Workers With in Small Firms Chapter I

Workers With in Small Firms

Chapter I outlines the problems this research aims to address, namely an information gap that may, if filled, enhance employment for potential and existing workers with disabilities. This chapter defines the problem background, purpose of research, theoretical framework through which conclusions will be drawn from survey data gathered in the field, the research questions the survey instrument seeks to answer, the definition of terms those questions employ and limits and delimitations of the intended research. Once those parameters are outlined, the claim this argument attempts to support, that closing a gap in information describing satisfaction and productivity for workers with disability in small firms below conventional definitions of 500 workers or less may improve employment for a historically marginalized population, leads to conclusions that thus inform the subsequent methodological and analytical chapters.

Problem Background: The Uncashed 'Triple Paycheck'

Title I of the Americans with Disabilities Act (ADA) bans discrimination against persons with disabilities in employment (42 U.S.C. § 12111, in U.S. Commission on Civil Rights, 2000, n.p.). Yet the courts increasingly restricted definitions of disability and the "reasonable accommodations" necessary for full integration in the U.S. workforce subsequent to ADA's passage in 1990 to the degree that the U.S. Congress found it necessary in 2008 to reaffirm the intent and redefine the letter in order to achieve the broad mandate against discrimination ADA was originally set out to prohibit. The 2008 Americans with Disabilities Act Amendment Act empowered the Equal Employment Opportunity Commission (EEOC) to broaden rules based on the original ADA definitions and those new EEOC regulations have only recently seen their first year of implementation since May 24, 2011 (U.S. Department of Labor, 2011, n.p.). Yet the most recent available American Community Survey employment statistics show that for 2009, "an estimated 36.0% (plus or minus 0.31 percentage points) of non-institutionalized, male or female, with a disability, ages 21-64, all races, regardless of ethnicity, with all education levels in the United States were employed" (Erickson, Lee and von Schrader, 2011, n.p.).

At the same time, once unemployed workers with disabilities sign on to the federal Social Security and Disability Insurance supplemental income transfer program after their assets are depleted enough, and then also earn more than $810 per month in income, they become liable for the health care coverage Medicaid otherwise pays for before earnings up to that income level. Given private insurance exclusions for pre-existing conditions, the result is a powerful disincentive against earning more than $810 per month if out-of-pocket health care immediately negates any subsequent earnings beyond that level. An individual with disability wanting to enter the workforce and become independent of health- and income-related transfers must earn more than $810 per month plus their entire monthly health care cost at cash prices paid out of pocket, given current legal health insurance exclusion for the pre-existing condition that got them on the transfer program in the first place. Given eight times less employment compared to workers without disabilities in the U.S. labor force (U.S. Department of Labor, 2012a, n.p.) and earnings barely two-thirds of the civilian nondisabled, noninstitutionalized population (U.S. Census Bureau, n.d.), or incidence of workers with disabilities comprising twice the share of 'in the labor force' below federal poverty levels than working plus above poverty in the Atlanta-Sandy Springs-Marietta, GA Metro Area as of the 2005-07 American Community Survey (U.S. Census Bureau b), is it thus no surprise that potential workers on SSDI leave the transfer program for paid employment at a rate of about half of one percent per year (Tremblay, Porter, Smith and Weathers, 2011, p. 19)? Yet employing these workers would perhaps be one of the most lucrative investments society could make if financial independence for workers on SSDI generated a 'triple paycheck' comprising disposable income and the spending that implies, reduced direct tax transfers, and increasing resources for other social programs like for example supporting an upcoming wave of retiring seniors.

