables and CUT.

The second model in Table 2 includes GENDER as an independent variable. When GENDER is included in the model, the Change in -2Log Likelihood statistic of 8.631 is highly significant (p-value < .01). The sign of the parameter estimate indicates that females are significantly more likely to have considered accepting a pay cut for reduced work hours (flexible work plan) than are males. This result is consistent with prior research suggesting females are more likely to consider family harmony, while males are more sensitive to self-relevant information. This result also may help explain why females tend to leave the accounting profession at a higher rate than males. Flexible work schedules are likely more appealing to female accountants than to male accountants.

Thus, we fail to reject H1. After controlling for experience, education hours worked, marital status, number of children, age, and level of salary, female accountants are more likely than males to accept pay cuts for reduced work hours (flexible work plan).

The final model in Table 2 contains CERTI added to the base model. Although the results for certification are not as strong as the results for GENDER, the Change in -2Log Likelihood statistic of 2.943 is still significant at conventional levels (p-value < .05). The sign of the parameter estimate is, as expected, positive.

This result may appear counterintuitive, but it is consistent with prior stress research. Certified accountants have greater responsibilities and stress related duties than noncertified accountants. Thus, working long hours in stressful situations would likely lead to a greater desire for more flexible work schedules. After controlling for other basic variables, certified accountants are more likely than noncertified accountants to have considered pay cuts in exchange for reduced work hours. Thus, we fail to reject H2.

Full Model

GENDER and CERTI could be surrogate variables measuring the same construct. Thus, a full model including all variables was tested. The results for the full model are reported in Table 3.

Although one would expect the results for the full model to be somewhat weaker than those for the incremental models (the results for the full model are likely to be impacted by collinearity), the results for the Wald Chi-Square statistics should be similar to those for the Change in -2Log Likelihood reported earlier.

The results for the full model are consistent with the incremental models. Both GENDER and CERTI are significant, but GENDER is the stronger predictor variable. The signs for both are positive as expected, indicating that GENDER and CERTI are positively related with CUT.

Contingency Tables

Two-way Contingency Tables were developed to investigate the relationship between the two variables of interest and CUT. Panel 1 of Table 4 contains the GENDER by CUT contingency table results, while Panel 2 of Table 4 contains the CERTI by CUT contingency table results. The first number in each cell (yii) represents the observed number of observations for that cell, while the second in each cell (mii) represents the number of observations one would expect by chance.

The GENDER by CUT contingency table results show that, as expected, a strong relationship exists between GENDER and CUT (the Pearson [chi square] is significant at a p-value < .01). The cells show that nineteen out of the thirty-five females (over 54%) had considered a pay cut in exchange for reduced work hours, while only nineteen of the seventy-four males (less than 26%) had done so.

The results for the CERTI by CUT contingency table are similar (Pearson [chi square] significant at p-value < .02). In this case, twenty-seven out of sixty seven certified accountants (40%) had considered a pay cut in exchange for reduced work hours, while only eight out of forty-two (19%) of the noncertified accountants had done so.

Thus, the results for the contingency tables are consistent with the hypotheses and agree with the log-linear models. Females and certified accountants are significantly more likely to have considered a pay cut for reduced work loads (flexible work schedule) than males and noncertified accountants. These results are consistent after controlling for various demographic attributes (log-linear models) and when looking at one-way relationships (contingency tables).

CONCLUSIONS

The issue of work loads, employee turnover, and flexible work schedules are very important issues facing the accounting profession. Past research suggests that accountants have high turnover and employers have a difficult time keeping female accountants. The results of this study provide empirical evidence concerning the attributes that may help explain which employees would desire flexible work schedules (a willingness to take a pay cut for a reduced work schedule).

The results of this study suggest that gender and certification status of the accountant impacts his/her desire for a flexible schedule. Female accountants are significantly more likely to desire a flexible schedule than males. Employers need to consider that the goals and ambitions of female accountants may be more inherent in nature than those of male accountants.

This study's results also suggest that certified accountants are more likely to desire pay cuts for reduced work loads than noncertified accountants. The authors are unaware of prior research suggesting this relationship. This result suggests that employers should be careful concerning work loads and expectations of its certified accountants. Failure of a company to create an equitable work environment for certified accountants will more likely result in an increased probability of losing its most qualified accountants. Apparently, income is not the only thing important to certified accountants.

The subjects used in this study are from the Middle Tennessee area (Nashville and surrounding counties). The results for this study may be geographic specific and not applicable to other areas of the country. Thus, the external validity of this research may be limited.

There has been much research concerning the working conditions in the accounting profession but little to address how individual and family attributes can impact work variables such as efficiency, satisfaction, etc. Further empirical research may be warranted concerning the family structures of accountants and how they impact work related issues such as flexible scheduling.

