Discussion: View Thread

I agree with Fred Nickols, and Deming before him, --- scrap appraisals ---
at least the kind of appraisals most organizations do. But there is another
way which has worked quite well (Honeywell Computer business, 1980's). It
sounds awfully mechanical at first but hold the howls until you think about
it a while.
First, presume that any organizational component with 15 or more employees
should exhibit a performance profile approximating a normal distribution.
Any group skewed higher simply is not being sufficiently challenged and any
skewed lower are not getting the competency development (quality of
management) they deserve.
Second, determine how much the median should be making (market rates).
Third, examine the gap of each individual and calculate the compensation
adjustment for the period of performance.
Fourth, watch the appraisals come in and if they deviate from the expected
normal distribution, challenge the managers to justify such. Any manager
who insists that his/her "portfolio of assets" is outstanding obviously has
it too easy so should not be compensated as much as is a manager with just a
normal distribution of overachievers.
Once you get this principle in place then you can start working the median.

Sound like Taguchi?

Maybe, but we were doing evolutionary optimization in the 1960's.
While he was still in knee pants?

Jack Ring
32712 N. 70th St., Snottsdale, AZ 85262
480-488-4615
Information, having no mass nor rigidity, cannot be pushed, only pulled.

----- Original Message ----- >
> Date: Thu, 27 Apr 2000 14:17:07 -0400
> From: "Jackson, Bonnie" <BJacks5@RESPARK.RTTONET.PSU.EDU>
> Subject: Compensation tied to appraisals
>
> I'm working with a company who is changing their appraisal process.
> Previously, their appraisals were tied to compensation and they found that
> all employees were rated very high on their appraisals. They have
recently
> gone to an objective review process. The question from their employees
> is..... how is this process tied to compensation. If you do well on the
> review, doesn't that count towards increased compensation. My question to
> the list is..What models have folks seen that work well to determine
> compensation while still rewarding high performers. What part should the
> review play in determining compensation? What other factors are
> organizations using?
>
> Thanks.
>
> Bonnie Jackson
>

On 27 Apr 00, at 22:45, Jack Ring wrote:

> I agree with Fred Nickols, and Deming before him, --- scrap appraisals ---
> at least the kind of appraisals most organizations do. But there is
> another way which has worked quite well (Honeywell Computer business,
> 1980's). It sounds awfully mechanical at first but hold the howls until
> you think about it a while. First, presume that any organizational
> component with 15 or more employees should exhibit a performance profile
> approximating a normal distribution.

This is straight-forward mumbo jumbo of the worst kind. The irony
is that you mention such an incorrect use of statistical concepts
and Deming in the same message.

There is NO WAY (repeat sotto voce -no way) that you can justify
or support applying a normal distribution to anything but randomly
selected populations. The only way this makes any sense is if you
hire randomly, completely randomly.

If you actually do have a normal distribution underlying employee
performance, your hiring practices would be absolutely broken.

Since the assumption about normal distributions is wrong, so is
everything that follows.

Lest you don't understand the concept here: If you take all the
heights of players in the NBA, do you think you will get a normal
distribution? Or weights of people who go to weight watchers?

It's bunk, and it's intro. statistics level stuff.

If this is the stuff people are telling managers in development, we
are in deep trouble.


Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

OK, then, don't hold the howls until you think.
Now, think.

Jack Ring
32712 N. 70th St., Snottsdale, AZ 85262
480-488-4615
Information, having no mass nor rigidity, cannot be pushed, only pulled.

----- Original Message -----
From: <rbacal@escape.ca>
To: Jack Ring <jring@AMUG.ORG>; Management Education and Development
Discussion <MG-ED-DV@MAELSTROM.STJOHNS.EDU>;
<MG-ED-DV@MAELSTROM.STJOHNS.EDU>
Sent: Thursday, April 27, 2000 11:44 PM
Subject: Re: [MG-ED-DV] Appraisals and Compensation


