A review of international experience

part of a profit-sharing plan and that the voluntary employee contributions would be 
treated as employer contributions for income tax purposes (thereby allowing the associ-
ated income tax to be deferred) providing that the following conditions were met:
 
• The applicable contributions were made by an irrevocable election before the end 
of the year in which the profits on which they were based were determined.
 
• Certain nondiscrimination measures were met (at the time, more than half the 
participants had to be in the lowest-paid two-thirds of all eligible employees).

56 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
 
• The applicable contributions were subject to the same restrictions on withdrawals 
and distributions as other employer contributions made to the plan.
Over time, a number of employers extended this bonus/profit-sharing structure to 
allow employees to elect to reduce their normal salary through such deferrals and expanded 
the salary reduction design to money purchase plans. Not surprisingly, many lower-paid 
workers chose to take these bonuses in cash, whereas higher-paid workers were inclined 
to defer the payment and the taxation of the payment (something that was particularly 
important in a period in which income tax rates were much more steeply progressive). 
Still, as recently as the early 1970s, fewer than a thousand CODAs were in existence.
Concerned that a disproportionate share of tax benefits from these arrangements was 
going to higher-paid workers, on December 6, 1972, the IRS issued a proposed regula-
tion that provided that contributions made at the election of an employee “in return for a 
reduction in his basic or regular compensation or in lieu of an increase in such compensa-
tion” would be taxed as if they had been received by the employee. Although the proposed 
regulation did not specifically reference CODAs involving bonuses or amounts paid in 
addition to regular compensation, it cast a shadow on their future treatment as well. At 
the same time, the U.S. Treasury Department, which has a broader policy role in the 
regulation of pensions than the IRS (which is a bureau of the U.S. Treasury), announced 
that it would further study these plan designs and determine how they should be treated. 
THEN CAME ERISA
Before any regulations were issued, in 1974 ERISA was enacted. This landmark federal 
law—the result of many years of debate and negotiations in Congress—established mini-
mum standards for pension plans sponsored by private sector employers as well as rules 
governing the federal income tax effects of transactions associated with employee ben-
efit plans. Included in the legislation was a provision specifically prohibiting the Treasury 
Department from issuing regulations that would affect the tax treatment of these arrange-
ments before January 1, 1977. This provision was intended to provide Congress with an 
opportunity to take action on the issue. This deadline was extended twice, once to January 
1, 1978 (by the Tax Reform Act of 1976) and a second time to January 1, 1980 (by the 
Tax Treatment Extension Act of 1978). In July 1978, the IRS withdrew its 1972 proposed 
regulation.
In the Revenue Act of 1978, Congress finally acted by adding Section 401(k) to the 
Internal Revenue Code (Raish n.d.). This new section of the tax code, which would even-
tually lend its name to the 401(k) plan, provided that a profit-sharing or stock bonus plan 
would not be barred from the favorable tax treatment afforded to qualified pension plans 
merely because it allowed employees to make a voluntary pretax contribution (as allowed 
under a qualified CODA). It also set forth the criteria for those arrangements.
Section 408(a)(8), added to the tax code in the same legislation, explicitly provided 
that contributions made by employees to a qualified trust at an employee’s election under 
a qualified CODA were to be treated as if they were employer contributions and there-
fore not subject to federal income taxation at the point of contribution. On November 
10, 1981, the IRS published proposed regulations under the new 401(k) that provided 
guidance about how employers could implement these arrangements—an event many 

