the impact of adoption of wrongful discharge laws on union

1

The Impact of Adoption of Wrongful Discharge Laws on Union Density: 1983-2014

Job Market Paper

31 October, 2016

Eric Hoyt, University of Massachusetts Amherst

Abstract: Union membership has declined precipitously in recent decades following the rise of

retaliatory firing and other harsh union avoidance tactics. This paper utilizes variation in the

time and location (i.e. state) of adoption of wrongful-discharge court doctrines in the U.S. over

the 1980s and 1990s as a source of numerous natural experiments regarding the impact of

employment protection on union membership and union coverage outcomes. This analysis

finds that adoption of one wrongful discharge law in particular, the implied contract doctrine, is

associated with small but consistent increases in union density, driven by increases in union

membership and coverage levels. The impact is greatest for the private sector, manufacturing

industry, and younger male and female workers with less than college education. This paper

proposes that the implied contract doctrine boosts union density by lowering the cost of seeking

union membership and coverage. The findings also suggest that employment protection policies

may foster greater equity in labor market outcomes by age and sex, as well as serve as a

counterweight to the pressures undermining union density over this period.

Keywords: labor unions, union density, wrongful discharge laws, employment protection

Acknowledgements: I am grateful for the guidance of Professors Fidan Kurtulus, Gerald

Friedman, and Eve Weinbaum as well as for the numerous helpful comments of professors and

fellow graduate students at the University of Massachusetts Amherst.

Author Contacts: [email protected], 200 Hicks Way University of Massachusetts

Amherst, Amherst, MA 01003, (413)-362-4517

3

I. Introduction

In the United States, wrongful-discharge common law doctrines have been growing in

importance as state courts have followed a trend of precedent-setting employment law rulings.

These decisions, though occurring with significant variation between states in timing of adoption

and broadness of applicability, have nonetheless altered the basic institutional foundation

underlying the U.S. labor market since the late 19th

century, the common-law doctrine of

employment-at-will. Employment-at-will is the legal principle which allows employers to

dismiss their employees for any reason or no reason at all.

A vibrant economic literature has arisen in recent decades concerning the impact of wrongful

discharge laws, and employment protection legislation more generally, on employment, wages,

productivity, the use of temporary workers, job flows, and hiring rates, among other labor market

outcomes. A contemporaneous literature has emerged within the labor movement seeking to

understand the dramatic fall in union membership and union economic influence in the U.S.

economy which began in the years immediately following WWII and has accelerated in recent

decades. This work is the first to investigate the impact of wrongful-discharge laws on union

membership and coverage outcomes, adding a previously unexamined dimension to the literature

on the labor market ramifications of wrongful discharge laws, while contributing a rigorous

analysis of the role played by employment law changes in mediating union membership trends to

the literature on union decline.

The literature on the labor market effects of wrongful discharge laws is founded on the

theoretical model articulated by Lazear (1990) and Blanchard and Katz (1997), which predicts

that employment protection policies, by increasing firing costs, have an indeterminate effect on

4

employment levels. That is, since employment protection makes it more difficult for employers

to fire workers, it slows both the rate of dismissal and the rate of hiring. The theory predicts that

in the short-run employment levels may increase, decrease, or be unaffected, depending on the

particular labor demand conditions in the economy. However, in the long run, it is predicted, as

Autor et. al. (2006) and others have noted, that employment levels and wages may fall if rising

labor costs caused by employment protection policies are not offset by productivity gains.

One method of analysis, adopted most notably by Autor, Donahue III, and Schwab (2006),

Miles (2000), Dertouzos and Karoly (1992, 1993), and Kugler and St. Paul (2004), and which is

the method used in this paper, utilizes variation in the time and location (i.e. state) of adoption of

wrongful-discharge court doctrines over the 1980s and 1990s as a source of numerous natural

experiments regarding the impact of employment protection on labor market outcomes. Autor

et. al. (2006), with monthly outcome variables for employment, wages, and explanatory

variables for policy adoptions, come to the conclusion that while adoption of two of the three

main wrongful discharge policies has no statistically significant effect on employment outcomes,

adoption of the implied contract doctrine, , which creates an exemption to employment-at-will

through counting employers’ verbal and written promises of job protection as contractually

binding, has moderate and consistent negative effects on employment. The authors find no

significant effect on wages. Further, they find that the effect of the implied contract doctrine is

felt strongest in the short-term by those most likely to switch jobs often, such as women and less

educated workers. They found that the long-run impact was felt strongest by older and more

highly educated workers. They also found stronger negative impacts of the implied contract

doctrine on manufacturing as opposed to nonmanufacturing employment.

5

Kugler and Saint-Paul (2004) reveal disemployment results consistent with an explanation

based on asymmetric information and employee-employer matching. They find evidence that

wrongful discharge doctrines motivate a process of adverse selection in which employers

discriminate in hiring against the unemployed. They argue that employers, since they have

imperfect information about employees, assume that workers who are already unemployed are

more likely to be “lemons”, or bad investments from the perspective of the firm, because their

status might reflect non-observable qualities such as a lack of motivation or work effort. The

authors predict that, after firing costs rise following the adoption of wrongful-discharge laws,

employers may become even less likely to hire unemployed workers. They find that this effect is

lessened for union members, whose status as unemployed is mediated by seniority and less likely

to reflect unobservable worker characteristics. Kugler’s results are similar to Autor et. al (2006)

in that they points toward stronger disemployment effects for precarious labor market

participants.

Autor (2003), while linking the adoption of wrongful discharge laws with greater use of

temporary help services, notes that states with slower union decline also witnessed greater

utilization of temporary workers and suggests that this reflects employers’ efforts to avoid costs

of union wage premia. Without discounting this highly reasonable line of argument, my analysis

contributes another angle to the correlation between wrongful discharge doctrines, unions, and

greater utilization of temporary workers; that is, my paper suggests that slower union decline

may itself be partially caused by passage of wrongful discharge laws.

My analysis adds to this literature by showing that adoption of one wrongful discharge law in

particular, the implied contract doctrine, is associated with modest but consistent increases in

union membership and coverage density. The greatest increase in union membership and

6

coverage rates occurs for the private sector (0.84% and 1.03%), manufacturing industry (1.39%

and 1.59%), and younger male workers with less than college education (1.91% and 1.97%).

Significant increases in union membership and coverage levels in the private sector (5.37% and

6.34%), manufacturing industry (8.06% and 9.15%), and for male workers with less than college

education (11.34% and 11.38%) appear to be the main driver of the rise in union density. The

implied contract doctrine is also associated with increases in union density for young female

workers with the less than college education (0.74% and 1.74%), but this result appears

explained by an only marginally significant 10.19% increase in the level of union coverage and a

nonsignificant increase in union membership levels of 4.55%.

One could reasonably suspect that these results point towards reverse causality because

greater union growth or decline over this period might increase the probability of states’ adoption

of wrongful discharge doctrines. Strong and growing unions could be more likely to demand job

security throughout the labor market, and large losses in collectively bargained job security in a

state could spur demands for broad employment protection. However, Autor (2003) and Miles

(2000) find that neither above average union growth or union decline is associated with greater

probability of states’ adoption of wrongful discharge doctrines.

My results seem to contradict previous findings regarding little or no impact of wrongful

discharge laws on the employment levels of unionized workers, given they are already covered

under employment protections within collective bargaining agreements. However, this analysis

proposes that the implied contract doctrine boosts union density, membership, and coverage

levels by lowering the costs to workers of seeking union membership and coverage, rather than

through an impact on employment levels. Some evidence emerges of statistically significant

disemployment effects for nonmembers and noncovered workers within particular industry and

7

demographic subgroups, but none of these correspond, except perhaps in the case of the

construction industry, to the instances of union density growth found within this analysis. This is

especially important since an increase in union density could occur if the numerator, consisting

of the level of employed union members (or covered workers), remains relatively fixed, while

the denominator, consisting of the sum of the level of employed union members and

nonmembers (or covered and noncovered workers), declines as result of employment losses for

nonmembers (or noncovered workers).