>
>
> On 27 Apr 00, at 22:45, Jack Ring wrote:
>
> > I agree with Fred Nickols, and Deming before him, --- scrap
appraisals ---
> > at least the kind of appraisals most organizations do. But there is
> > another way which has worked quite well (Honeywell Computer business,
> > 1980's). It sounds awfully mechanical at first but hold the howls until
> > you think about it a while. First, presume that any organizational
> > component with 15 or more employees should exhibit a performance profile
> > approximating a normal distribution.
>
> This is straight-forward mumbo jumbo of the worst kind. The irony
> is that you mention such an incorrect use of statistical concepts
> and Deming in the same message.
>
> There is NO WAY (repeat sotto voce -no way) that you can justify
> or support applying a normal distribution to anything but randomly
> selected populations. The only way this makes any sense is if you
> hire randomly, completely randomly.
>
> If you actually do have a normal distribution underlying employee
> performance, your hiring practices would be absolutely broken.
>
> Since the assumption about normal distributions is wrong, so is
> everything that follows.
>
> Lest you don't understand the concept here: If you take all the
> heights of players in the NBA, do you think you will get a normal
> distribution? Or weights of people who go to weight watchers?
>
> It's bunk, and it's intro. statistics level stuff.
>
> If this is the stuff people are telling managers in development, we
> are in deep trouble.
>
>
> Visit the work911.com supersite at http://www.work911.com
> for work related articles, or to find almost anything including
> book reviews and suggestions, discussion lists and more.
>

On 28 Apr 00, at 7:35, Jack Ring wrote:

> OK, then, don't hold the howls until you think.
> Now, think.

Is that your best shot at justifying the use of an assumed normal
distribution?


Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

Earlier, Jack Ring wrote that he agreed with me about scrapping performance
appraisals. He added the possibility of assuming a normal
distribution. Robert Bacal termed that "mumbo jumbo of the worst kind" and
added.

> > There is NO WAY (repeat sotto voce -no way) that you can justify
> > or support applying a normal distribution to anything but randomly
> > selected populations.

<snip a little bit>

> >> Lest you don't understand the concept here: If you take all the
> > heights of players in the NBA, do you think you will get a normal
> > distribution? Or weights of people who go to weight watchers?
> >
> > It's bunk, and it's intro. statistics level stuff.

Robert, I think your comments above assume something that needs to be made
explicit, namely, the reference point. While it is true that the
distribution of height among players in the NBA and the distribution of
weight among people who go to weight watchers won't match the distribution
of height and weight in the general population, owing to preselection, the
distribution of height among NBA players and the distribution of weight
among those who attend weight watchers might well assume a normal
distribution with respect to those two populations. I learned this a long,
long time ago as the principle of "bell curves within bell curves." The
range of heights among members of the NBA and weights among people who
attend weight watchers would be very different from the range in the
general population but the distributions in these two subsets of the
general population might in fact be quite bell-shaped. I confirmed this
possibility with three of my colleagues at ETS (where statistics approaches
a religion). As one of them put it to me, "It's possible that you won't
get a bell shaped curve of heights among the NBA players (or the weight
watcher attendees) but that's the way to bet."

That's good enough for me. So Jack's admonition to hold the howls in
abeyance seems sound enough to me.
--

Fred Nickols
The Distance Consulting Company
"Assistance at A Distance"
http://home.att.net/~nickols/distance.htm
nickols@worldnet.att.net
(609) 490-0095

Jack,

I have given your concept a lot of thought and like it very much. Nice stuff. Thank you for sharing it. I particularly like how it elegantly creates a dual pressure on the manager: to ensure performance to a median (standard) and to report/appraise honestly.

Question: Am I incorrect when I sense some relationship or kinship of this approach to some of the production control methodologies that "Six Sigma"and the like are based upon?

Best wishes.

Ed
Drive On!


>>> jring@AMUG.ORG 04/28/00 01:45AM >>>
I agree with Fred Nickols, and Deming before him, --- scrap appraisals ---
at least the kind of appraisals most organizations do. But there is another
way which has worked quite well (Honeywell Computer business, 1980's). It
sounds awfully mechanical at first but hold the howls until you think about
it a while.
First, presume that any organizational component with 15 or more employees
should exhibit a performance profile approximating a normal distribution.
Any group skewed higher simply is not being sufficiently challenged and any
skewed lower are not getting the competency development (quality of
management) they deserve.
Second, determine how much the median should be making (market rates).
Third, examine the gap of each individual and calculate the compensation
adjustment for the period of performance.
Fourth, watch the appraisals come in and if they deviate from the expected
normal distribution, challenge the managers to justify such. Any manager
who insists that his/her "portfolio of assets" is outstanding obviously has
it too easy so should not be compensated as much as is a manager with just a
normal distribution of overachievers.
Once you get this principle in place then you can start working the median.