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
57
practitioners now consider the birth of the 401(k) plan. Within two years, surveys showed 
that nearly half of all large firms were either offering, or considering offering, a 401(k) 
plan.
PARTICIPATION BOOSTERS
The Tax Reform Act of 1984 established two mandatory nondiscrimination tests for 
CODAs, which were applicable to 401(k) plans. Rather than merely complying with the 
general nondiscrimination tests, these 401(k) programs now also had to pass the muster of 
a new nondiscrimination test called the “average deferral percentage” test. Passing this test 
required that the average rate of pretax contributions deferred by non–highly compen-
sated employees (NHCEs; at that time, essentially the lower two-thirds of the workforce 
eligible to participate in the plan) fall within a narrowly defined range of the average per-
centage of the deferrals made by the group defined as highly compensated.
Contributions made in excess of those limits could either be returned to highly 
compensated employees (generally employees who deferred at the highest rates, not always 
the most highly compensated) and taxed or recharacterized as after-tax contributions to 
the plan (if the plan permitted such contributions). Alternatively, certain additional con-
tributions (called qualified matching contributions or qualified nonelective contributions) 
could be made by the employer to effectively increase the deferral levels of NHCEs. These 
qualified contributions must be immediately 100 percent vested; they can therefore be an 
expensive solution to bringing the plan into compliance. Additionally, the first method 
cited above—returning contributions—was viewed as extremely disruptive in employee 
relations, particularly with a key constituency (the highly compensated). The available 
correction methods provided a strong incentive for plan sponsors to avoid falling short of 
the nondiscrimination test standards, reinforcing the emphasis on encouraging participa-
tion by NHCEs.
Despite the imposition of these restrictions, the continued popularity of these pro-
grams—and their tax-deferral design—had real implications for government revenue 
flows. In fact, to reduce fiscal pressures, in 1984 the Treasury Department proposed elimi-
nating Section 401(k) from the Internal Revenue Code.
The Tax Reform Act of 1986 brought further restrictive changes to the 401(k), nota-
bly a ceiling on allowable pretax contributions of $7,000 per year (adjusted annually based 
on the consumer price index) and changes to the nondiscrimination test.
4
 Changes to the 
nondiscrimination test included a revised definition of compensation as well as a revised, 
and generally narrower, definition of highly compensated employee. At the same time, 
another nondiscrimination test (the average compensation percentage test)—designed to 
achieve the same objective of limiting differences in matching contributions and after-
tax employee contributions—was introduced, putting additional pressure on the amounts 
that could be contributed by more highly compensated workers.
These new nondiscrimination tests provided a particularly strong incentive for 
employers to encourage lower-paid workers to contribute as much as possible to the plan. 
Indeed, targeting a high participation rate among NHCEs was necessary not only to max-
imize the deferral levels of the higher-paid group but to ensure legal compliance. And 
while the ability to contribute (and save) pretax provided incentives to the NHCEs, the 
employer matching contribution already in place in many thrift or savings plans (many of 

58 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
which were to add a 401(k) feature) also acted as a powerful financial contribution incen-
tive. With more stringent standards to meet, interest in boosting the saving behaviors of 
the NHCEs was higher than ever.
5
LATER CHANGES
The Small Business Job Protection Act of 1996 was the first law to move back toward 
encouraging the expansion of 401(k)s by providing design-based “safe harbor” methods 
for satisfying the nondiscrimination tests applicable to 401(k) plans.
6
 These methods basi-
cally allowed employers to avoid passing the standard 401(k) nondiscrimination test by 
either making a 3 percent contribution for all eligible workers or providing a basic match 
of at least 100 percent on the first 3 percent of pay deferred plus 50 percent on the next 
2 percent. The act also repealed the limits imposed under the Internal Revenue Code’s 
Section 415(e), which reduced the amount that could be contributed to defined contri-
bution plans (including 401(k) plans) if the employer also sponsored a defined benefit 
plan for the same employees and greatly simplified the definition of highly compensated 
employees.
In 1998, the IRS issued Revenue Ruling 98−30, which gave a stamp of approval for 
employers to make “negative elections” (that is, automatic enrollment) into 401(k) plans 
for newly eligible employees. In 2000, the IRS followed up with Revenue Ruling 2000−8, 
providing additional guidance on negative elections by allowing automatic enrollment 
in 401(k) plans for already eligible employees who were deferring at a rate less than the 
automatic enrollment rate.
The Economic Growth and Tax Relief Reconciliation Act of 2001 (EGTRRA) made 
significant changes to the 401(k) (Facts from EBRI 2005). These changes, subject to the 
nondiscrimination testing limits, were designed to increase individual savings in these 
programs. The law dramatically increased elective annual deferral limits (to $11,000 in 
2002, rising by $1,000 a year until 2006, when inflation indexing would take over), per-
mitted additional “catch-up” contributions by participants 50 and older (up to $1,000 
in 2002, $2,000 in 2003, $3,000 in 2004, and $4,000 in 2005), increased the maxi-
mum compensation limit (as a percentage of salary) to $200,000 from $170,000 (with 
amounts indexed thereafter), and increased the annual defined contribution dollar limit 
from $35,000 to $40,000 under Section 415(c) and allowed annual indexing thereafter in 
$1,000 increments. Additionally, the compensation limit in Section 415(c) was increased 
to 100 percent of compensation, from 25 percent at the time EGTRRA was passed, and 
the vesting requirements of employer matching contributions were accelerated beginning 
after 2001 (requiring that they be vested at least as rapidly as three-year cliff vesting or 
two- to six-year graded vesting).
7
EGTRRA also laid out the provisions for a Roth 401(k), to be effective in 2006, slated 
to sunset at the end of 2010 along with the rest of the EGTRRA provisions. Roth contri-
butions are deferred as after-tax dollars (that is, income tax is paid or withheld in the year 
contributed). Combined with regular 401(k) pretax deferrals in determining the maximum 
annual deferral, qualified distributions from a designated Roth 401(k) account, including all 
income, were to be tax free. (A traditional 401(k) account is funded with pretax dollars. In 
general, tax must be paid when the original contribution and earnings are withdrawn.) All 
employer matching funds are deposited into the account on a pretax basis, even if all of the 