In the context of large and precipitous decline in union membership and coverage, and given

the significant role of retaliatory firings in propelling this development1, the finding in this paper

on union density within the private sector, manufacturing industry, and among less educated and

younger men and women suggests that employment protection policies may serve as a

counterweight to the pressures undermining union membership. Since the private sector and

manufacturing industry experienced some of the largest declines in union membership and

coverage over this period, and younger men and women have been historically underrepresented

among union members, these findings suggest that employment protection may also be one way

to grow union membership within, as well as beyond, its traditional base.

II. Institutional Background of Wrongful Discharge Laws

Before proceeding further, it is worthwhile to comment briefly on the institutional

background of wrongful-discharge doctrines and union density in the United States. There have

been three primary wrongful-discharge doctrines that have proliferated in recent decades: the

1 Researchers have found that since the 1980s nearly 1-in-4 union elections is marred by retaliatory discharge, and

pro-union workers face a 2percent-3percent chance of suffering retaliatory discharge (Schmitt and Zipperer 2007,

LaLonde and Meltzer 1991, and Weiler 1983). Even following successful union representation elections, research

shows that only 48percent of organized workplaces have a collective bargaining agreement one year following the

election (Bronfenbrenner 2009).

8

implied contract doctrine, the public policy doctrine, and the covenant of good faith and fair

dealing.

The implied contract doctrine states that language in employee manuals and oral promises

made by supervisors detailing a specific duration of employment or standard procedures for

dismissal can override the default of employment-at-will. In some states, the implied contract

doctrine arises in the absence of written or oral promises even when employers’ actions establish

a repeated past practice of discharging only for justified reasons (Hirsch, Secunda, and Bales

2013). Toussaint v. Blue Cross Blue Shield of Michigan is frequently cited as one of the best

cases, though not the first, to articulate the implied contract doctrine. Charles Toussaint worked

for many years for his employer Blue Cross Blue Shield. When he started the position, he was

given told that his employment was secure as long as he preformed his job well. After he was

fired without an explanation, Toussaint sued for wrongful discharge. The Michigan Supreme

Court ruled that his employer’s statement, as well as language in the company’s employee

handbook, constituted legally enforceable guarantees of dismissal for “good cause” (Muhl 2001).

The implied contract doctrine, as it has developed in practice, is the most broadly applicable of

wrongful discharge laws, since it applies to every discharge within a workplace, creating a “good

cause” standard rather than prohibiting particular “bad faith” discharges. The damages awarded

to litigants are contractual, meaning employers owe the employee the salary and wages they

would have paid that employee had they continued working since the date of discharge, less any

amount of wages and salary the employee was able to earn in new employment. In many late-

adopting states, and in certain states that initially adopted strong interpretations of the implied

contract doctrine, courts have weakened this policy by sanctioning employers’ use of disclaimers

9

in employee handbooks which counteract implied contractual obligations by explicitly

announcing to employees their status as at-will.

The public policy doctrine is an exception to the at-will rule that arises when an employee is

discharged after invoking a publicly protected right or obligation, such as filing a worker’s

compensation claim or attending jury duty. The doctrine also arises when an employee is fired

for refusing an employer’s request to violate public policy, ranging from a clear infraction like

investment fraud to broader like endangering public health (Hirsch, Secunda, and Bales 2013).

The first case to illustrate the public policy doctrine was decided in August of 1959 in California.

An employee of the Teamsters labor union was terminated immediately after having provided

truthful testimony to a state governmental body on internal corruption within the union, violating

instructions given by his employer to deny such accusations under oath. The court ruled in his

favor when he challenged the dismissal, establishing a general principle that has been developed

further in court rulings (Muhl 2001). The doctrine permits tort damage awards for successful

litigants. These damages can be much larger than the contract damages of the implied contract

doctrine, as they can also include punitive damages and legal fees, for instance, in addition to the

employees’ lost wages and salary. The per litigant cost to employers is larger than for the

implied contract doctrine. Tort damages are included as an additional disincentive to employers

since the harm inflicted is both to the individual discharged employee as well as to the broader

community. Since cases involving the public policy doctrine are less common than those arising

under the implied contract doctrine, given they pertain only to particular instances of discharge

for “bad cause”, it is likely that the overall cost of the doctrine to employers is smaller than that

of the implied contract doctrine.

10

Finally, the good faith doctrine, if followed to the letter, requires that every discharge

decision be justified. However, in practice the good faith doctrine has developed, like the public

policy doctrine, only to prohibit particular cases of morally reprehensible termination. For

instance, one of the first court decisions to adopt the good faith doctrine was Fortune vs.

National Cash Register Co. in Massachusetts in 1977. Fortune was a sales employee who just

after finalizing a deal entitling him to a substantial commission was discharged for a period of

time during which he would have received payment. He was rehired shortly afterwards. The

company structured its commission payouts to accrue to its sales personnel in two installments:

when a deal was first signed with a customer, and then upon final delivery of the goods. The

company discharged the salesperson without justification after the deal was signed but before the

delivery of goods, leaving the employee with less of the commission owed to him. Fortune sued

for wrongful termination, the Massachusetts Supreme Court affirmed that that he was discharged

in bad faith, and he was awarded damages. Cases involving the good faith doctrine, like the

public policy doctrine, are relatively rare since they prohibit specific instances of “bad cause”

discharge. While tort damages were permitted in most early interpretations of the doctrine, the

policy has developed in practice to limit litigant awards to contract damages in most adopting

states. Thus, since cases involving the good faith doctrine are less common than the implied

contract doctrine yet entail the same per litigant costs to employers, it seems likely this doctrine

has the smallest overall cost to employers (Hirsch, Secunda, and Bales 2013).

III. Data Sources

This paper uses state court decisions from 1970 through 2014 to construct explanatory

variables on wrongful discharge doctrine adoptions. I employ the method of Morris (1995),

which was adopted by Autor et. al. (2006) for the years 1970 through 1999, by coding the first

11

state Supreme Court or Intermediate Appellate court ruling that demonstrates a clear acceptance

of each of the three wrongful discharge doctrines.2 While Autor et al. (2006) utilize this data to

construct variables on a monthly basis from the year 1978 through 1999, this paper, in order to

match union membership and coverage data available only at the yearly level for all states, and

only from 1983 onwards, constructs annual policy adoption variables for 1983 until 2014, the

most recent year of available union data. The data set over this period includes the sustained

(i.e. at least five year long) adoption of the implied contract doctrine by 32 states, the public

policy doctrine by 26 states, and the good faith doctrine by 6 states. Table 1 includes a complete

list of all wrongful discharge doctrine recognitions by state as of 2014, the last year of the study,

as well as the closest year3 to the date of policy adoption or repeal. Maps 1 through 3 give a

geographic snapshot of state recognition of the three wrongful discharge doctrines as of 2014.

Table 1 and Maps 1 through 3 show the complete list of recognitions, adoptions, and repeals

stretching back to 1960, long before the period of analysis in this paper. They show a total of 41

2 While I first drew solely upon the legal adoption data for 1970 through 1999 constructed by Autor et. al. (2006)

and are publicly accessible on David Autor’s MIT faculty website,

http://economics.mit.edu/faculty/dautor/data/autdonschw06, and were first accessed on November 14, 2012, I

updated the legal adoptions to 2014 based on my own analysis of the case law. My analysis conforms to Morris

(1995), but differs from Autor et. al. (2006), in counting the implied contract doctrine as being adopted without

subsequent repeal in Arizona from 1983 onwards. My coding differs from Autor et. al. (2006) in not coding

Louisiana as having adopted the good faith doctrine in 1998. Autor et. al. (2006) may have misclassified Barbe v.

A.A. Harmon & Co. (1998) as an adoption of a good faith exception to employment at-will when the text of the case

decision shows that it only created a good faith covenant in the provision of employee bonuses. Finally, my reading

of the case law and secondary literature made clear that the New Hampshire Supreme Court, in its Centronics V.

Genicom (1989) decision, readopted the good faith doctrine. This finding contradicts Morris(1995) and Autor et. al.