Sound like Taguchi?

Maybe, but we were doing evolutionary optimization in the 1960's.
While he was still in knee pants?

Jack Ring
32712 N. 70th St., Snottsdale, AZ 85262
480-488-4615
Information, having no mass nor rigidity, cannot be pushed, only pulled.

----- Original Message ----- >
> Date: Thu, 27 Apr 2000 14:17:07 -0400
> From: "Jackson, Bonnie" <BJacks5@RESPARK.RTTONET.PSU.EDU>
> Subject: Compensation tied to appraisals
>
> I'm working with a company who is changing their appraisal process.
> Previously, their appraisals were tied to compensation and they found that
> all employees were rated very high on their appraisals. They have
recently
> gone to an objective review process. The question from their employees
> is..... how is this process tied to compensation. If you do well on the
> review, doesn't that count towards increased compensation. My question to
> the list is..What models have folks seen that work well to determine
> compensation while still rewarding high performers. What part should the
> review play in determining compensation? What other factors are
> organizations using?
>
> Thanks.
>
> Bonnie Jackson
>

On 28 Apr 00, at 18:13, Fred Nickols wrote:


> Robert, I think your comments above assume something that needs to be made
> explicit, namely, the reference point. While it is true that the
> distribution of height among players in the NBA and the distribution of
> weight among people who go to weight watchers won't match the distribution
> of height and weight in the general population, owing to preselection, the
> distribution of height among NBA players and the distribution of weight
> among those who attend weight watchers might well assume a normal
> distribution with respect to those two populations.

Or it may not. As it is in the workplace. Probably a better example
is the weight of professional linemen in football. This is an empirical
question. If I get some time, I may take a few minutes to
demonstrate, but you all can do it at home. Take any population
that is heavily selected (not randomly, but filtered). Compute the
mean and the median (that's a quick and dirty way to check about
normalcy.



I learned this a
> long, long time ago as the principle of "bell curves within bell curves."
> The range of heights among members of the NBA and weights among people who
> attend weight watchers would be very different from the range in the
> general population but the distributions in these two subsets of the
> general population might in fact be quite bell-shaped. I confirmed this
> possibility with three of my colleagues at ETS (where statistics
> approaches a religion). As one of them put it to me, "It's possible that
> you won't get a bell shaped curve of heights among the NBA players (or the
> weight watcher attendees) but that's the way to bet."

Ok, even if we leave aside the math of it, I'm sure those ETS folks
will also tell you that you don't ASSUME a normal distribution and
then operate and apply statistical tests as if it WAS normal. You
calculate it first. That's hard to do in the realm of appraisals, to test
form normal distributions, but like much in the field, people fudge
and fake to sound scientific.

In any event to talk a silly example. I only hire people who weigh
exactly 212 pounds. I have a staff of 250 people weighting 212
pounds. You're going to tell me that this selected group will
comprise a normal distribution?

You get what you select for. If the hiring practice is random you
should get a normal distribution, all things being equal. The more
effective you are at selecting good people the less normal your
distribution will be, all other factors being equal.

If we are going to invoke statistical concepts here, could we please
at least do it correctly?

There's also another aspect to this that Dr. Sullivan talked about at
the performance management conference...if this continues, I'll
mention it.



Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

I read Jack's post on employees approximating the normal distribution,
setting up the appraisal system according to that and then determine why the
distribution may deviate. That type of system has been around for years.
Many organizations will dictate where employees can fall in the appraisal
system and make sure the values assigned meet those criteria. If your
employees appraisal is too high, they are just changed. The problem with
doing the system this way, is it ensures that some employees will be below
standards and some will be above and the rest will be in the middle (normal
distribution). What this ignores is establishing performance goals that are
criterion based rather than norm based. You see the same thing in school
testing, criterion versus norm referenced. If the goals are set correctly
and the measurement system is proper, raters are trained properly, and the
appraisals are conducted properly, then your employees can all be skewed
high or low and they may assume a normal distribution. If the performance
management and development system works correctly, then all employees can
achieve ratings well above the middle of a normal distribution, because the
ratings are based on absolute scores rather than normed on the performing
group norms. Just look at Army physical fitness testing in various units.
Initial testing may well approximate normal, due to differences in fitness
and other characteristics. Since doing 'x' number of pushups is 'x' number
of pushups, then there is no norm reference for the scoring. When the people
are trained and their fitness levels improves, they will gradually migrate
to the upper end of the scale, but it is still absolute, no norming as far
as the rating is concerned. Now, developing a measurement system for
physical tasks such as pushups, sit-ups, and running specific distances, is
fairly easy, it is more difficult to do the same with human performance
measures on the job. This is why compensation should be based on a formula
of organizational performance rather than individual performance. the
fairest form of compensation is piece work, just read about Lincoln Electric
and you can see how a true pay for performance program works. To look at a
theoretically good performance management system, look at the military.
Performance appraisals are used for assignments, schooling, and development.
they contribute some to promotion but pay decisions are automatic. Pay
raises come with longevity. if you want to make more, you do those things
necessary to be promoted. If you do not perform satisfactorily, you are sent
packing after some developmental and other attempts to improve performance.

Six sigma is a whole different issue. That deals with keeping production
deviations outside of the three standard deviations above and below the
median, which means, you re-tool or correct processes that fall outside,
somewhat less than 2% deviation rate. You can't apply that to performance
appraisals, apples and oranges.

Ken Rossi, Ed.D.
Asst. Professor of Information Systems
Hawaii Pacific University
Honolulu, HI 95813
(808) 544-1412
kgrossi@hpu.edu
rossik@hawaii.rr.com
----- Original Message -----
From: "Edward Hampton" <ehampton@MAIL.UCF.EDU>
To: <MG-ED-DV@MAELSTROM.STJOHNS.EDU>
Sent: Friday, April 28, 2000 12:30 PM
Subject: Re: Appraisals and Compensation


Jack,

I have given your concept a lot of thought and like it very much. Nice
stuff. Thank you for sharing it. I particularly like how it elegantly
creates a dual pressure on the manager: to ensure performance to a median
(standard) and to report/appraise honestly.

Question: Am I incorrect when I sense some relationship or kinship of this
approach to some of the production control methodologies that "Six Sigma"and
the like are based upon?

Best wishes.

Ed
Drive On!


>>> jring@AMUG.ORG 04/28/00 01:45AM >>>
I agree with Fred Nickols, and Deming before him, --- scrap appraisals ---
at least the kind of appraisals most organizations do. But there is another
way which has worked quite well (Honeywell Computer business, 1980's). It
sounds awfully mechanical at first but hold the howls until you think about
it a while.
First, presume that any organizational component with 15 or more employees
should exhibit a performance profile approximating a normal distribution.
Any group skewed higher simply is not being sufficiently challenged and any
skewed lower are not getting the competency development (quality of
management) they deserve.
Second, determine how much the median should be making (market rates).
Third, examine the gap of each individual and calculate the compensation
adjustment for the period of performance.
Fourth, watch the appraisals come in and if they deviate from the expected
normal distribution, challenge the managers to justify such. Any manager
who insists that his/her "portfolio of assets" is outstanding obviously has
it too easy so should not be compensated as much as is a manager with just a
normal distribution of overachievers.
Once you get this principle in place then you can start working the median.

Sound like Taguchi?

Maybe, but we were doing evolutionary optimization in the 1960's.
While he was still in knee pants?

Jack Ring
32712 N. 70th St., Snottsdale, AZ 85262
480-488-4615
Information, having no mass nor rigidity, cannot be pushed, only pulled.