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
59
employee’s contributions are Roth contributions. Employer contributions may be subject to 
vesting rules set by the plan documents requiring the employee to reach a certain number of 
years of service before he or she is entitled to keep the matching funds.
Actual implementations of this Roth provision at the plan level were modest, 
because of the late implementation relative to the other provisions of the law, the relatively 
brief window of applicability (2006 to the sunset in 2010), the cost of implementing 
the change, and the potential confusion among participants long accustomed to hearing 
about the benefits of deferring on a pretax basis.
However, in August 2006, the Pension Protection Act was signed into law, with a 
major emphasis on fostering retirement saving and 401(k) plan participation. The act 
provided a safe harbor for automatic enrollment that included a provision allowing for 
automatic annual increases in deferrals. It also resolved the potential conflict between 
the practice of automatic enrollment and some state laws prohibiting wage garnishment. 
Additionally, it made permanent certain provisions of EGTRRA that were set to expire 
or sunset at the end of 2010, including increased deduction limits, increased rollover 
options, and a start-up tax credit for small-employer plans.
At the same time, the Roth 401(k) provisions that became effective in 2006 gained 
new life with the removal of the 2010 sunset. After years of hesitation, by 2011, about 
30 percent of employer respondents to the PLANSPONSOR Defined Contribution Survey 
had adopted a Roth provision, though enrollment by participants remained weak.
The Pension Protection Act’s automatic enrollment safe harbor provided fresh vigor 
to the adoption of the design, even among plan sponsors that, either deliberately or inad-
vertently, failed to qualify for the full protections of the legislative safe harbor (Adams 
2011). A 2008 analysis by the Employment Benefit Research Institute (EBRI) of the pro-
vision indicates that, even under the most conservative assumptions for auto-escalation of 
contributions, switching 401(k) plans to automatic enrollment was seen as likely to have 
a very significant positive impact on generating additional retirement savings for many 
workers, especially low-income workers (EBRI 2008).
Overall, the emergence of 401(k) plans brought a new dynamic to defined contri-
bution plans in the United States, enabling workers of all income levels to defer payment 
of taxes on compensation contributed to these plans. Popular as these programs were to 
become, their stricter nondiscrimination rules made the participation of NHCEs a criti-
cal factor. Employer matches were an important source of encouragement, not only for 
participation but also for participation at an effective level.
The Level of the Match 
Cost was a key factor for employers in determining the level of deferral at which to match, 
but plan sponsors also wanted to avoid the necessity of having to refund “excess” con-
tribution deferrals to highly compensated workers—contributions exceeding the limits 
imposed by the average deferral percentage nondiscrimination tests. Many opted to match 
deferrals of 4‒10 percent of pay at various rates (commonly $0.25 or $0.50 on the dollar), 
with deferrals above those amounts left unmatched.
The new nondiscrimination requirements—and heightened interest in the rates of 
plan participation of NHCEs—also led to greater interest in employee education about 