(2006) as they seem to have both missed this development in their legal coding.

3 In order to approximate the detail of Autor et. al. (2006) and Morris (1995)’s data regarding month of policy, I

lump adoptions into the calendar year of passage if occurring from January through July, and into the following

calendar year for policy adoptions occurring during July through December. For instance, the adoption by New

Hampshire of the implied contract exception in September of 1988, because it fell in the second half of the calendar

year and was technically closer to January 1989 than January 1988, was classified as occurring in the year 1989. As

one sees, for instance in Appendix A of Autor et. al. (2006), the actual date of the court case decision in this case is

in August of 1988 in New Hampshire, however I follow the method of Autor et. al. (2006) in classifying passage as

taking effect in the first full month following the court decision, September 1988, for this and all adoptions when

observed at the monthly-level of analysis.

12

adoptions of the implied contract doctrine, 43adoptions of the public policy doctrine, and 13

adoptions of the good faith doctrine as of 2014. Only Florida, Georgia, Louisiana, and Rhode

Island never adopted any wrongful discharge doctrines. Also, no change occurs in states’

recognition of wrongful discharge doctrines from the adoption of the good faith doctrine in

January 1994 through the end of the analysis in December 2014.

Yearly state-level union membership and coverage outcome variables for the years 1983 to

2014 were constructed from the Current Population Survey Outgoing Rotation Group monthly

files following the methodology of economists Barry Hirsh and David Macphearson (2003).

Monthly microdata data was extracted and then averaged, first over all individuals in each state

for each month, and then over each year in each state, for the number of employed wage and

salary workers over aged 16 and below 65 years of age, the subset of these workers who are

union members, and the subset of these workers who are covered by union collective bargaining

agreements. These figures were constructed for total state workforces, for public and private

sectors, for major industrial categories, and for eight sex-age-educational attainment

demographic categories. This data was then used to construct yearly state union membership

and coverage density figures for each state over the period 1983 to 2014, per the following

formula:

% Memst = (Membersst/Employmentst)100

% Covst = (Coveredst/Employmentst)100

This study utilizes the union membership rate and union coverage rate, defined as the %age of

employed wage and salary employees who are union members and %age who are covered by a

union collective bargaining agreement, in order to capture the broadest possible window into the

13

impact of wrongful discharge laws on union density. The subscripts s and t on the union density

measurers, as in the econometric specifications in the following section, indicate the relevant

state and year for each observation. While multiple attempts were made to extract and construct

figures by race and ethnicity, zero-valued union membership and coverage variables, even using

the broadest (i.e. non-white vs. white or nonhispanic vs. hispanic) demographic categories,

persisted for several states during multiple years over the period 1983 to 2014. In this paper,

outcome variables for the levels of nonmembership and noncoverage are constructed from the

union membership and coverage level data by subtracting from the total, sectoral, industrial, and

demographic group aggregate employment levels the respective levels of union membership and

coverage for each category.

IV. Econometric Models

Firstly, to give a picture into the dynamics of the impact of wrongful discharge doctrines on

union membership and coverage outcomes over the year surrounding adoption in treatment

relative nonadopting control states, this paper utilizes, before delving into the full fixed effects

specification, a more base line differences-in-differences regression with yearly leads and lags

from two years before to two years after adoption, which is nearly identical to the dynamic

specification included in Autor et. al. (2006). The model is

Yst = αs + δt + Σ2

τ= -2 γτ Ls,t-τ + δt x Regions+ ԑst (1)

Yst is the outcome variable which, in this specification, captures the % change in union

membership and union coverage rates in a state in each year. The variable αs is a state intercept

that captures idiosyncratic, unobserved, and pre-existing features of each relevant state which

might impact the change in union outcomes irrespective of the year of observation. δt represents

14

year fixed effects which control for yearly state-invariant shifts in the regression intercept,

accounting for common national shocks to union outcomes. δt x Regions is a term composed of

fixed effects representing the four U.S. Census regions interacted with the year fixed effects in

order to control for passing regional shocks to union membership and coverage outcomes. The

dummy variable Ls,t switches on from zero to one only during the year a state court adopts a

wrongful discharge doctrine, and permits the contemporaneous estimation of the impact of each

doctrine on union density in adopting states relative to nonadopting states. The parameter of

interest γτ gives the first-year impact of the adoption of wrongful discharge laws on the average

change in union membership and coverage outcomes between adopting and nonadopting states.

The specification is weighted by the share of national population aged 16 to 64 in each state in

the given year, and uses Huber-White robust standard errors. Identification of γτ comes from

within-region variation in union membership and coverage outcomes between adopting and

nonadopting states.

This specification is used to capture the impact of wrongful discharge laws on union density

for young males with less than college education, to give a representative view of the dynamics

surrounding policy adoption. A much broader investigation of outcomes by sector, industry, and

demographic groups is reserved for regression of the full fixed effects specification. In order to

match the results for the full fixed-effects regression, the overall sample window of adoptions

analyzed was restricted to 1985 to 2011. By omitting analysis on adoptions from before 1985, as

well as data on one and two year lags for adoptions before 1985, like with regressions using

equation 2, the overall number of adoptions actually studied becomes 19 implied contract

doctrine adoptions, 21 public policy doctrine adoptions, and 6 good faith doctrine adoptions.

15

The treatment observations in the full fixed effects difference-in-differences regression

analysis, following equation 1 of Autor et. al. (2006), consist of the five years surrounding policy

adoption of one of the three wrongful discharge doctrines, from two years before to two years

after the year of policy change. The control groups are composed of the states which do not

adopt any of the three wrongful discharge laws over the five years surrounding each adoption in

a treatment state. States which adopt a wrongful discharge doctrine before or after the five-year

period surrounding policy change in a treatment state enter the control group for that adoption.

The five-year treatment and control window is used instead of a panel composed over the entire

sample period to account for possible serial correlation in yearly state union membership and

coverage outcome data, as well as to identify discrete changes in the outcome variables

attributable to policy adoption.

The full fixed effects difference-in-differences regression specification is:

Yst = αs + αs * Treatst + β1Postst + β2TreatstPostst + β3 Postpostst + δt + δt x Regions+ ԑst (2)

Treatst is an indicator for the period 2 years before to 2 years after the year of adoption of a

wrongful-discharge law in treatment state s. Postst is an indicator for the period from the year of

adoption4 to 2 years after the year of adoption for all states, both adopting and nonadopting.

TreatstPostst is an interaction term representing the three year period following policy adoption,

in the treatment state only, beginning with the first year of adoption. The coefficient of interest

β2 is an estimate of the pre-post change in the outcome variable in adopting states relative to the

corresponding change in non-adopting states. Postpostst is a dummy variable that turns on for

4 This analysis does not retain the “doughnut hole” restriction of Autor et. al. (2006); that is, it does not omit data for

the first twelve months immediately following policy adoption to account for a response period during which

employers become aware and begin to react to the policy change. My analysis is yearly so this restriction based on

implementation delay over the first few months did not seem appropriate.