ACKNOWLEDGMENT

We wish to thank participants of the Allied Academies Fall Conference for their helpful comments on an earlier version of this paper. Also, this paper had benefited greatly from the comments of two anonymous reviewers.

REFERENCES

Agresti, A. (1984). Analysis of Ordinal Categorical Data. New York: John Wiley and Sons, Inc.

Almer, E. D. & S. Kaplan (2002). The effects of flexible work arrangements on stressors, burnout, and behavioral job outcomes in public accounting. Behavioral Research in Accounting, 14, 1-34.

Aziz, A., D. C. Emanuel & G. H. Lawson (1988). Bankruptcy prediction--An investigation of cash flow based models. Journal of Management Studies, 25 (September), 419-437.

Aziz, A. & G. H. Lawson (1989). Cash flow reporting and financial distress models: Testing of hypotheses. Financial Management, 18 (spring), 55-63.

Casey, C. J. & N. J. Bartczak (1985). Using operating cash flow data to predict financial distress: Some extensions. Journal of Accounting Research, 23 (spring), 384-401.

Choo, F. (1986). Job stress, job performance, and auditor personality characteristics. Auditing: A Journal of Practice and Theory, 5, 17-32.

Collins, K. M. & L. N. Killough (1992). An empirical examination of stress in public accounting. Accounting, Organizations, and Society, 17, 535-548

Dalton, D. R., J. W. Hill & R. J. Ramsay (1997). Women as managers and partners: context specific predictors of turnover in international public accounting firms. Auditing: A Journal of Practice and Theory, (Spring), 29-50.

Fisher, R. T. (2001). Role stress, the Type A behavior pattern, and external auditor job satisfaction and performance. Behavioral Research in Accounting, 13, 143-170.

Fogarty, T. J., J. Singh, G. K. Rhoads & R. K. Moore (2000). Antecedents and consequences of burnout in accounting: Beyond the role stress model. Behavioral Research in Accounting, 12, 31-67.

Gaertner, J. & K. Ruhe (1981). Job-related stress in public accounting. Journal of Accountancy, 151 (June), 68-74.

Gentry, J. A., P. Newbold & D. T. Whitford (1985). Classifying bankrupt firms with funds flow components. Journal of Accounting Research, 23 (spring), 146-160.

Gentry, J. A., P. Newbold & D. T. Whitford (1987). Funds flow components, financial ratios, and bankruptcy. Journal of Business Finance and Accounting, 14 (winter), 595-606.

Greenhaus, J. H. & N. J. Beutell (1985). Sources and conflict between work and family roles. Academy of Management Review, 10, 76-88.

Hosmer, D. W. & S. Lemeshow (1989). Applied Logistic Regression. New York: AICPA.

Kelly, T. & L. Margheim (1990). The impact of time budget pressure, personality, and leadership variables on dysfunctional auditor behavior. Auditing: A Journal of Practice and Theory, 9, 21-42.

Kennedy, D. B. (1992). Classification techniques in accounting research: Empirical evidence of comparative performance. Contemporary Accounting Research, 8 (spring), 419-442.

Meyers-Levy, J. (1988). The influence of sex roles on judgment. Journal of Consumer Research, 14 (March), 522-530.

Meyers-Levy, J. & B. Sternthal (1991). Gender differences in the use of message cues and judgments. Journal of Marketing Research, 28 (February), 84-96.

Padgett, M., K. Paulson, G. Hughes, S. Hughes & C. Born (2005). The relationship between pre-employment expectations, experiences, and length of stay in public accounting. Journal of Leadership & Organizational Studies, 12, 82-103.

Rebele, J. E., R. E. Michaels & R. Wachter (1996). The relationship of career stage to job outcomes and role stress: A study of external auditors. Advances in Accounting, 14, 241-258.

Sanders, J. C.,D. L. Fulks & J. K. Knoblett (1995). Stress and stress management in public accounting. The CPA Journal, 65 (8), 46.

Scheuermann, S. B. & J. B. Finch (1998). Why women leave public accounting. Business Forum, 23, 36-40.

Snead, K. & A. Harrell (1991). The impact of psychological factors on the job satisfaction of senior auditors. Behavioral Research in Accounting, 3, 85-96.

Solnik, C. (2005). No day of rest for weary accountants as work piles up. Long Island Business News, (July 22), 1.

Sweeney, J. T. & S. L. Summers (2002). The effect of busy season workload on public accountants' job burnout. Behavioral Research in Accounting, 14, 223-246.

Viator, R. E. (2001). The association of formal and informal public accounting mentoring with role stress and related job outcomes. Accounting, Organizations and Society, 26, 73-93.

Ward, T. J. (1992). The incremental predictive ability of net and gross cash flows using four-state ordinal models of financial distress. Southern Business and Economic Journal, 16 (October), 26-44.