60 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
these workplace programs, both to explain the program and to encourage participation. 
This heightened interest was particularly true in the period following the introduction of 
the new (and more restrictive) tests of the Tax Reform Act of 1986. Although nondiscrim-
ination testing constraints remain an active concern for many plan sponsors, they have 
become less of an issue because of subsequent modifications in the test structure, as well as 
the introduction of new safe harbor plan designs (with minimum employer contribution 
requirements, which preclude the need for a nondiscrimination test) and the expanding 
impact of automatic enrollment (which serves to boost participation levels and the defer-
ral averages for NHCEs).
Total contributions to 401(k) plans more than doubled between 1984 and 1987, 
from $16.3 billion to just over $33.0 billion. The percentage of contributions to defined 
contribution plans stemming from CODAs also rose, from 42 percent in 1984 to 55 per-
cent in 1987 (Andrews 1992).
Among the early adopters of the 401(k) design were smaller businesses that had 
either not previously sponsored a plan or in some cases had offered a profit-sharing pro-
gram. For businesses that had not sponsored a plan, the 401(k) did not require employer 
contributions (unlike profit-sharing plans, which, by definition, involved the sharing of 
employer profits). Small businesses were to grow even more enamored of the design after 
passage of the Tax Reform Act of 1986, which prohibited individuals above certain com-
pensation levels from making pretax contributions to an individual retirement account 
(IRA) if they were covered by a profit-sharing plan.
According to PLANSPONSOR’s 2011 Defined Contribution Survey of plan spon-
sors, only about 3 in 10 plan sponsors (fewer among larger programs) said that “all or 
nearly all” participants were deferring enough income to take full advantage of the maxi-
mum employer match (Adams 2011). Less than a quarter of the largest programs (401(k) 
assets in excess of $1 billion) reported that 90 percent or more of their participants were 
deferring at a level sufficient to receive the full employer match. Among large and mid-
size plans, the percentage was even lower (table 3.1). Among the 23 percent of respon-
dents who reported decreasing their workplace plan contribution percentage at some 
point, 46 percent said they needed extra money; 9 percent reported that the reduction was 
caused by elimination of the company match.
Although an employer match of 50 percent of employee deferrals up to 6 percent of 
pay is often presented as the norm, there is actually a great deal of variety in how defined 
contribution matches are structured. Seventy-seven percent of the defined contribu-
tion client base of Vanguard (a large mutual fund company) did, in fact, incorporate the 
50 percent of deferral up to 6 percent of pay as a structure, but it applied to only 60 per-
cent of its participant accounts. In contrast, 15 percent of the plans—and 35 percent of 
participants—operated under a dollar-for-dollar match on the first 3 percent of pay, with 
50 percent on the next 2 percent. PLANSPONSOR’s 2011 Defined Contribution Survey 
of nearly 7,000 plan sponsors also reveals an array of options (tables 3.2 and 3.3).
Impact of Matching Contributions on 401(k) Saving
The sensitivity of participation and contributions to plan characteristics—notably the 
employer matching rate—may play a critical role in retirement saving. It has long been 

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
61
TABLE 3.1  Percentage of active participants deferring enough salary to take full advantage of the 
maximum employer match, by company size
Percentage of companies
Participation
All firms
Micro
Small
Midsize
Large
Mega
All or nearly all participants (90% or more)
29.2
37.9
25.4
21.1
16.8
22.2
Vast majority (75% or more)
20.6
16.9
21.8
22.3
27.7
27.8
About half
25.6
18.2
28.5
33.1
32.0
38.9
Less than half
24.7
27.0
24.4
23.6
23.5
11.1
Average
65.9
66.7
64.9
64.2
65.0
71.9
Median
74.0
75.0
70.0
70.0
70.0
74.5
SOURCE: PLANSPONSOR’s 2011 Defi ned Contribution Survey.
NOTE: The PLANSPONSOR survey categorized plans as follows: “mega”: more than $1 billion; “large”: $200 million–$1 bil-
lion; midsize: $50 million–$200 million; small: $5 million–$50 million; and “micro”: less than $5 million in plan assets.
TABLE 3.2  Level of match offered by companies offering matching contributions, by company size
Percentage of companies
Level of match (as percentage of 6% of 
salary)
All firms
Micro
Small
Midsize
Large
Mega
More than 100 
4.2
4.2
4.7
2.7
3.7
5.0
100
6.2
4.3
4.8
8.0
9.9
17.8
51–99
27.3
24.4
25.1
28.9
37.0
42.6
50
30.7
33.7
29.8
28.4
30.3
19.3
Less than 50 
31.7
33.4
35.6
31.9
19.1
15.3
SEE table 3.1.
TABLE 3.3  Type of employer contributions, by company size
Percentage of companies
Type of contribution
All firms
Micro
Small
Midsize
Large
Mega
Match of participant contribution
66.4
56.4
69.7
78.7
80.4
83.3
Nonelective contribution
14.9
10.8
13.4
21.0
26.9
27.5
Profi t-sharing contribution
28.6
26.0
35.0
29.2
22.9
16.4
Other 12.4
12.7
12.2
10.5
12.8
14.5
No contribution
11.0
16.2
7.6
6.6
5.5
5.6
SEE table 3.1.