16

each state that reenters the control group 3 years after the year of policy adoption to account for

any bias caused by the enduring effect of the policy adoption after the five-year treatment

window expires. These window lengths are the same as the windows in Autor et. al. (2006) for

consistency.5

Yst, the outcome variable6, is identical to that in equation 1, with the exception that it is also

represents the approximate %age change in the level of union membership (i.e. the change in the

natural logarithm of the level of union membership), the approximate %age change in the level

of nonmembership, the approximate %age change in the level of union coverage, and the

approximate %age change in the level of noncoverage over sectors, industries, and sex-age-

education groups. The variables αs, δt , and δt x Regions are defined identically as for equation

5 To illustrate the legal doctrine adoption variables, consider the following example. Massachusetts adopted the

implied contract doctrine in 1988, while neighboring Rhode Island did not adopt any wrongful discharge law over

the period studied. Massachusetts is a treatment state, so the variable Treatst takes on the value one for the years

1986, 1987, 1988, 1989, and 1990, and takes on the value of zero in years before 1986 and after 1990. Since Rhode

Island did not adopt any wrongful discharge law over the five years surrounding adoption of the implied contract

doctrine in Massachusetts, Treatst always remains zero in this state. In this case, Rhode Island is included within the

control group. The variable Postst takes on the value one in Massachusetts and Rhode Island, and in all other states,

for all years that this policy remains in effect in Massachusetts, beginning with the year of adoption, which in this

case is the year 1988 to the end of the dataset in 2014. Treatst* Postst is an interaction term representing the three

year period, in the treatment state only (i.e. Massachusetts), beginning with the year of adoption of the implied

contract doctrine in 1988. In other words, Treatst* Postst takes on the value one only in Massachusetts and only for

the years 1988, 1989, and 1990. Maine, another neighboring state, also did not adopt any wrongful discharge laws

over the five years surrounding 1988, so it is included within the control group. However, Maine adopted the

implied contract doctrine in 1978. In this case, the variable PoststPostst switches from zero to one for Maine starting

the year after the treatment window for its adoption ended, which is 1981, and takes on the value one for the rest of

sample period, that is, until 2014. If Massachusetts becomes eligible as a control for a later state adoption, its

PoststPostst variable, which switches on beginning in 1991, would take on the same role as it plays in Maine in this

example. The illustration described here is summarized in Table 2. One important consequence of this five year

treatment window structure for the fixed effect specification is that it omits analysis of policy adoptions occurring

from 1983 and 1984, and from 2012 to 2014.

6 The outcome variables for the full fixed effect specifications are investigated over a slightly broader range of

economic categories than in Autor et. al. (2006), expanding beyond an industrial breakdown of manufacturing and

nonmanufacturing to include construction, service, and transportation-communication-utilities industries, and by

including analysis at the level of economic sector. The eight sex-age-educational attainment demographic categories

are nearly identical to those used by Autor et. al. (2006), categorizing young workers as aged 16 to 39 as opposed to

18 to 39 in order to conform to the method of union membership and coverage data construction of Hirsch and

Macphearson (2003).

17

1. The term αs * Treatst is a state-time specific intercept that accounts for changes in union

membership and coverage rates, and the levels of union membership, nonmembership, coverage

and noncoverage attributable to idiosyncratic, unobserved, and pre-existing features of the

relevant treatment state over the respective five-year treatment window in that state. As with

equation 1, each regression is weighted by the share of national population aged 16 to 64 in each

state in the given year, and Huber-White robust standard errors are used to account for within-

state error correlations. Identification of β2 comes from variation in union membership and

coverage outcomes between adopting and nonadopting states within the same census regions.

This specification gives an estimate of the causal effect of policy adoption controlling for a

variety of state, regional, national, and time-specific characteristics that may influence union

membership and coverage outcomes.

V. Results

A. Event Study Difference-in-Differences Regression

Figure 1 illustrates the dynamic effects, obtained using equation 1, of state adoption of the

wrongful discharge doctrines on union membership and coverage outcomes from two years prior

to three years post in adopting states relative to nonadopting states, and normalized to zero for

the year of adoption. These regressions, in contrast to the broader investigation of an extended

list of outcome variables used in the full fixed-effects analysis, are restricted to results for young

males with less than college education in order to highlight the dynamics surrounding policy

adoption, since union outcomes for this demographic group witnessed the most significant

impact of policy adoption.

18

Young less educated male union membership and coverage rates grew substantially to a peak

of 1.30% to 1.19%, for union membership rates and coverage rates respectively, over the two

years following the year of adoption of the implied contract doctrine. No clear trend emerges for

either the public policy or good faith doctrine adoptions. Young less educated male union

density appears to decline somewhat following good faith doctrine adoption, though there is no

steady pretreatment trend and significant year-to-year variation in union density. However, these

findings should be taken with some caution given the small size (6 treatment states) of the good

faith adoption sample over this period and the outsized role of large membership and coverage

declines, as well as the role played by possible confounding events, at least in one state,

Arizona7.

B. Fixed Effects Difference-in-Differences Regression

The fixed-effect full difference-in-difference regression specification from equation 2

reveals8 that, in the case of the implied contract doctrine, policy adoption is associated with

increases in aggregate private sector union membership and coverage rates of 0.84% and 1.03%,

as seen in column 3 of the first rows of Panels A and D of Table 2. The increase in union

coverage rate appears largely driven by a respective increase in the level of union membership

7 A closer investigation reveals large and disproportionate changes among the five sustained good faith doctrine

adoptions over the five years 1983 to 1988 surrounding good faith doctrine adoption in 1985 in Arizona, with a

nearly 30percent decline in union membership and coverage levels and a roughly 20percent increase in

nonmembership and noncoverage levels. Perhaps helping to explain this result, from 1983 to 1986 in Arizona the

copper industry experienced an historic strike of miners and mill workers that resulted in the replacement of nearly

all strikers and the decertification of their unions (Kingsolver 1989).

8 Sample statistics tables containing the mean, standard deviation, and number of observations for all fixed effects

regression samples of union outcome variables were constructed but omitted from this paper to save space. They are

available from author upon request. These tables are divided by subsamples for the implied contract, public policy,

and good faith doctrine combined treatment and control samples. The sample means for all outcome variables

across all three doctrine regressions approximate the national average levels of union membership and coverage and

union density over the years studied.

19

and coverage of 5.37% and 6.64%, visible in column 6 of the first row of Panels B and E of

Table 2. While there is evidence of a marginally significant 11.07% decline in private sector

union coverage levels associated with adoption of the good faith doctrine, in column 6 of the last

row of Panel E of Table 2, this does not exhibit any significant impact on union density, and is

not reflected in a statistically significant decline in union membership in column 6 of the last row

of Panel B of Table 2. As noted previously, these large declines could be at least in part related

to the disproportionately large declines in union membership and coverage levels in Arizona,

though the majority of the six good faith doctrine treatment states exhibit union decline well

above the national average in the years surrounding policy adoption.9

Turning to state industry groups in Table 3, analysis reveals that the implied contract doctrine

is associated with modest but statistically significant increases in manufacturing union

membership and coverage rates of 1.39% and 1.59 %, visible in columns 1 and 6 of the first row

of Panels A and B. These findings appear largely driven by a highly significant increase in the

level of manufacturing union membership and coverage of 8.06% and 9.15%, shown in columns

1 and 6 of the first row of Panels C and D. In columns 2 and 7 of the first row of Panels A and

B, one sees that adoption of the implied contract doctrine is associated with 0.38% and 0.68%

increases in nonmanufacturing union membership and coverage rates. Among the three major

9 If this high profile instance of union membership and coverage level decline and nonmembership and noncoverage

level growth does not in itself account for the disproportionate impacts for Arizona as a whole, the resolution of

such a large and public labor dispute could nonetheless set the tone of union membership development throughout

the Arizona labor market over these years. Delaware, New Hampshire, and Wyoming also exhibited union

membership and coverage level declines in the range of 10percent to 20percent per year over the five years

surrounding good faith doctrine adoption in each state, well above the national average for those years. On the other

hand, Idaho and Nevada, the remaining two good faith adopting states within the 1985 to 2011 sample window,

exhibit union decline at about the national average and slightly below the national average, respectively, over the

five years surrounding good faith doctrine adoption. These above average declines in other states might be

explained by an uptick in plant closings in Delaware and New Hampshire, and employment fluctuations connected

to swings in mineral export prices and long-run mechanization in extractive industries in Wyoming. Nevada’s

slower union decline could be explained by the growth of a highly organized service sector in Las Vegas over this

period.

20

industrial subdivisions of nonmanufacturing listed in Table 3 (i.e. construction, transportation-

communication-utilities, and service industries), construction industry union density increases

seem to drive the overall increase in nonmanufacturing union density, with a 2.76% and 2.3%

increase in union membership and coverage rates, respectively, visible in columns 3 and 9 of the

first row of Panels A and B. This result itself, unlike the situation for manufacturing industries,

appears driven not by statistically significant increases in the level of union membership and

coverage, but by 9.73% and 9.3% declines in the level of nonmember and noncovered

employment in columns 3 and 9 of the first row of Panels E and F of Table 3. While seeming to

show evidence of standard disemployment effects as demonstrated by Autor et. al. (2006),

among others, this result is not obvious since there is no reason one would expect nonmember

and noncovered employment within the construction industry to be composed of a higher

percentage of workers with less attachment to the labor market than, say, the service industry,

which does not exhibit comparable declines.