Ward, T. J. (1994). An empirical study of the incremental predictive ability of Beaver's naive operating flow measure using four-state ordinal models of financial distress. Journal of Business Finance and Accounting, 21 (June), 547-561.

Ward, T. J. (1999). A review of financial distress literature and recommendations for future research. Academy of Accounting and Financial Studies Journal, 3 (winter), 136-151.

Ward, T. J., B. Foster & J. Woodroof (2006). Estimated operating cash flow, reported cash flow from operating activities, and financial distress. Advances in Quantitative Analysis of Finance and Accounting, 4 (summer), 97-120.

Weick, K. E. (1983). Stress in accounting systems. The Accounting Review, 58, 350-369.

Terry J. Ward, Middle Tennessee State University

Meredith T. Albright, Accounting Manager, Nash Lipsey Burch, LLC

Table 1: Demographics

Position: Owner/Partner 7%
 CFO 7%
 Manager 32%
 Controller 18%
 Senior 10%
 Staff 13%
 Finance 5%
 Misc. 8%

Experience: 1-5 years 21%
 6-10 years 14%
 11-15 years 9%
 over 15 years 56%

Gender: Female 36%
 Male 64%

Certification: CPA 48%
 CMA 14%
 No certification 38%
 Both CPA and CMA 7%

Salary Level: $30,000-$45,000 11%
 $45,000-$60,000 21%
 $60,000-$75,000 17%
 $75,000-$90,000 18%
 $90,000+ 32%
 No response 1%

Highest Level of PhD 1%
Education Masters 36%
 Bachelors 59%
 Associates 3%
 High School Diploma 1%

Table 2: Multiple Models: Base Model of Control Variables with Each
Variable of Interest Added Separately

 Base Model Parameter

Variables Estimates Wald
 [chi square]

Intercept -2.615 2.248
EXPER -0.119 0.167
ED 0 vs. 2 0.670 0.710
ED 1 vs. 2 -0.113 0.058
HOURS 0.086 4.142 **
MARITAL -0.359 0.345
NO_CHILD -0.056 0.066
AGE 0.383 0.882
SALARY 0 vs. 4 1.414 2.232
SALARY 1 vs. 4 -0.023 0.002
SALARY 2 vs. 4 -0.539 0.270
SALARY 3 vs. 4 -0.904 3.301 *
GENDER
CERTI

Model -2Log Likelihood 15.624 (11df)
[Delta] in -2Log Likelihood N/A

 Base Model with
 GENDER Variable
 Parameter

Variables Estimates Wald
 [chi square]

Intercept -2.599 0.875
EXPER -0.019 0.004
ED 0 vs. 2 1.206 1.844
ED 1 vs. 2 -0.422 0.736
HOURS 0.083 3.289 *
MARITAL -0.575 0.801
NO_CHILD -0.168 0.535
AGE 0.329 0.618
SALARY 0 vs. 4 1.691 2.640
SALARY 1 vs. 4 0.079 0.020
SALARY 2 vs. 4 -0.412 0.627
SALARY 3 vs. 4 -1.117 4.312 **
GENDER 0.757 7.963 ***
CERTI

Model -2Log Likelihood 24.255 ** (12df)
[Delta] in -2Log Likelihood 8.631 *** (1df)

 Base Model with
 CERTI Variable
 Parameter

Variables Estimates Wald
 [chi square]

Intercept -2.477 1.200
EXPER -0.079 0.068
ED 0 vs. 2 0.485 0.303
ED 1 vs. 2 -0.149 0.095
HOURS 0.080 3.636 **
MARITAL -0.256 0.169
NO_CHILD -0.066 0.086
AGE 0.370 0.804
SALARY 0 vs. 4 1.141 1.321
SALARY 1 vs. 4 -0.199 0.130
SALARY 2 vs. 4 -0.433 0.743
SALARY 3 vs. 4 -0.802 2.455
GENDER
CERTI 0.466 2.813 *

Model -2Log Likelihood 18.567 * (12df)
[Delta] in -2Log Likelihood 2.943 ** (1df)

The base model includes the control variables EXPER, ED, HOURS,
MARITAL, NO_CHILD, AGE, and SALARY regressed on CUT. CUT = whether a
subject considered a pay cut for reduced work hours, coded 0 if
subject did not consider taking a pay cut and 1 if subject considered
taking a pay cut. EXPER = years of experience for subject. ED =
level of education with three levels; coded 0 if subject did not
have a college degree, coded 1 if subject had a Bachelors degree,
and coded 2 if subject had a Masters or higher degree. HOURS =
number of hours, on average, that subject worked each week.
MARITAL = whether subject was married with two levels, coded 0 if
subject was not married and coded 1 if subject was married.
NO_CHILD = number of children subject had. AGE = age of subject.
SALARY = annual salary of subject with five levels; coded 0 for
incomes between $30,000 to $45,000, coded 1 for incomes between
$45,000 and $60,000, coded 2 for incomes between $60,000 and
$75,000, coded 3 for incomes between $75,000 and $90,000, and coded
4 for incomes over $90,000.