62 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
assumed that matching employer contributions—the allure of “free money” to partici-
pants (and would-be participants)—provided a strong financial motivation to contrib-
ute to defined contribution plans, notably 401(k)s. Industry surveys have suggested that 
employee contribution levels tend to cluster around the matching levels, reinforcing a 
cause-and-effect connection.
Historically, providing employer matching contributions to 401(k) plans was seen 
as a primary means of increasing the likelihood of passing the nondiscrimination (aver-
age deferral percentage) tests (Brady 2007). However, Ippolito (1997) finds an alterna-
tive explanation might be more plausible: in essence, employers use the 401(k) match 
to attract and retain a workforce with specific characteristics, and matches reward work-
ers with lower discount rates. Mitchell, Utkus, and Yang (2005) posit that employee 
demand could offer an alternative explanation, with the result that highly compensated 
employees demand more generous tax-deferred employer matches. Both of these argu-
ments view the employer match as a workforce management tool rather than a regula-
tory response.
EMPIRICAL STUDIES ON THE IMPACT OF MATCHING CONTRIBUTIONS ON 
PARTICIPATION
Over the past 20 years, empirical studies have analyzed the effect of matching contribu-
tions on the probability of participating in 401(k) plans that rely on voluntary enrollment 
(see Andrews 1992; Bassett, Fleming, and Rodrigues 1998; Engelhardt and Kumar 2007; 
Even and Macpherson 2005; GAO 2007; Kusko, Poterba, and Wilcox 1998; Mitchell, 
Utkus, and Yang 2005; Papke 1995; Papke and Poterba 1995; Yakoboski 1994). The 
magnitude of the results varies considerably depending on the type of database used, the 
methodologies employed, and the assumptions made. However, the consensus is that for 
401(k) plans that have not employed automatic enrollment, an employer match has had 
a positive impact on plan participation (see chapter 15 for additional discussion of and 
experience with these issues).
An important caveat is that most survey data do not contain detailed information 
on plan design. In an attempt to mitigate this shortcoming, Mitchell, Utkus, and Young 
(2005) use 2001 data on 500 401(k) retirement plans covering nearly 740,000 employ-
ees to evaluate how employer matching incentives affect retirement saving levels. Their 
analysis includes two notable innovations. First, they evaluate employee saving behavior 
separately for highly compensated and non–highly compensated employees at the firm 
level. Second, in an attempt to deal with nonlinear 401(k) matching formulas (explained 
in more detail below), they bifurcate these formulas into an “incentive element” (the 
degree to which the employer matches various increments of employee compensation) 
and a “liquidity element” (indicating how much the employee must contribute in order to 
receive the entire employer incentive payment).
Their ordinary least squares regression analysis finds that every 10 percent increase 
in the match rate raises NHCE participation rates by about 1 percentage point. However, 
for this group, the participation incentives (3–6 percent of pay) are statistically insignifi-
cant and turn negative for matches above 6 percent of pay. As a consequence, the authors 
of this analysis conclude that the incentive effects of employer matching contributions are 
small:

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
63
The empirical model implies that close to 65 percent of NHCEs at the typical firm 
would join their 401(k) plan regardless of the presence of a match. Plan participation 
would be estimated to rise over a narrow range, by 5 to 15 percentage points, respond-
ing to a range of match offerings, from a modest ($0.25 per dollar on the first 3 per-
cent of pay) to a very generous match ($1.00 per dollar up to 6 percent of pay). At the 
modal promised employer match ($0.50 per dollar on 6 percent), over one-quarter of 
NHECs fails to participate in the 401(k) plan; even with a generous match, more than 
20 percent still fail to join (Mitchell, Utkus, and Yang 2005, 16).
Given that the participation percentages for certain groups of eligible participants 
(especially young and low-income workers) have increased substantially under automatic 
enrollment, many observers have wondered whether the matching contributions would 
continue to be associated with higher participation rates under these plans.
Beshears and others (2007) estimate the impact of the employer match on savings 
plan participation under automatic enrollment in two ways. First, they analyzed a plan 
sponsor with an automatic enrollment 401(k) plan that replaced its employer match with 
a nonelective contribution.
8
 They find that plan participation rates decreased by at most 
5–6 percentage points among new hires after the plan change.
Second, they pooled data for nine firms with automatic enrollment to identify the 
relationship between participation rates and the match. They find that a 1 percentage 
point decrease in the maximum potential match was associated with a 1.8–3.8 percentage 
point decrease in plan participation at six months of eligibility.
Based on these findings, the authors estimate that for a typical employer match (that 
is, 50 percent match on the first 6 percent of pay), eliminating the match under an auto-
matic enrollment plan could reduce plan participation by 5–11 percentage points.
Dworak-Fisher (2011) uses microdata from the National Compensation Survey to 
develop a new line of research on the impact of employer matches on 401(k) participa-
tion rates. The study finds that for those with the lowest income, employer matches have 
little or no effect on participation, but automatic enrollment has dramatic effects. Among 
those in the middle income group, employer matches have substantial effects, which may 
be larger than the effects of automatic enrollment. However, these results should be inter-
preted with care, as they are based on microdata from the respondents in 2002–03, and 
only a small percentage of the plans in the sample (6 percent) were governed by automatic 
enrollment provisions at that time.
EMPIRICAL STUDIES ON THE IMPACT OF MATCHING CONTRIBUTIONS ON 
CONTRIBUTION BEHAVIOR IN VOLUNTARY ENROLLMENT 401(K) PLANS
Logically, the notion that an employer match increases the incentive for an employee to 
contribute to a 401(k) plan appears incontrovertible. The analysis becomes more complex 
with respect to the level of contributions the employee will make, for two reasons.
9
First, although a larger match rate provides a greater financial incentive for the 
employee to contribute (at least within a specified range), employees may have a certain 
target in mind with respect to the total (employee and employer) contribution that needs 
to be made each year to satisfy their personal financial planning objectives. For example, 
if an employee has determined that he or she needs to save a total of 9 percent of compen-
sation, the required employee contribution would be 6 percent if the employer matched 

64 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
50 percent up to 6 percent of compensation but only 4.5 percent if the employer matched 
100 percent up to (at least) 4.5 percent of compensation. Thus, for some employees, a 
higher match rate may result in a lower employee contribution rate.
Second, empirical analysis that considers only the match rate (as opposed to the 
match cap or the interaction between the two) may provide unexpected results. For exam-
ple, if an individual employee’s primary concern is to make sure he or she receives the 
maximum match possible, the employee would be more likely to contribute up to the 
level of the match cap. In that case, an employer match of 50 percent of the first 6 per-
cent of compensation would likely generate a larger employee contribution rate than one 
incented by matching 100 percent of the first 3 percent of compensation—even though 
the maximum total employer match for the worker would be 3 percent of compensation 
in either case.
These phenomena help explain the results from some of the early empirical work 
in this area. Using plan data from the Form 5500 filed annually by ERISA-qualified 
plans with the U.S. government, Papke (1995) finds substantial employee contribution 
increases when an employer moves from a zero to a small or moderately sized match rate 
proxy. At higher match rates, however, employee contributions fall. Using a subset of 
the EBRI/Investment Company Institute (ICI) 401(k) database with salary information, 
Holden and VanDerhei (2001) perform a regression analysis of the influence of the match 
rate on participants’ contribution rates. They find that participant before-tax contribution 
rates fell minimally as the employer match rate rose.
10
 However, as the match cap chosen 
by the employer increased, participant contribution rates rose.
Kusko, Poterba, and Wilcox (1998) use employee-level data from the 401(k) plan 
at a medium-size U.S. manufacturing firm to analyze the participation and contribution 
decisions of eligible workers.
11
 Their analysis suggests that contribution decisions of eligi-
ble employees are relatively insensitive to the rate of employer matching on worker contri-
butions and that most employees maintain the same participation status and contribution 
rate year after year, despite substantial changes in the employer’s match rate. Moreover, 
they find that institutional constraints on contributions, imposed by either the employer 
or the IRS, are an extremely important influence on contributor behavior.
Yakoboski and VanDerhei (1996) confirm these results in their analysis of partici-
pant data from three large 401(k) sponsors. Moreover, they find significant clustering 
around the match cap, as illustrated below:
 
• Company A had a maximum pretax contribution of 9 percent of earnings and a 
match rate of 30 percent for the first 5 percent of earnings. A total of 21 percent 
of participants contributed 5 percent of pay to the plan, 45 percent contributed 
9 percent of pay, and 1 percent contributed up to the allowable maximum for 
that year under Section 402(g). The average deferral percentage for Company A 
was 6.7 percent.
 