The impact of the good faith doctrine adoption may be reflected in the significant 12.6% and

9.83% respective declines in manufacturing union membership and coverage levels, shown in

columns 1 and 6 of the last row of Panel C and D of Table 3. These large declines could at least

in part be related to the outsized role of high profile union membership declines in Arizona,

Wyoming, Delaware, and New Hampshire, attributable to the small sample size and events

confounded with the year of policy adoption.

Analysis of the eight sex-age-educational attainment demographic groups in Tables 4 and 5

suggests that the implied contract doctrine is associated with a roughly 1.91% and 1.97% highly

significant increases in union membership and coverage rates for young male workers with less-

than-college education, as shown in the column 1 of the first row of Panel A in both tables. This

21

growth appears driven by a large and statistically significant roughly 11.34 and 11.38% increase

in the level of union membership and coverage for both young less educated males, as seen in

column 1 of the first row of Panel B in Tables 4 and 5. In the case of young females with less

than college education, union membership and coverage rates also increase by 0.74% and 1.14%

following adoption of the implied contract doctrine, as visible in column 5 of the first row of

both tables. However, only in the case of union coverage is it clear that this increase in density is

driven by a marginally significant 10.19% increase. This discrepancy between young less

educated female union membership and coverage level growth could suggest that, for certain

demographic groups, employment protection policies might increase free-riding behavior in

terms of increased utilization of union coverage without increased participation in union

membership participation. One could imagine that groups which have the most time

commitments outside of work, such as women who often bear the bulk of parenting

responsibilities, might be more inclined to free-ride, out of sheer necessity, on the benefits of

active union membership without participating themselves.

The implied contract doctrine also appears associated with a sizeable and marginally

significant 11.16 %and 10.13% decline in college educated older female membership and

coverage levels, visible in column 8 of the first row of Panel B in Tables 4 and 5, and a 8.82%

and 9.30% decline in college educated older female nonmember and noncovered employment

levels. However, this result is somewhat perplexing given that no significant or similarly-sized

disemployment effect emerges for less-educated younger and older females, whether union

members and covered or nonmembers and noncovered, who one might expect to have even

lower attachment to the labor force, and because the decline is greater for union members and

covered workers than for nonmember and noncovered workers, who one might expect to be

22

more precarious. Finally, the good faith doctrine adoption appears associated with a reallocation

of employment away from younger males and females with less than college education, whether

unionized or nonunionized, and toward older male college education nonmember and

noncovered employment. As noted previously, these declines could be confounded by

employment trends in the small sample of largely nonrepresentative states.10

VI. Conclusion

The implied contract doctrine does appear to increase unionization, and this growth seems to

occur in those (i.e. private), industries (i.e. manufacturing), and demographic groups (i.e. young

men and women with less-than-college education) where unions have both a strong base, as well

as potential to grow. Younger workers, especially younger women, without college education

have been underrepresented compared to older and more educated workers. While having the

most to gain from union membership, they are often too precarious to take the risks necessary to

form or join unions, or to secure a first collective bargaining agreement. Policies which lower

the costs to seeking union membership and union coverage could nudge these workers into

action, boosting union membership and coverage outcomes for themselves and the industries and

sectors in which they reside.

Past economic research has linked employment protection to increased labor productivity and

worker morale, as well as to slowed employment and job flows, small but consistent declines in

employment, and increased use of temporary workers. These new findings demonstrate, at least

10

One can imagine that such steep union decline in the highly organized manufacturing industries in Delaware and

New Hampshire, in which younger male and female workers, union and nonunion alike, predominate, along with

disorganization of the heavily unionized copper sector in Arizona, could spur erosion of union membership and

coverage across demographic groups, while paving the way for expansion of employment of older college educated

male nonmember and noncovered workers, a demographic who would be likely to staff new positions within

nonunion mineral extraction and processing industries.

23

in the context of the U.S., that wrongful discharge doctrines may also promote a main vehicle

workers have for voice and power at their jobs, union representation and collective bargaining.

At the very least, one could argue that these policies, passed during a period of steep union

decline and deindustrialization in the United States, helped stem losses in union membership and

coverage. Given current debates regarding the persistent and growing inequality following the

Great Recession, employment protection and unions could be a mutually reinforcing set of tools

to foster growth and fairness in labor markets and the economy at large.

24

Works Cited

Autor, David H., John J. Donauhue III, and Stewart J. Schwab. “The Costs of Wrongful-

Discharge Laws.” The Review of Economics and Statistics. Vol. 88, No. 2 (2006): 211-

231.

Autor, David H. “Outsourcing At Will: The Contribution of Unjust Dismissal Doctrine to the

Growth of Employment Outsourcing.” Journal of Labor Economics. vol. 21, no. 1,

2003.

Blanchard, Olivier Jean, and Lawrence F. Katz. "What We Know and Do Not Know About the

Natural Rate of Unemployment." Journal of Economic Perspectives. 11 (1997), 51-72.

Bronfenbrenner, Kate. “No Holds Barred: The Intensification of Employer Opposition to

Organizing.” Economic Policy Institute. 235. (May 2009): 1-31.

Dertouzos, James N., and Lynn A. Karoly. Labor-Market Responses to Employer Liability.

Santa Monica, CA: Rand, 1992.

Dertouzos, James N., and Lynn A. Karoly. "Employment Effects of Worker Protection: Evidence

from the United States." in Christoph F. Buechtemann (Ed.), Employment Security and

Labor Market Behavior: Interdisciplinary Approaches and International Evidence.

Ithaca, NY: ILR Press, 1993.

Hirsch, Barry T. and David A. Macphearson. "Union Membership and Coverage Database from

25

the Current Population Survey: Note." Industrial and Labor Relations Review, Vol. 56,

No. 2, January 2003, pp. 349-54.

Hirsch, Jeffrey M., Secunda, Paul M., and Richard A. Bales. Understanding Employment Law.

2nd

edition., Matthew Bender & Company., 2013. Print.

Kingsolver, Barbara. Holding the Line: Women in the Great Arizona Mine Strike of 1983.

Ithaca, NY: ILR Press, 1989.

Kugler, Adriana, and Gilles Saint-Paul. "How Do Firing Costs Affect Worker Flows in a World

with Adverse Selection?" Journal of Labor Economics. 22:3 (2004), 553-584.

LaLonde, Robert J. and Bernard D. Meltzer. “Hard Times for Unions: Another Look the

Significance of Employer Illegalities.” University of Chicago Law Review. 58. (1991):

953-1014. Print.

Lazear, Edward P. "Job Security Provisions and Employment." Quarterly Journal of Economics

105:3 (1990), 699-726.

Miles, Thomas J. "Common Law Exceptions to Employment at Will and U.S. Labor Markets."

Journal of Law, Economics and Organizations. 16:1 (2000), 74-101.

Morris, Andrew P. "Developing a Framework for Empirical Research on the Common Law:

General Principles and Case Studies of the Decline of Employment-at-Will." Case

Western Reserve Law Review. 45 (1995), 999-1148.

26

Muhl, Charles J. 2001 “The Employment-At-Will Doctrine: Three Major Exceptions.”

Washington, DC: Monthly Labor Review.

http://www.bls.gov/opub/mlr/2001/01/art1full.pdf.

Schmitt, John and Ben Zipperer. “Dropping the Ax: Illegal Firings During Union Election

Campaigns, 1951-2007.” Washington, DC: Center for Economic and Policy Research,

2009.

Weiler, Paul. “Promises to Keep: Securing Workers’ Rights to Self-Organization Under the

NLRA.” Harvard Law Review. 96.8 (June 1983): 1769-1827. Print.