These are the trends this research seeks to counteract. Any contribution toward employing potential workers with disabilities should be encouraged as urgently as available resources allow, for these and other reasons. Yet while there is copious data and theoretical research achieving best practice in vocational rehabilitation and employment services from the academic, agency and helping professions sectors, the 'silver bullet' policy for job placement and retention for potential and existing workers with disability remains elusive, ADA Title I notwithstanding. At the same time, with the aging of the U.S. baby boomer generation, disabled veterans returning from foreign military campaigns, and sweeping national health care policy change before the U.S. Supreme Court within the month following this writing, the need for successful placement and retention increases rather than recedes every day. The volume of academic, government, industry and nonprofit-sector research continues to expand but not all studies or even government agencies define disability the same; not all employment statistics that include disability are published below the state level (U.S. Department of Labor, 2012b, n.p.), and not all local, regional or metropolitan statistical area employment statistics screen for disability. What information is to be had is often dated and/or focused at too widespread a level to necessarily relate local employment conditions where policy makers, advocates, employers and job-seekers can implement that information in the field (Fisher and Sousa-Poza, 2007, p. 4).

At the same time, macroeconomic events on the national and global level including obvious shocks like recent housing and credit crises but also necessary and useful but relentless technological and process innovation, including productivity-enhancing management or software innovation, or manufacturing automation or even trade policy, for that matter, presents a new and evolving workplace every day as well. The unprecedented magnitude of many recent global events undermine much of the canonical research precedent going back say before the advent of email, cell phones or even the 2008-era financial disruption and thus may not describe the workplace employers and job seekers encounter today. Advances in adaptive technology for workers with disability have changed so much even since President Bush signed ADA in 1990 for example as to render conventions of 'reasonable accommodation' older than a dozen years or so functionally if not necessarily theoretically obsolete. These limitations underscore the ongoing need for and utility of regional and local information of recent vintage and quality researchers can defend to robust statistical significance.

A further gap this research seeks to fill is the lack of differentiation between what are typically classified as "small to medium employers," firms employing under 500 (Linnan and Birken, 2006, p. 433). Lumping together firms under 500 employees may cloud more than it clarifies, if small employers are a significant source of even temporary job growth, or if smaller-than-500 firms do not face identical conditions or perform the same within that classification. Most of the existing statistics that include factors for disability or local employment like the federal Current Population Survey or American Community Survey fail to differentiate by employer size and disability, but it does not seem to stretch reason to speculate that a small enterprise of say five or ten stakeholders may face different challenges and opportunities than a local branch of a global multinational employing hundreds of thousands with corresponding assets and resources. Including for example self-employed entrepreneurs or a growing share of part-time workers with disability (Hothckiss, 2004, p. 25) in the group "under 500 employees" and then comparing that to a multinational like say Wal-Mart or IBM just for example demonstrates how misleading such classification could become. Yet very few academic or government theoretical or data sources differentiate by firm size below 500 employees if at all, although a few but growing body of research suggests smaller workforces can deliver enhanced productivity and job satisfaction for historically marginalized workers but at the same time have less resources for employee health care (Day and Greene, 2008, p. 644). The current study may contribute to a growing realization that generalizations above and below what many entrepreneurs would consider a huge firm at 500 employees, mask differences that bear investigation if they can increase employment for workers with disabilities in firms of any size.

This project seeks to provide such data but also support the assertion that the typical definition of 'SME' as under 500 employees glosses over potential 'low-hanging fruit' if for example employment in micro-enterprises provides higher job satisfaction and productivity than attempting to place workers with disability in more articulated, hierarchical multinationals. On the other hand perhaps the larger employers are in fact better able to accommodate diverse employees than struggling small employers, in which case placement agencies and vocational planners may increase performance from existing resources targeting larger, medium or smaller firms as results reveal. If this research does not exhaustively answer such questions in our region, supporting the utility of further inquiry may contribute if differences within the under-500 size and between those and Fortune 500-size employers convincingly emerge from robust, defensible evidence. Since ADA only applies to firms with over 15 employees for example, not all 'SME's by conventional definitions of 500 or under share even the same institutional constraints, let alone access to resources or economies of scale. The difference may be material for more than 18 million potential U.S. workers with disability (Erickson, Lee, and von Schrader, 2011, n.p.). Rarely-asked questions of whether workers disclose invisible disabilities to potential employers or if those conditions include only "registered disabilities" for example demonstrate how generalizations like "workers with disabilities" and/or (often at the same time) "small employers" constituting under 500 employees, may mislead perhaps even to the point of harm or foregone productivity from increasingly scarce resources. This project may help demonstrate the value of refining such definitions in an evolving workplace.