• NHCEs at Company B were allowed to contribute a maximum of 15 percent 
of pretax income, with a 100 percent match on the first 3 percent of earnings. 
Twenty-one percent contributed 3 percent of pay, 10 percent contributed 15 per-
cent, and 0.1 percent contributed the allowable limit.
12
 The average deferral rate 
was 5.4 percent.

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
65
 
• Highly compensated employees at Company B were allowed to contribute a 
maximum of 10 percent of pretax income, with a 100 percent match on the first 
3 percent of earnings. Fifteen percent contributed 3 percent of pay, 10 percent 
contributed 10 percent, and 15 percent contributed at the Section 402(g) limit. 
The average deferral rate was 5.9 percent.
 
• Company C had a maximum pretax contribution of 16 percent of earnings and a 
match rate of two-thirds on the first 6 percent of earnings. A total of 30 percent 
of participants contributed 6 percent of pay to the plan, 7 percent contributed 
16 percent of pay, and 12 percent contributed up to the Section 402(g) limit. The 
average deferral percentage for Company C was 6.3 percent.
Although this analysis is based on the experience of only three plan sponsors, it 
shows that in addition to individual-specific characteristics (for example, age, wages, and 
tenure), employee contribution behavior can be influenced to a large extent by plan design 
variables (the match cap and plan limits for pretax contributions) as well as the maximum 
legal annual deferral limits.
VanDerhei and Copeland (2001) attempt to deal with these plan design influ-
ences on employee contribution behavior by working with a small subset of the EBRI/
ICI 401(k) database,which allowed them to track 137 “pure” matching formulas (that is, 
formulas without a nonelective contribution).
13
 Participants in the database were excluded 
if they were under age 20 or over 64, had been with the current employer for less than 
one year, or earned less than $10,000 a year. Applying each of these screens and deleting 
participants with existing account balances who did not make employee contributions in 
1998 left a total of 163,346 participants for analysis.
In previous research, the level of contributions was estimated by assuming that it 
was a function of demographic variables and some measure of a match rate of the plan. 
However, this approach fails to account for the fact that some plans have different match 
rates for different levels of the percentage of compensation contributed. For example, a 
plan may offer a dollar-for-dollar match for the first 2 percent of compensation contrib-
uted and a 50 percent match for the next 3 percent of compensation contributed. In addi-
tion, this approach does not clearly distinguish between a plan that matches 50 percent of 
contributions for the first 4 percent of compensation from plans that match 50 percent of 
contributions for the first 6 percent of compensation.
As the data used in this research contain plan-specific matching formulas, the actual 
match rate at each percentage level of contributions is known. Therefore, VanDerhei and 
Copeland (2001) use an estimation procedure that takes advantage of knowing the differ-
ing incentives that an employee eligible to contribute to a 401(k) faces at each percentage 
of compensation level of contributions.
The parameters of a model for the first increment can be estimated from the entire 
sample by dividing it into two groups: employees who make the contribution and those 
who do not. The parameters of a model for the second increment can be estimated by 
dividing the subsample of employees who make the first incremental contribution into 
employees who make the next 1 percent of compensation contribution and employees 
who do not. Successive iterations are estimated until the maximum plan limit of all match 
formulas is obtained. In this model, the decision of an eligible employee is examined at 