----- Original Message ----- >
> Date: Thu, 27 Apr 2000 14:17:07 -0400
> From: "Jackson, Bonnie" <BJacks5@RESPARK.RTTONET.PSU.EDU>
> Subject: Compensation tied to appraisals
>
> I'm working with a company who is changing their appraisal process.
> Previously, their appraisals were tied to compensation and they found that
> all employees were rated very high on their appraisals. They have
recently
> gone to an objective review process. The question from their employees
> is..... how is this process tied to compensation. If you do well on the
> review, doesn't that count towards increased compensation. My question to
> the list is..What models have folks seen that work well to determine
> compensation while still rewarding high performers. What part should the
> review play in determining compensation? What other factors are
> organizations using?
>
> Thanks.
>
> Bonnie Jackson
>

Robert Bacal wrote:

>In any event to talk a silly example. I only hire people who weigh
>exactly 212 pounds. I have a staff of 250 people weighting 212
>pounds. You're going to tell me that this selected group will
>comprise a normal distribution?

No, I won't tell you that. I agree with you: it's a silly example. But
NBA players aren't all 6' 10" tall and those who go to weight watchers
don't all weigh 300 pounds.

>You get what you select for. If the hiring practice is random you
>should get a normal distribution, all things being equal.

A normal distribution of what, Robert? Performance? I don't think so. I
think most of us would agree that the performance of any and all
individuals in the workplace is affected by factors beyond those
attributable to the individual. Consequently, even if I selected really
good people (the employee-quality equivalent of 212 pound people), their
performance could still be distributed in a bell shaped, normal curve.

>The more effective you are at selecting good people the less normal your
>distribution will be, all other factors being equal.

The less normal your distribution of what? Good people? Performance?

>If we are going to invoke statistical concepts here, could we please
>at least do it correctly?

That seems to imply that I incorrectly invoked a statistical
concept. Please be specific. Which concept and how was it invoked
incorrectly?
--

Fred Nickols
The Distance Consulting Company
"Assistance at A Distance"
http://home.att.net/~nickols/distance.htm
nickols@worldnet.att.net
(609) 490-0095

I would assume that they try to hire only outstanding employees,
Thus the problem of assuming normal distribution may be statistically accurate,
but in reality (we hope! the difference between our best and worst employees is
not significant enough to worry about distributions.

I've enjoyed lurking around this topic since I'll be teaching a(n) HRM class
for the first time in about 5 years this summer.

William.Sharbrough@citadel.edu


>On 28 Apr 00, at 18:13, Fred Nickols wrote:


>> Robert, I think your comments above assume something that needs to be made
>> explicit, namely, the reference point. While it is true that the
>> distribution of height among players in the NBA and the distribution of
>> weight among people who go to weight watchers won't match the distribution
>> of height and weight in the general population, owing to preselection, the
>> distribution of height among NBA players and the distribution of weight
>> among those who attend weight watchers might well assume a normal
>> distribution with respect to those two populations.

>Or it may not. As it is in the workplace. Probably a better example
>is the weight of professional linemen in football. This is an empirical
>question. If I get some time, I may take a few minutes to
>demonstrate, but you all can do it at home. Take any population
>that is heavily selected (not randomly, but filtered). Compute the
>mean and the median (that's a quick and dirty way to check about
>normalcy.



>I learned this a
>> long, long time ago as the principle of "bell curves within bell curves."
>> The range of heights among members of the NBA and weights among people who
>> attend weight watchers would be very different from the range in the
>> general population but the distributions in these two subsets of the
>> general population might in fact be quite bell-shaped. I confirmed this
>> possibility with three of my colleagues at ETS (where statistics
>> approaches a religion). As one of them put it to me, "It's possible that
>> you won't get a bell shaped curve of heights among the NBA players (or the
>> weight watcher attendees) but that's the way to bet."

>Ok, even if we leave aside the math of it, I'm sure those ETS folks
>will also tell you that you don't ASSUME a normal distribution and
>then operate and apply statistical tests as if it WAS normal. You
>calculate it first. That's hard to do in the realm of appraisals, to test
>form normal distributions, but like much in the field, people fudge
>and fake to sound scientific.

>In any event to talk a silly example. I only hire people who weigh
>exactly 212 pounds. I have a staff of 250 people weighting 212
>pounds. You're going to tell me that this selected group will
>comprise a normal distribution?

>You get what you select for. If the hiring practice is random you
>should get a normal distribution, all things being equal. The more
>effective you are at selecting good people the less normal your
>distribution will be, all other factors being equal.