Each variable of interest, GENDER and CERTI, was added individually
to the Base Model to determine whether each variable has incremental
usefulness over the base control variables. GENDER = gender of
subject with two levels, coded 0 for males and 1 for females.
CERTI = whether subject is certified with two levels, coded 0 if
subject is not certified and 1 if subject has certification (CPA or
CMA).

The Wald [chi square] (two degrees of freedom for ED, four degrees
of freedom for SALARY, and one degree of freedom for each of the
remaining independent variables) tests the significance of each
individual variable, while the -2Log Likelihood tests the predictive
significance of the overall model.

Change in -2Log Likelihood = the change in the -2Log Likelihood
statistic from the Base Model to the Added Model with GENDER, CERTI,
or SALARY added to the Base Model. A significant Change in -2Log
Likelihood chi-square indicates that the added variable has
incremental usefulness over the accounting information measured by
the control variables. Since the hypotheses are based on directional
tests, the tests are a one-tailed test for the two variables of
interest using the Change in -2Log Likelihood Statistic.

*** Significant at p-value [less than or equal to] .01.
** Significant at p-value [less than or equal to] .05.
* Significant at p-value [less than or equal to] .10.

Table 3: Full Model with all Independent Variables Together

Variables Parameter Standard Wald
 Estimates Error [chi square]

Intercept -2.094 2.466 0.721
EXPER 0.087 0.323 0.072
ED 0 vs. 2 0.971 0.926 1.100
ED 1 vs. 2 -0.352 0.507 0.481
HOURS 0.077 0.047 2.733 *
MARITAL -0.457 0.657 0.483
NO_CHILD -0.192 0.231 0.688
AGE 0.314 0.429 0.535
SALARY 0 vs. 4 1.481 1.143 1.677
SALARY 1 vs. 4 -0.191 0.591 0.104
SALARY 2 vs. 4 -0.280 0.544 0.266
SALARY 3 vs. 4 -1.034 0.559 3.412 *
GENDER 0.810 0.277 8.552 ***
CERTI 0.572 0.306 3.481 *
Model -2Log 27.995 (13df) ***
 Likelihood

All variables and statistics are as described in Table 2.
*** Significant at p-value [less than or equal to] .01.
** Significant at p-value [less than or equal to] .05.
* Significant at p-value [less than or equal to] .10.

Table 4: Two-Way Contingency Tables

Panel 1, GENDER by CUT Contingency Table

 GENDER

 Male Female

CUT No pay cut [y.sub.11] 55 [y.sub.12] 16
 [m.sub.11] 48.20 [m.sub.12] 22.79
 Pay cut [y.sub.21] 19 [y.sub.22] 19
 [m.sub.21] 25.79 [m.sub.22] 12.20
 Totals [y.sub.+1] 74 [y.sub.+2] 35

Pearson [chi square] = 8.564 (p-value = .00, with 1df)

Panel 2, GENDER by CERTI Contingency Table

 CERTI

 Not Yes

CUT No Pay Cut [y.sub.11] 34 [y.sub.12] 40
 [m.sub.11] 28.51 [m.sub.12] 45.48
 Pay Cut [y.sub.21] 8 [y.sub.22] 27
 [m.sub.21] 13.48 [m.sub.22] 21.51
 Totals [y.sub.+1] 42 [y.sub.+2] 67

Pearson [chi square] = 5.648 (p-value = .02, with 1df)

 Totals

CUT No pay cut 71 [y.sub.1+]

 Pay cut 38 [y.sub.2+]

 Totals 109 n

Pearson [chi square] = 8.564 (p-value = .00, with 1df)

Panel 2, GENDER by CERTI Contingency Table

 Totals

CUT No Pay Cut 74 [y.sub.1+]

 Pay Cut 35 [y.sub.2+]

 Totals 109 n

Pearson [chi square] = 5.648 (p-value = .02, with 1df)

CUT = whether a subject considered a pay cut for reduced work hours,
coded 0 if subject did not consider taking a pay cut and 1 if
subject considered taking a pay cut.
GENDER = gender of subject with two levels, coded 0 for males and 1
for females.
CERTI = whether subject is certified with two levels, coded 0 if
subject is not certified and 1 if subject has certification (CPA or
CMA).

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE TO ASCII]

where [y.sub.ii] is the observed count and [m.sub.ii] is the
expected or estimated count, calculated as [m.sub.ii] =
[([y.sub.+I]) ([y.sub.i+])] [approximately equal to] n.