29

Figure 1. --- Union Membership and Coverage Rates for Male Workers Aged 16 to 39 With Less

Than College Education Before and After Adoption of Wrongful Discharge Laws: Yearly Leads

From 2 Years Before to 3 Years After Adoption

Panel A. --- Implied Contract

100 x (Members/Employment) 100 x (Covered/Employment)

Panel B. --- Public Policy


Panel C. --- Good Faith


Note: Each figure represents a separate weighted OLS baseline difference and differences regression with two years of leads and lags in which the dependent variable is 100 times the number of

male less than college educated employed wage and salary workers between 16 and 39 years of age who are union members or covered by collective bargaining agreements over all male less

than college educated aged 16 to 39 employed wage and salary workers in 50 U.S. states. The number of union members, covered workers, and total wage and salary employees were constructed

from Current Population Survey Outgoing Rotation Group monthly earnings files for the years 1983 to 2014. All models include state main effects and indicators for each year in the sample, as

well as interactions between four Census-region dummies and calendar year dummies. Models are weighted by state’s share of national population aged 16-64 in each year using CPS earnings

weights. Huber-White robust standard errors were used to allow for unrestricted error correlations across observations within states.

-3

-2

-1

0

1

2

3

-2 -1 0 1 2

Point Estimate

Robust 90 Percent Confidence Interval

Year Relative to Adoption -3

-2

-1

0

1

2

3

-2 -1 0 1 2

Point Estimate


Year Relative to Adoption

-3

-2

-1

0

1

2

3

-2 -1 0 1 2

Point Estimate


Year Relative to Adoption -3

-2

-1

0

1

2

3

-2 -1 0 1 2

Point Estimate



-3

-2

-1

0

1

2

3

-2 -1 0 1 2

Point Estimate



-6

-4

-2

0

2

4

6

-2 -1 0 1 2

Point Estimate



30

TABLE 1. --- Recognition of Wrongful Discharge Laws by State as of December 31, 2014 and

Closest Calendar Year to Date of Policy Adoption and Repeal

State Implied Contact Y/N Adopt Repeal

Public Policy Y/N Adopt Repeal

Good Faith Y/N Adopt Repeal Readopt

Alabama………………….

Alaska………………….....

Arizona…………………..

Arkansas………………….

California…………………

Colorado…………………

Connecticut……………....

Delaware………………….

Florida……………………

Georgia…………………..

Hawaii…………………...

Idaho……………………..

Illinois……………………

Indiana…………………..

Iowa………………………

Kansas……………………

Kentucky…………………

Louisiana…………………

Maine…………………….

Maryland…………………

Massachusetts…………….

Michigan………………….

Minnesota…………………

Mississippi………………..

Missouri………………….

Montana…………………..

Nebraska………………….

Nevada……………………

New Hampshire…………..

New Jersey……………….

New Mexico………………

New York…………………

North Carolina……………

North Dakota……………..

Ohio………………………

Oklahoma…………………

Oregon……………………

Pennsylvania……………..

Rhode Island………………

South Carolina……………

South Dakota……………..

Tennessee…………………

Texas……………………..

Utah……………………….

Vermont…………………..

Virginia……………………

Washington……………….

West Virginia……………...

Wisconsin…………………

Wyoming………………….

Total

Y 1988

Y 1983

Y 1983

Y 1984

Y 1972

Y 1984

Y 1986

N

N

N

Y 1987

Y 1977

Y 1975

Y 1988

Y 1988

Y 1985

Y 1984

N

Y 1978

Y 1985

Y 1988

Y 1980

Y 1983

Y 1992

N 1983 1988

Y 1987

Y 1984

Y 1984

Y 1989

Y 1985

Y 1980

Y 1983

N

Y 1984

Y 1982

Y 1977

Y 1978

N

N

Y 1987

Y 1983

Y 1982

Y 1985

Y 1986

Y 1986

Y 1984

Y 1978

Y 1986

Y 1985

Y 1986

42 43 1

N

Y 1986

Y 1985

Y 1980

Y 1960

Y 1986

Y 1980

Y 1992

N

N

Y 1983

Y 1977

Y 1979

Y 1973

Y 1986

Y 1981

Y 1984

N

N

Y 1982

Y 1980

Y 1976

Y 1987

Y 1988

Y 1986

Y 1980

Y 1988

Y 1984

Y 1974

Y 1981

Y 1984

N

Y 1985

Y 1988

Y 1990

Y 1989

Y 1975

Y 1974

N

Y 1986

Y 1989

Y 1985

Y 1984

Y 1989

Y 1987

Y 1985

Y 1985

Y 1979

Y 1980

Y 1990

43 43 0

N

Y 1983

Y 1985

N

Y 1981

N

Y 1980

Y 1992

N

N

N

Y 1990

N

N

N

N

N

N

N

N

Y 1978

N

N

N

N

Y 1982

N

Y 1987

Y 1974 1981 1990

N

N

N

N

N

N

N 1985 1989

N

N

N

N

N

N

N

N

N

N

N

N

N

Y 1994

11 12 2 1 Notes: Based on author’s analysis of case law and secondary literature. “Y” denotes that a policy is recognized and “N” denotes that it is not recognized by a state as of December 31, 2014.

Adoptions and repeal dates are rounded to the calendar year in which they occurred if the first full month following policy change was from January to May of that year, and are rounded forward

to the following calendar year if the first full month following policy change is from June to December of the original year.

31

TABLE 2. --- Simplified Illustration of the Legal Policy Variables for the Full Fixed Effects

Differences in Differences Regression Using the Adoption of the Implied Contract Doctrine in

Massachusetts in 1988 for Treatment Observations and Rhode Island and Maine for Control

Observations

Panel A. --- Massachusetts (Adopts the Implied Contract Doctrine in 1988)

Legal Policy

Variables 1985 1986 1987 1988 1989 1990 1991 1992

Treatst 0 1 1 1 1 1 0 0

Postst 0 0 0 1 1 1 1 1

Treatst Postst 0 0 0 1 1 1 0 0

Postst Postst 0 0 0 0 0 0 1 1

Panel B. --- Rhode Island (Never Adopts A Wrongful Discharge Doctrine)

Legal Policy

Variables 1985 1986 1987 1988 1989 1990 1991 1992

Treatst 0 0 0 0 0 0 0 0

Postst 0 0 0 1 1 1 1 1


Postst Postst 0 0 0 0 0 0 0 0

Panel C. --- Maine (Only Adopted the Implied Contract Doctrine in 1978)

Legal Policy

Variables 1985 1986 1987 1988 1989 1990 1991 1992

Treatst 0 0 0 0 0 0 0 0

Postst 0 0 0 1 1 1 1 1


Postst Postst 1 1 1 1 1 1 1 1 Note: Constructed from author’s analysis of Autor et. al. (2003). Treatst is an indicator for the period 2 years before to 2 years after the year of adoption of a wrongful-discharge law in

treatment state s. Postst is an indicator for the period from the year of adoption11 to 2 years after the year of adoption for all states, both adopting and nonadopting. TreatstPostst is an

interaction term representing the three year period following policy adoption, in the treatment state only, beginning with the first year of adoption. The coefficient of interest β2 is an estimate of

the pre-post change in the outcome variable in adopting states relative to the corresponding change in non-adopting states. Postpostst is a dummy variable that turns on for each state that

reenters the control group 3 years after the year of policy adoption to account for any bias caused by the enduring effect of the policy adoption after the five-year treatment window expires.

