1. Using organization services for personal use (e.g.

long distance telephone calls)

2. Giving gifts/favors for preferential treatment
3. Taking longer than necessary to do a job
4. Divulging confidential information.
5. Doing personal business on work time
6. Concealing ones errors
7. Passing blame for errors to an innocent co-worker
8. Claiming credit for someone elses work
9. Falsifying time/quantity/quality reports
10. Padding expense account more than 10 %
11. Calling in sick to take a day off
12. Authorizing a subordinate to violate rules
The first form of counterproductive behavior is sloppy work, i.e. making errors in the
identification task. Coins could be confused, or there could be too many or too few coins
described in comparison to those found in the plastic bag. Of course, participants could differ in
their abilities and therefore could have different costs of complying with the instructions, but
they could make sure they are doing a perfect job by putting enough effort into it. Consequently,
all errors can be interpreted
as counterproductive behaviour rather than incompetence.8 On average, participants make 6.2
mistakes, which corresponds
to 5 percent of the average productivity. The error rate (proportion of errors relative to the
apparent output) ranges from 0 to
0.9 with a mean value of 0.05. There are no significant treatment differences in the number of
errors made (KruskalWallis
test, p = 0.522) or in the error rate (KruskalWallis test, p = 0.867). Out of 108 participants, 21
fulfilled the identification task
without any errors.
The second form of counterproductive behavior is an incorrect report. We call the discrepancy
between the report and the productivity the total reporting error. On average the report is 6.2
coins above the actual productivity. This difference is significantly different from 0 (Wilcoxon, z
= 6.42, p < 0.01, two-tailed). We find significant differences in the total reporting error between
treatments (KruskalWallis test, p = 0.096). It is highest under competition (U-test, z = 1.98, p =
0.05, two-tailed) and lowest in the piece rate treatment compared to the other treatments (U-test,
z=
1.74, p = 0.08, two-tailed).
The reporting error could be partly due to mistakes in the identification task (i.e. participants
could report the number

of coins identified and not correct for mistakes). We define such discrepancies due to mistakes as
indirect inflation. On top
of that, participants could also directly inflate the report (report a higher number than the number
of coins identified on the
identification sheet).
To see to what extent participants indirectly inflate the report through mistakes in the
identification task, we calculate
the difference between the apparent output (number of bars on the identification sheet) and the
actual output (productivity).
This difference is equal to 4.98 coins on average, that is, most of the reporting error comes from
the mistakes but not all of
it. On top of that, participants also inflate the report directly. The average difference between the
report and the apparent
output is 1.26 coins. We observe significantly more indirect inflation than direct inflation
(Wilcoxon, z = 5.78; p < 0.01, twotailed).
There are no significant differences in indirect inflation across pay schemes (KruskalWallis test,
p = 0.590), but we
do observe significant treatment differences in direct inflation (KruskalWallis test, p = 0.037).
Direct inflation is significantly
higher under competition (U-test, z = 2.57, p = 0.01, two-tailed) than in the other two treatments.
There are 21 participants (19 percent) who inflate the report by 5 coins or more (which we call
strong inflators).9 The
average total inflation of these strong inflators is 30.14 coins, which is 27 percent of productivity
and accounts for 94 percent
of the observed inflation in this experiment.10
Finally, a third way of behaving counterproductively is theft. Only one individual in the piece
rate treatment stole 3 two
euro coins and 5 (all) one euro coins.
Summarising, we find evidence of counterproductive behaviour under all three pay schemes. On
average, the total amount
of counterproductive behaviour amounts to 10 percent of productivity and is driven by errors in
the identification task as
well as reporting errors. Indirect inflation of the report is more prevalent than direct inflation,
although we do find more
direct inflation under competition than under the other two treatments. There is essentially no
theft taking place.
Our findings are consistent with different social norms associated with different forms of
counterproductive behaviour.
Theories of social norms propose that violating a social norm involves a non-monetary (moral)
fixed cost. The assumption is
that offenses against social norms create a disutility for individuals, for example a feeling of guilt
(Battigalli and Dufwenberg,
2007) or a disadvantageous change in self-perception (Bnabou and Tirole, 2011). The social
norm is probably strongest for
theft (Robinson and Benett, 1995), which is in fact illegal. A participant who cares about social
norms will rather choose a

different type of counterproductive behaviour, such as inflation of the report. We would also
anticipate to see more indirect
inflation of the report (inflation through erroneous work) than direct inflation (inflation through
lying), because the latter
may appear more deliberate and dishonest than the former. Errors in the identification task may
more easily pass as honest
mistakes and may be perceived as more acceptable than other forms of counterproductive
behaviour.
The reason why we observe higher productivity and more direct inflation under competition
could be because of otherregarding
preferences (Rabin, 1993; Fehr and Schmidt, 1999; Bolton and Ockenfels, 2000). The
competition treatment always
leads to an unequal outcome and higher productivity or inflating the report are ways to shift the
inequality in ones favour
(Grund and Sliwka, 2005)