Hence the focus on productivity and job satisfaction in small firms, especially under ADA-threshold size, compared to increasingly larger employers. Exhaustive definitions of productivity and job satisfaction are about as elusive as individual definitions and perceptions of disability, but the correlation between job satisfaction, productivity, and the foregone social cost from lost earnings, hiring and training cost, job search, unemployment insurance and vocational support programs et cetera has been demonstrated throughout the industrial organization and management literature since the early 1950s at least (Mount, Ilies and Johnson, 2006, p. 604). When people like their jobs, they are more productive, however defined, the consensus agrees, although measuring productivity and satisfaction often reduce to the semantic problem of defining words in terms of other words (Mount, Ilies and Johnson, 2006, p. 599). Productivity and satisfaction also face confounds where for example higher productivity leads to higher satisfaction, which leads to higher productivity. In many studies, better health leads to higher productivity and satisfaction, which then promote increased perceptions of health (Fisher and Sousa-Poza, 2007, p. 3). This research will not attempt to solve such potentially inherent paradoxes but may contribute toward better benchmarks where instrument items and Likert-scale values mask subjective definitions that can be punctured with the most perfunctory analysis (Jones, Jones, Latreille and Sloane, 2008, p. 14).

Without solving the problem of individual perceptions of the value of "5" on a Likert scale of 1-7 or the definition of "like" in "I like my job" for every worker, demonstrating higher performance and satisfaction through innovations on metrics modeled off of widely-cited precedent has value field testing a more accurate yardstick if results deliver interesting and defensible improvement. These methodological contributions are peripheral to the core objective of increasing employment for workers with disability in the Atlanta region by identifying performance advantages by firm size for those subjects, but if this research contributes to a better measurement, it will join a growing body of theory that recognizes the need for finer generalizations than "with a disability" or SME including under 500 employees when the difference between a sole proprietor and even 30 employees may be effectively the same as one employing 30,000, and not all workers with disabilities ever even disclose those to employers, plausibly perhaps out of fear of triple unemployment rates and 2/3 earnings of those who do not or can get by without disclosure.

Study Purpose

The purpose of this study is to fill a gap in information describing the perceptions of job satisfaction and productivity in workers with disabilities in the Atlanta-Sandy Springs-Marietta metropolitan statistical area. The purpose of providing this information is to enhance the ability of employers, job seekers, agency and vocational planners to achieve stable, fulfilling and productive employment for potential workers with disabilities, and also to explore the potential utility of further research considering performance of firms smaller than 500 employees.

Theoretical Framework

The theoretical frameworks this research will rely on are well-established. This dissertation attempts no innovation to canonical practices but will employ inferential statistics using Statistical Package for the Social Sciences (SPSS) to test hypotheses by a variety of methods. Such diversity aims toward redundancy given the potential confounds that often undermine parametric analysis. Analysis of Variance (ANOVA) for example relies on assumptions of independence, normality and identical variance within groups for comparing deviation around means between the groups (Rosenholtz, 2004a, p. 3). SPSS calculates a vast array of "post hoc" statistics that measure the validity of these assumptions, and thus support or refute conclusions suggested by parametric tests like ANOVA. At the same time and as part of the analysis of variance, SPSS has the capacity to run multiple cross-tabulations of for example Pearson's Product-Moment correlation between variables of interest such that the output presents a tabular matrix allowing quick identification of potential relationships for further study. This screening capability is augmented by visual inspection of means plots graphed between two variables, which allows for quick, informal assessment of potential relationships between characteristics. If the likelihood that samples with particular characteristics would arise in unexpected groupings simply by chance becomes low enough, then probability theory describes how unlikely such results should occur by accident and therefore suggests two (or more) groups really differ. Groups in this case would be responses on variables like job satisfaction, presence of disability, number of employees at work and demographic variables like age or gender. If workers with disability in large and small firms do not differ in productivity for example, then a diverse array of statistical procedures support how strongly such assertion can be defended.