66 
MATCHING CONTRIBUTIONS FOR PENSIONS: A REVIEW OF INTERNATIONAL EXPERIENCE
each level of possible contributions. This strategy captures changes in the incentives of con-
tributing an additional percentage of compensation and controls for whether the partici-
pant is allowed to contribute (for example, in some plans a highly compensated employee 
might be barred from making additional contributions beyond 6 percent of compensa-
tion, whereas an NHCE might be allowed to contribute 15 percent of compensation).
Because each percentage of compensation contributed is modeled as a separate deci-
sion, an employee either contributes or does not contribute the ith percent of compensa-
tion. As a binomial variable is the dependent variable in the estimation, a probit regression 
is used to represent the nature of the dependent variable. The conditional estimation 
model for an individual to contribute (C) is
P
(C
i 
= 1) = a
i 
+ b1
iMTCHi 
+ b2
iAMTCHi 
+ b3
iage
 + b4
iage 
2 
+ b5
iage 
3 
+ 
b6
iwage
 + b7
iwage 
2
+ b8
iwage 
3 
+ b9
itenure
 + b10
itenur
e 
2 
+ b11
itenure 
3
+ e
i
, 
where i represents the ith interval of contributions; MTCH and AMTCH are the plan 
match variables; age, wage, and tenure (and their squared and cubed values) are the demo-
graphic variables; and ei is the error term for each interval regression.
The demographic variables are self-explanatory. The plan match variable, MTCHi, 
is the match rate percentage at each level of contribution. However, this variable was 
thought to be insufficient to capture the entire incentive of contributing at each interval. 
For example, a participant facing a match rate of 50 percent for the first 6 percent of com-
pensation might have more incentive to contribute the second percent of compensation 
than the fifth percent because of the value of additional matches that would be foregone if 
he or she decides not to contribute at that level. In other words, given that the employee 
already contributed 4 percent, the decision not to defer the fifth percent of compensation 
costs the employee the 50 percent match on the fifth percent plus the option to receive a 
50 percent match on the sixth percent (a current match of 0.5 and a future match of 0.5). 
In contrast, the decision not to defer the second percent of compensation, given that an 
employee already contributed 1 percent, costs the employee the 50 percent match on the 
second percent plus the option to receive a 50 percent match on the third, fourth, fifth, 
and sixth percent (a current match of 0.5 and a future match of 0.5 * 4 = 2.0 percent). 
Therefore, an additional match value variable (AMTCHi) was included, representing the 
option value of the additional match value at a given percentage of contributions.
A series of probit regressions was conducted for each possible level of percentage 
of compensation contributed in order to estimate the total level of contributions to this 
model. Each estimation of the percentage of compensation contributed has a different set 
of results for the differing incentives of contributing that level of compensation. In the 
first interval, the probability of an eligible employee contributing anything to the plan 
is estimated. In the second interval, the probability that a participant who contributed 
1 percent of compensation contributes a second percent of compensation is estimated; 
people who do not contribute 1 percent of compensation are not included in the estima-
tion of the second percent of compensation. Under this process, only employees who are 
faced with the decision to contribute the next percentage are investigated.
14
Two specifications of this basic model structure were estimated for employees’ deci-
sions to contribute to a 401(k) up to 18 percent of compensation. The first specification 
included only the match rate (MTCHx, x = 1, 2, … , 18) immediately facing the employee 

3.  MATCHING CONTRIBUTIONS IN 401(K) PLANS IN THE UNITED STATES 
67
at each level of compensation without any plan dummies. The second specification added 
the additional plan match value variable (AMTCHx, x = 1, 2, … , 18).
To illustrate this model, VanDerhei and Copeland (2001) compute the probability 
that a 22-year-old employee with one year of tenure with his or her current employer and 
wages of $15,000 who already contributed 4 percent of compensation will contribute 
an additional percent. This value is estimated to be as low as 81 percent if this is the last 
interval of compensation that is matched by the employer (that is, the additional match is 
equal to zero). In contrast, the same employee is estimated to have a 90 percent probabil-
ity of contributing the extra percent of compensation if the option to earn an extra 1 per-
cent of employer match would be forfeited if the employee continued to contribute the 
maximum percentage of compensation matched. In each of the three intervals illustrated, 
the model predicts that employees with the lowest estimated probability of contributing 
the extra percent of compensation when the additional match is set equal to zero (young 
employees and employees with lower levels of wage and tenure) will be the most sensitive 
to increases in the additional match level.
Figure 3.1 displays the predicted contributions for stylized participants under typi-
cal plan matching formulas. It shows that older participants and participants with higher 
levels of wage and tenure are expected to make higher employee contributions for a given 
plan design, and it also allows investigation of how the change in plan design will affect 
the expected contribution behavior. For example, a change from a 50 percent match on 
the first 6 percent of compensation to a 75 percent match over the same range results in an 
expected increase in employee contributions for all of the stylized participants. Figure 3.1 
also reveals the ability of the model to predict contributions under a two-tier matching 
formula (for example, a 75 percent match on the first 2 percent of compensation, decreas-
ing to 50 percent for the next 3 percent of compensation), as well as its ability to model 
employees participating in a plan with no employer match.
FIGURE 3.1  Predicted employee contributions for selected persons and plan matching formulas
2
4
6
8
10
12
age = 55,
wage = 45 
age = 40,
wage = 25 
age = 35,
wage = 20 
age = 22,
wage = 15