You are on page 1of 3

Primary question:

Do we learn more when we choose to or when we are told to?


Voluntary (Internal) vs. Involuntary (External)

Background: Sources of automotive recalls
1. External mandate: National Traffic and Motor Vehicle Safety Act (NTMVSA) (1966)
- All vehicles and equipment must comply to the standard
o Laboratories at National Highway and Traffic Safety Administration (NHTSA)
o Vehicle owners ! NHTSA ! NTMVSA
o Other government agencies !NHTSA ! NTMVSA
2. Voluntary recall by automakers (prior to NHTSA involvement)
- Automakers own tests, inspection procedures and info-gathering systems

Hypotheses
*** Test of the basic learning effects (H1)
ASSUMPTION: Traditional research on learning curve
o Organizational learning ! decrease in unfavorable outcomes (Lave and March,
1975)
" Proxy for learning/knowledge improvement = production experience
" Decrease in unfavorable outcome = reduction in recall #
- Experience ! org. knowledge ! improved error detections/design/org. process ! less errors
Therefore,
H1: The higher the level of prior production experience, the lower the number of subsequent recalls.

*** Contesting theories of volition and external mandate:
Voluntary learning (H2)
ASSUMPTION: Autonomy in the voluntary action
- Marcus and Nichols, 1999 (Nuclear plants); Hackman and Oldham, 1980; Pfeffer, 1995
(Individual level research on benefits of autonomy over jobs)
- Organizational autonomy = acceptance by members + permanent change
- Problem ! voluntary action ! low resistance/improved competencies/acceptance ! future
recall reduction
Therefore,
H2: Voluntary recalls will lower the subsequent recall rate.

Additional support for voluntary learning (H2A)
ASSUMPTION: No autonomy in involuntary action
- Involuntary (externally mandated) events ! defensive response = shallow, unwilling
response ! no knowledge to penetrate into the routines & practices of org. (Marcus, 1988)
- Centralization of authority ! demotivation ! mechanical compliance ! increase in future
error (Perrow, 1984)
- Institutional pressure ! decoupling ! symbolic compliance ! no deep learning ! future
errors (Meyer and Rowan, 1977)
Therefore,
H2A: Involuntary recalls will result in shallower learning process than voluntary recalls.

Involuntary learning (H3)
ASSUMPTION: Involuntary recalls draws more attention to a problem
- Organization as systems of structurally distributed attention (March and Simon, 1958; Cyert
and march, 1963; Hutchins, 1995)
- Learning occurs when attention is focused on the problem (Radner and Rothchild, 1975;
Winter, 1981)

ASSUMPTION: Organizational inertia can be broken by external shock
- Salient events help overcome organizational inertia (March et al., 1991)
- Only external shock can stimulate inert org. to learn and innovate (Cyert and March, 1963,
Greve, 1998; Sitkin, 1992)
- Involuntary recalls ! Salience/attention-getting ! external shock ! overcome org. inertia
! reduction in error.
Therefore,
H3: Involuntary recalls will lower the subsequent recall rate.

*** Generalist and specialists (H4A, H4B)
- ASSUMPTION: Not all automakers learn equally from recall events. Generalists are more
complex.
- Barnett et al., 1994 (specialists, bank, higher returns); Ingram and Baum, 1997a (generalist
less affected by environment)
- Generalists = broader resource space/ products ! problems of controlling/coordination
- Complex org are less likely to change (Barnett and Carroll, 1995) ! less likely to learn
Therefore,
H4A: Generalist automakers will have more recalls than specialists
H4B: The ability to learn from prior voluntary and involuntary recalls will be weaker for generalist
automakers than specialists.

Method
Sample and Data
- Unit of analysis: Year-automaker level
- Data source: NHTSA (recall info), Wards Automotive Yearbook, Motor Vehicle Facts and
Figures, Automotive News, Automotive News Market Data Book, Standard Catalog of
Imported Cars 1946-1990, Worldscope, COMPUSTAT, Disclosure, and 10-K Reports
(organizational info)
- Final sample: 47 automakers, who experienced 2,287 recalls between 1966-1999

Variables
Dependent variables:
1. Annual number of severe recalls for each automaker
*Severe recalls: NHTSAs criteria on hazard level of recall. (a>b>c>d) only a, b level recalls are
counted.
2. All severe recalls separated into voluntary severe recalls and involuntary severe recalls.

Independent variables:
1. Production experience (H1): # of cars produced by a focal automaker since the beginning of its
production. (lagged and logged)
2. Cumulative voluntary and involuntary recalls (H2,2A,3):
I. Absolute number of voluntary/involuntary recall experience within past 3 years.
II. Relative proportion of voluntary recall to involuntary recall experience within past 3 years.
3. Generalism and Specialism (H4A,4B): Range of engine capacity = range of automakers
technological offerings (Largest Smallest)

Control variables:
1. Age (current year production begin year)
2. Size (sales)
3. Presidential administration (democratic admin ~(+)~ automotive recalls)
4. Industry competition (population density)
5. Time elapsed from 1966 (trend effects or development of error detection tech)
6. Automakers home country (factor variable, within-group estimator)

Statistical model:
Negative binomial model (count model + over-dispersion problem) + conditional fixed models

Results

DV1: All recalls DV2: Involuntary recalls DV3: Voluntary recalls
H1: Production experience
Neg Sig
(H1 marginally
supported)
-
Neg Sig
(H1 supported)
H2: Voluntary recalls -
Neg Sig
(H2 supported)
Pos Sig
(H2 not supported)
H2: Voluntary recalls (ratio) -
Neg Sig
(H2 supported)
Neg Insig
(H2 not supported)
H3: Involuntary recalls -
Neg Insig
(H3 not supported)
Pos Sig
(H3 not supported)
H4A: Generalism -
Positive insig
(H4A not supported)
-
H4B: Generalism X
Voluntary (ratio)
-
Negative sig
(H4B not supported)
-
H4A: Generalism - -
Positive sig
(H4A supported)
H4B: Generalism X
Voluntary (ratio)
- -
Negative insig
(H4B not supported)

H2A is also supported separately from above regressions.
- Pages of technical reports associated with voluntary recalls are significantly longer than
voluntary recall.
- Number of solved cases in voluntary recalls are significantly greater than those of involuntary
recalls.
Summary:
1. Basic learning effect reduces subsequent voluntary recalls. (H1)
2. Learning from voluntary recalls reduces only subsequent involuntary recalls. (H2)
3. Learning from involuntary recalls are shallower than voluntary recalls. (H2A)
4. Generalists have more voluntary recalls than specialists (H4a)
5. Generalists learn better from voluntary recalls than specialists (H4b x)

Further discussion
The role of volition in learning is effective in reducing involuntary errors. ! Interesting boundary
condition.
1. There exist different types of failure. Voluntary recall may be viewed positively, while
involuntary recalls may be seen as severely negative. Therefore, learning from failure should be
sensitive to the context of failure. Only when failures are viewed negative, learning takes place.
2. Recalls are positively related to future voluntary recalls. Learning do not reduce errors, but they
seem to trigger proactively respond and initiate failures to avoid others.
3. Production experience may imply exploitative learning, which corresponds to internal error
detection, but not with error detections from external sources.

You might also like