The hypotheses in this inquiry are deliberately set up to provide such information. Comparing means and other parameters like deviation (variance) and spread around means and medians (skew, kurtosis) will establish whether groups or classes fulfill the requirements of normality, independence and similar enough variance that ANOVA results are valid. These assumptions will be supported with post-hoc tests including but not restricted to Tukey's Honestly Significant Difference (Tukey's HSD), Levene's Statistic, ?-squared goodness of fit, and Cronbach's Alpha for example along with other indicators of kurtosis (shape of distribution) and skew. In case data turn out to be too skewed or show too much difference in variance between classes to defend ANOVA results, for example if say the youngest cohort by age had enough difference in variance compared to older workers on a dependent variable like job satisfaction to undermine assumptions underlying claims from ANOVA (indicated by Levene's statistic and also ? -squared 'goodness of fit'), the commands to SPSS will include non-parametric inferential analysis. In short, the analyst will command SPSS to run multiple, in some cases redundant and thus supporting analyses, and also the post-hoc tests that will demonstrate compliance with assumptions underlying parametric analysis, and also non-parametric analyses at the same time including graphic relationships between variables and their resulting confidence intervals all with several clicks of a mouse in the SPSS program. Relationships indicated by ANOVA will then be tested for strength and direction using correlation or regression as variable characteristics entail.

Non-parametric statistics like for example Wilcoxon's Rank-Sum test allow for comparing nonparametric classes like 'first, second, third' against continuous variables like scores from one to ten. Given the possibility of outliers that may introduce skew within the various groups however, like for example firm size where there is a lower bound of one employee but no upper bound for number of employees above 500, this inquiry limits groups in the category 'firm size' to discrete units that will hopefully satisfy the stringent normality requirements measured by Levene's statistic, Tukey's HSD and/or ? -squared goodness of fit tests for example within the various classes of firm size. Should the necessary assumptions underlying ANOVA be violated, then inference by for example regression and/or two-way ? -squared test for independence will support or refute non-parametric relationships to the point this research will still be able to claim that workers with and without disabilities do or do not perceive the same or different levels of job satisfaction or productivity in small firms than do those in large firms or without disability. If one class of respondents differ enough in response to a second, third or 'n' variable(s), then statistical inference allows researchers to support the unlikelihood of such accidental results at various levels of error using 'two (or more)-way' or 'factorial' procedures that measure interaction between more than two variables. If control groups from the general population for example are interesting compared to say workers with and without disabilities, or if multiple dependent variables for one independent factor show potential correlation or dependence, MANOVA and ANCOVA variants of ANOVA can also test those relationships simply by assigning SPSS such a task. These tests allow for presentation of main effect direction and strength between multiple variables for the same independent variable, which could also be identified through multiple regression but with far more labor in the absence of statistical modeling programs like SPSS.

As mentioned, parametric statistics require similar enough variance (dispersion about the mean) within categories in order to demonstrate difference from the total sample mean (all groups), i.e. true difference from model or instrumental error. The Central Limit Theorem explains that given large enough sample sizes, variance within all classes will approach normal distribution but if not Students' t-distribution approximates the norm the larger grows the sample size (Rosenholtz, 2004b, p. 33). This therefore suggests that if limiting the survey instrument to fewer question items allows wider participation and thus increasing sample size through subject convenience, then a shorter survey should deliver higher participation rates that will then drive variance within answer cohorts toward normal distribution and thus allow for ANOVA-type analysis of variance. Parametric tests deliver stronger results than non-parametric statistics, but nonparametrics can however be used in place of parametric statistics should the assumptions of normality within classes etc. not hold true, or also at the same time to support claims via analysis of variance should the data approach normality as they eventually do given large enough sample size. If a shorter survey induces higher participation, the resulting greater sample size will allow for stronger, more defensible claims through inferential hypothesis testing. The reason of course for setting up all this multiple redundancy in writing hypotheses from the outset is to avoid wasting scarce resources collecting useless data. On the contrary: Even given no more than a baker's dozen parameters on the survey instrument, exhaustive analysis of all data relationships will exceed the space for discussion in this report and hopefully thus leave information for further research, if gathered in robust enough manner as to be defensible; hence the attention to achieving required assumptions underlying parametric statistics, to provide opportunity for further analysis with minimal research cost.