32

TABLE 2. --- Difference-In-Differences Estimates of the Impact of Wrongful Discharge Laws on

State Union Membership and Coverage Outcomes by Sector: Contrasting Outcomes in Years 2

and 3 Following Adoption with Years 1 and 2 Preceding Adoption: 1983-2014

A. Membership Rates B. Membership Levels C. Nonmembership Levels

All Public Private All Public Private All Public Private Doctrines (1) (2) (3) (4) (5) (6) (7) (8) (9)

Implied 0.67* -0.38 0.84** 3.22 -2.20 5.37+ -1.16 1.32 -1.59

Contract

R-Sq

(0.29)

0.99

(0.82)

0.99

(0.25)

0.99

(2.65)

1.00

(3.17)

0.99

(2.93)

1.00

(1.73)

1.00

(2.44)

0.99

(1.81)

1.00

N 455 455 455

Public 0.03 -0.88 0.12 -2.13 -3.66 -1.19 -1.41 0.17 -1.55

Policy

R-Sq

(0.27)

0.99

(0.89)

0.99

(0.27)

0.99

(3.39)

1.00

(5.35)

0.99

(3.68)

1.00

(1.88)

1.00

(2.15)

0.99

(1.95)

1.00

N 465 465 465

Good 0.10 -0.08 -0.11 -3.44 6.98 -9.90 -2.11 11.44** -4.51

Faith

R-Sq

(0.67)

0.99

(1.83)

0.99

(0.41)

0.98

(7.53)

1.00

(11.39)

0.99

(5.94)

1.00

(3.38)

1.00

(4.23)

0.99

(4.27)

1.00

N 493 493 493

D. Coverage Rates E. Coverage Levels F. Noncoverage Levels

Implied 0.95** 0.21 1.03** 4.94+ 1.18 6.64* -1.51 0.48 -1.83

Contract

R-Sq

(0.34)

0.99

(0.79)

0.99

(0.31)

0.98

(2.92)

1.00

(3.12)

0.99

(3.13)

1.00

(1.70)

1.00

(2.16)

0.99

(1.82)

1.00

N 455 455 455

Public 0.23 -0.37 0.24 -0.84 -1.71 -0.15 -1.66 -0.55 -1.69

Policy

R-Sq

(0.32)

0.99

(0.94)

0.98

(0.32)

0.98

(3.20)

1.00

(4.76)

0.99

(3.71)

1.00

(1.90)

1.00

(2.28)

0.99

(1.94)

1.00

N 465 465 465

Good 0.09 -0.39 -0.26 -3.50 7.38 -11.07+ -2.14 11.10* -4.35

Faith

R-Sq

(0.62)

0.99

(1.16)

0.98

(0.45)

0.98

(6.14)

1.00

(7.85)

0.99

(5.91)

1.00

(3.35)

1.00

(4.65)

0.99

(4.28)

1.00

N 493 493 493 Note: Each figure represents a separate weighted OLS fixed-effects difference and differences regression in which the dependent variable is 100 times the number of employed wage and salary

workers between 16 and 64 years of age in a givens sector who are union members or covered by collective bargaining agreements over all employed wage and salary workers between the ages

of 16 and 64 in a given sector, 100 times the natural log of the number of employed wage and salary workers between 16 and 64 years of age in a given sector who are union members or covered

by collective bargaining agreements over all employed wage and salary workers between the ages of 16 and 64 in a given sector, or 100 times the natural log of the number of employed wage

and salary workers between the ages of 16 and 64 in a given sector who are not union members nor covered by collective bargaining agreements over all employed wage and salary workers

between the ages of 16 and 64 in a given sector in 50 U.S. states. The number of union members, covered workers, nonmembers, noncovered workers, and total wage and salary workers in a

given sector is constructed from Current Population Survey Outgoing Rotation Group monthly earnings files for the years 1983 to 2014. All models include state main effects and indicators for

each year in the sample, as well as interactions between four Census-region dummies and calendar year dummies. Models are weighted by state’s share of national population aged 16-64 in each

year using CPS earnings weights. Huber-White robust standard errors were used to allow for unrestricted error correlations across observations within states.

33

TABLE 3. --- Difference-In-Differences Estimates of the Impact of Wrongful Discharge Laws on State

Union Membership and Coverage Outcomes by Industry: Contrasting Outcomes in Years 2 and 3

Following Adoption with Years 1 and 2 Preceding Adoption: 1983-2014

Panel A.--- Membership Rates Panel B. --- Coverage Rates

Man. Nonman. Const. Trans. Serv. Man. Nonman. Const. Trans. Serv. Doctrines (1) (2) (3) (4) (5) (6) (7) (8) (9) (10)

Implied 1.39* 0.38+ 2.76* -0.22 0.15 1.59* 0.68* 2.30+ 0.65 0.50

Contract

R-sq

(0.59)

0.97

(0.22)

0.99

(1.12)

0.96

(1.02)

0.92

(0.45)

0.98

(0.63)

0.97

(0.28)

0.99

(1.20)

0.96

(1.13)

0.92

(0.39)

0.98

N 455 455

Public -0.34 0.04 0.88 0.15 -0.12 -0.21 0.24 0.92 -0.08 0.08

Policy

R-Sq

(0.65)

0.97

(0.26)

0.99

(1.09)

0.96

(0.85)

0.92

(0.49)

0.98

(0.81)

0.97

(0.32)

0.99

(1.13)

0.96

(0.80)

0.92

(0.54)

0.98

N 465 465

Good -1.42 0.36 0.61 1.31 -0.38 -1.67+ 0.36 0.73 -0.21 -0.24

Faith

R-sq

(0.85)

0.97

(0.76)

0.99

(1.60)

0.96

(2.51)

0.92

(0.55)

0.98

(0.96)

0.96

(0.74)

0.99

(1.67)

0.96

(3.17)

0.91

(0.70)

0.98

N 493 493

Panel C. --- Membership Levels Panel D. --- Coverage Levels

Implied 8.06* 1.42 2.70 -0.42 --3.98 9.15* 3.80 0.31 1.11 -0.31

Contract

R-sq

(3.42)

0.99

(2.43)

1.00

(8.75)

0.98

(3.81)

0.99

(4.39)

0.99

(3.41)

0.99

(2.77)

1.00

(8.73)

0.98

(3.91)

0.99

(3.22)

0.99

N 455 455

Public -1.64 -1.46 7.95 -5.06 -4.75 -0.45 -0.11 5.46 -5.66 -2.98

Policy

R-sq

(4.34)

0.99

(3.85)

1.00

(10.07)

0.98

(4.86)

0.99

(6.15)

0.99

(4.48)

0.99

(3.56)

1.00

(9.85)

0.98

(4.47)

0.99

(4.88)

0.99

N 465 465

Good -12.6** 0.58 13.98 6.40 -10.73 -9.83* -0.46 10.82 2.43 -6.08

Faith

R-sq

(4.49)

0.99

(8.96)

1.00

(13.63)

0.97

(14.80)

0.99

(7.71)

0.99

(3.95)

0.99

(7.77)

1.00

(13.52)

0.97

(15.81)

0.99

(6.27)

0.99

N 493 493

Panel E. --- Nonmembership Levels Panel F. --- Noncoverage Levels

Implied -0.52 -1.35 -9.73* -2.00 -1.69 -0.86 -1.72 -9.30* -3.26 -2.15

Contract

R-sq

(2.28)

1.00

(1.69)

1.00

(4.58)

0.99

(2.55)

0.99

(1.69)

1.00

(2.34)

1.00

(1.65)

1.00

(4.60)

0.99

(2.80)

0.99

(1.54)

1.00

N 455 455

Public -0.57 -1.59 -6.16 -4.23 -2.21 -0.82 -1.83 -6.21 -3.92 -2.45

Policy

R-sq

(2.70)

1.00

(2.03)

1.00

(4.60)

0.98

(2.66)

0.99

(2.13)

1.00

(2.70)

1.00

(2.07)

1.00

(4.46)

0.99

(2.79)

0.99

(2.13)

1.00

N 465 465

Good 3.33 -2.68 -2.57 3.00 -1.00 3.66 -2.78 -2.96 5.03 -1.38

Faith

R-sq

(4.80)

1.00

(3.34)

1.00

(7.33)

0.98

(4.88)

0.99

(3.52)

1.00

(4.70)

1.00

(3.33)

1.00

(7.20)

0.98

(4.08)

0.99

(3.76)

1.00

N 493 493 Note: Each figure represents a separate weighted OLS fixed-effects difference and differences regression defined identically to those in Table 2 with the exception that investigation is conducted for selected

industries.