In summary, means testing attempts to compare groups with each other and with the total sample in order to identify how likely similarities and differences would arise by simple chance. This requires enough data points within each group (and in total), and for parametric statistics enough similarity in variance within the subcategories, as well as assumptions of independence between survey questions and respondents. The most stringent normality would require absolutely random sampling which exceeds available resources due to the extensive sample size fulfilling that assumption would require absent some form of compulsion, but Students' t-distribution approximates the norm given large enough convenience sampling where such constraints are acknowledged. SPSS allows the researcher to nearly-effortlessly run a vast sequence of parametric and non-parametric statistical tests, which identify and verify potential similarities, relationships and independence or differences between groups to defend claims those would not occur in even an exhaustive sampling of the entire universe of data toward which such inference attempts to generalize. The result will allow in this case determination of similarities and differences between workers with and without disabilities in small firms and large firms, and between workers with disabilities and without in small or large firms, which then can be supported for robustness using post-hoc tests whether those results are parametric or nonparametric. If the required parametric assumptions prove invalid, non-parametric inferential statistics will still generate robust ability to test hypotheses regarding the variables of interest, job satisfaction and productivity (as defined) for workers with disabilities in small firms.

Research Questions (RQs) and Hypotheses

RQ1: Do workers with disabilities have higher, lower or the same perceptions of job satisfaction at larger or smaller firms?

H1: Workers with self-perceived disability will report higher job satisfaction in small firms than large firms.

RQ2: Do workers with disabilities have higher, lower or the same perceptions of productivity at larger or smaller firms?

H2: Workers with self-perceived disability will report higher productivity in small firms compared to large firms.

RQ3: Do workers with self-reported disabilities have higher, lower or the same perceptions of job satisfaction than workers without disabilities in smaller firms or larger firms?

H3: Workers with self-reported disabilities will have higher perceptions of job satisfaction than workers without disabilities in firms of all size.

RQ4: Do workers with self-assessed disabilities have higher, lower or the same perceptions of productivity compared to workers without disabilities in small firms or larger firms?

H4: Workers with self-reported disabilities will indicate higher productivity than workers without disabilities in firms of all size.

Definition of Terms

Disability will be defined by subjects, on grounds that not all disabilities are registered or even acknowledged to employers, with nearly triple unemployment rates providing powerful disincentive for such reporting where that can be avoided. Further, ADA covers the perception of disability, although perception itself is not grounds for reasonable accommodation (42 U.S.C. § 12111, in U.S. Commission on Civil Rights, 2000, n.p.). Productivity is defined indirectly by lack of missed work on grounds that less absenteeism should result in greater output regardless of product. "Work" includes paid work and labor required to bring sellable goods to market.

Wage level is also asked as productivity measurement on the questionable thesis that higher output eventually achieves higher wages. This assumption is theoretical and may not empirically hold in every case, but if not, the result is that productivity is not captured by wages in which case the content of study effectively defies statistical description if there is no causal linkage between productivity and its resulting wage. Many studies, some referenced in the review of literature, demonstrate productivity is defined in many subjective ways and controlling for quality of output for example defies the resources available for this research. Including the answer option 'per piece' in the survey instrument for pay attempts to include a sector that may be expanding as some claim part-time employment is (Hotchkiss, 2004, p. 25).. Likewise several questions triangulate "satisfaction" through search for different work or willingness to switch, due to varying personal valuations of satisfaction; intrapersonal value of foregone activity; and the problems of identifying a universal metric for utility or the value of time or discomfort. Remaining terms particularly those used in statistical inference will be explained in their context closer to results they describe rather than this section in order to improve coherence / reduce referencing effort.