34


Union Membership Outcomes by Sex, Age, and Education: Contrasting Outcomes in Years 2 and 3


Panel A.--- Membership Rates

M.H.Y. M.H.O. M.C.Y. M.C.O. F.H.Y. F.H.O. F.C.Y. F.C.O.

Doctrines (1) (2) (3) (4) (5) (6) (7) (8) Implied 1.91** 0.84 0.83 -0.46 0.74+ 0.18 0.22 0.19 Contract R-sq

(0.71) 0.97

(0.98) 0.97

(0.54) 0.94

(0.87) 0.94

(0.39) 0.93

(0.71) 0.96

(0.33) 0.94

(0.88) 0.93

N 455

Public 0.20 -0.14 0.58 0.04 0.47 -0.58 -0.67+ -0.63 Policy R-sq

(0.54) 0.96

(0.90) 0.98

(0.63) 0.94

(0.68) 0.94

(0.43) 0.93

(0.63) 0.96

(0.38) 0.93

(0.98) 0.93

N 465

Good 0.14 4.08 0.06 -1.60 0.11 -2.03* 0.18 0.17 Faith R-sq

(0.96) 0.96

(3.80) 0.97

(0.69) 0.94

(1.18) 0.93

(0.44) 0.92

(0.77) 0.95

(0.56) 0.93

(1.56) 0.92

N 493

Panel B. --- Membership Levels

Implied 11.34* 2.22 4.79 -4.93 4.55 -5.21 -2.85 -11.16+ Contract R-sq

(4.33) 0.99

(4.76) 0.99

(5.25) 0.99

(6.95) 0.99

(6.24) 0.97

(7.13) 0.98

(5.20) 0.98

(6.51) 0.98

N 455

Public -3.08 -1.36 -0.60 -2.81 -0.90 -5.10 -10.15+ -6.20 Policy R-sq

(4.76) 0.98

(4.44) 0.99

(5.90) 0.99

(4.48) 0.99

(6.00) 0.97

(6.15) 0.98

(6.03) 0.98

(8.89) 0.98

N 465

Good -18.13 21.97 -2.44 0.43 -19.24* -20.65+ -9.07 -3.47 Faith R-sq

(15.29) 0.98

(18.67) 0.99

(9.38) 0.99

(5.92) 0.99

(7.19) 0.97

(11.65) 0.98

(9.87) 0.98

(6.62) 0.98

N 493

Panel C. --- Nonmembership Levels

Implied -1.25 -2.50 -1.39 0.43 -0.73 -0.08 -2.00 -8.82** Contract R-sq

(2.72) 1.00

(2.28) 0.99

(2.55) 1.00

(2.75) 1.00

(2.29) 1.00

(2.30) 0.99

(2.85) 1.00

(3.02) 0.99

N 455

Public -2.00 -0.94 -2.89 -0.16 -1.60 0.41 -0.65 -2.08 Policy R-sq

(2.44) 1.00

(2.52) 0.99

(2.60) 1.00

(2.70) 1.00

(2.32) 1.00

(3.20) 1.00

(2.66) 1.00

(2.88) 0.99

N 465

Good -8.12* -4.81 -1.28 8.97* -7.23** 5.53 -1.60 0.05 Faith R-sq N

(3.49) 1.00

(6.95) 0.99

(7.46) 1.00

(3.41) 0.99 493

(2.49) 1.00

(5.50) 0.99

(5.17) 1.00

(6.34) 0.99

Note: Each figure represents a separate weighted OLS fixed-effects difference and differences regression on union membership outcomes defined identically to those in Table 2, though

investigated for given each of eight age-sex-education demographic category. MHY=males with less than college education between the ages of 16 and 39, MHO=males with less than college

education between the ages of 40 and 64, MCY=males with college or more education between the ages of 16 and 39, MCO=males with more than college or more education between the ages of

40 and 64, FHY=females with less than college education between the ages of 16 and 39, FHO=females with less than college education between the ages of 40 and 64, FCY=females with

college or more education between the ages of 16 and 39, FCO=females with college or more education between the ages of 40 and 64.

35


Union Coverage Outcomes by Sex, Age, and Education: Contrasting Outcomes in Years 2 and 3


Panel A.--- Coverage Rates

M.H.Y. M.H.O. M.C.Y. M.C.O. F.H.Y. F.H.O. F.C.Y. F.C.O.

Doctrines (1) (2) (3) (4) (5) (6) (7) (8) Implied 1.97* 1.04 1.13+ -0.02 1.14* 0.53 0.61 0.60 Contract R-sq

(0.75) 0.96

(0.97) 0.97

(0.57) 0.94

(0.91) 0.93

(0.45) 0.91

(0.78) 0.95

(0.40) 0.93

(0.98) 0.91

N 455

Public 0.50 0.19 0.41 0.08 0.99* -0.06 -0.58 -0.56 Policy R-sq

(0.62) 0.96

(0.89) 0.97

(0.69) 0.94

(0.81) 0.93

(0.43) 0.91

(0.65) 0.95

(0.49) 0.92

(1.04) 0.91

N 465

Good 0.17 3.60 0.10 -1.80 -0.13 -1.60* 0.24 0.07 Faith R-sq

(0.96) 0.96

(2.89) 0.97

(0.53) 0.94

(1.52) 0.93

(0.51) 0.91

(0.72) 0.94

(0.66) 0.92

(2.05) 0.91

N 493

Panel B. --- Coverage Levels

Implied 11.38* 2.98 6.42 -3.15 10.19+ 1.15 0.55 -10.13+ Contract R-sq

(4.51) 0.99

(4.67) 0.99

(5.07) 0.99

(5.96) 0.99

(5.88) 0.97

(6.04) 0.98

(4.33) 0.98

(6.03) 0.98

N 455

Public -1.83 -0.48 -1.80 -3.10 7.07 -2.39 -7.11 -5.53 Policy R-sq

(4.54) 0.98

(4.47) 0.99

(4.98) 0.99

(4.85) 0.99

(4.91) 0.97

(5.90) 0.98

(4.53) 0.98

(7.39) 0.98

N 465

Good -16.98 15.61 -3.68 0.73 -20.04 -18.59* -5.70 -3.44 Faith R-sq

(15.18) 0.98

(12.04) 0.99

(8.95) 0.99

(6.40) 0.99

(12.57) 0.97

(8.19) 0.98

(8.19) 0.98

(5.82) 0.98

N 493

Panel C. --- Noncoverage Levels

Implied -1.34 -2.78 -1.86 0.17 -1.24 -0.49 -2.48 -9.30** Contract R-sq

(2.71) 1.00

(2.37) 0.99

(2.58) 1.00

(2.84) 1.00

(2.16) 1.00

(2.31) 0.99

(2.81) 1.00

(2.98) 0.99

N 455

Public -2.44 -1.34 -2.69 -0.23 -2.33 -0.16 -0.64 -2.16 Policy R-sq

(2.48) 1.00

(2.45) 0.99

(2.84) 1.00

(2.78) 1.00

(2.27) 1.00

(3.23) 0.99

(2.79) 1.00

(2.83) 0.99

N Good

-8.40*

-4.00

-1.33

465 9.13*

-7.27**

5.25

-1.69

-0.11

Faith R-sq N

(3.57) 1.00

(5.83) 0.99

(7.32) 1.00

(3.63) 0.99 493

(2.29) 1.00

(5.81) 0.99

(5.38) 1.00

(6.86) 0.99

Note: Each figure represents a separate weighted OLS fixed-effects difference and differences regression on union coverag outcomes defined identically to those in Table 2, though investigated

for given each of eight age-sex-education demographic category. MHY=males with less than college education between the ages of 16 and 39, MHO=males with less than college education

between the ages of 40 and 64, MCY=males with college or more education between the ages of 16 and 39, MCO=males with more than college or more education between the ages of 40 and

64, FHY=females with less than college education between the ages of 16 and 39, FHO=females with less than college education between the ages of 40 and 64, FCY=females with college or

more education between the ages of 16 and 39, FCO=females with college or more education between the ages of 40 and 64.

the impact of adoption of wrongful discharge laws on union

Documents