>If we are going to invoke statistical concepts here, could we please
>at least do it correctly?

>There's also another aspect to this that Dr. Sullivan talked about at
>the performance management conference...if this continues, I'll
>mention it.



>Visit the work911.com supersite at http://www.work911.com
>for work related articles, or to find almost anything including
>book reviews and suggestions, discussion lists and more.

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
William C. Sharbrough, Ph.D.
Associate Professor of Business Administration
The Citadel
171 Moultrie Street
Charleston, SC 29409

Office (843) 953-5164 FAX (843) 953-6764 or Home (843) 763-8512
E-Mail: SHARBROUGHW@CITADEL.EDU

~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

On 29 Apr 00, at 7:07, Fred Nickols wrote:

> Robert Bacal wrote:
>
> >In any event to talk a silly example. I only hire people who weigh
> >exactly 212 pounds. I have a staff of 250 people weighting 212
> >pounds. You're going to tell me that this selected group will
> >comprise a normal distribution?
>
> No, I won't tell you that. I agree with you: it's a silly example. But
> NBA players aren't all 6' 10" tall and those who go to weight watchers
> don't all weigh 300 pounds.

Fred, the principle of selection is the same, whether you select a
broader range or not. The difference is in magnitude.

So at what point does a selection process get wide enough to give
you your assumed normal curve?

How about if we only select people who are 6.10 and 6.11? Is that
a normal curve? (no)

What about 6.9 - 6.11?

At what point does the distribution of a selected population
magically become normal?


Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

On 28 Apr 00, at 22:26, Kenneth Rossi wrote:

> I read Jack's post on employees approximating the normal distribution,
> setting up the appraisal system according to that and then determine why
> the distribution may deviate. That type of system has been around for
> years. Many organizations will dictate where employees can fall in the
> appraisal system and make sure the values assigned meet those criteria. If
> your employees appraisal is too high, they are just changed. The problem
> with doing the system this way, is it ensures that some employees will be
> below standards and some will be above and the rest will be in the middle
> (normal distribution).

You are absolutely correct (although I'm not sure that's how Ring
was suggesting it be used. In any event the application of a normal
curve, particularly in lower levels makes more sense because the
testing process can be tested itself for validity and reliability while
our employee measurements can't be (or not at a reasonable cost).

Further as you go up in ed. you get less and less randomness
selection, as the lower ability or achievement levels get dropped
out in university and grad school.

Good message...gotta run.



Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

Everywhere I have ever worked, whatever the job, it has always taken 2
people
to replace me when I have left. That is the volume and quality of my
performance.

I set my own standards... and they are very high.

I don't want to be measured relative to others, because that might cause me
to get lazy and reduce output to the lower standards of the so-called norm.

The other people that I know who are like me have expressed the same
feelings.

Measure and reward me on my progress and improvement, not someone else's
lack of it.

Conna Condon
DBA candidate Nova Southeastern Univ.

----- Original Message -----
From: "Robert Bacal" <rbacal@ESCAPE.CA>
To: <MG-ED-DV@MAELSTROM.STJOHNS.EDU>
Sent: Saturday, April 29, 2000 10:45 AM
Subject: Re: Appraisals and Compensation


> On 28 Apr 00, at 22:26, Kenneth Rossi wrote:
>
> > I read Jack's post on employees approximating the normal distribution,
> > setting up the appraisal system according to that and then determine why
> > the distribution may deviate. That type of system has been around for
> > years. Many organizations will dictate where employees can fall in the
> > appraisal system and make sure the values assigned meet those criteria.
If
> > your employees appraisal is too high, they are just changed. The problem
> > with doing the system this way, is it ensures that some employees will
be
> > below standards and some will be above and the rest will be in the
middle
> > (normal distribution).
>
> You are absolutely correct (although I'm not sure that's how Ring
> was suggesting it be used. In any event the application of a normal
> curve, particularly in lower levels makes more sense because the
> testing process can be tested itself for validity and reliability while
> our employee measurements can't be (or not at a reasonable cost).
>
> Further as you go up in ed. you get less and less randomness
> selection, as the lower ability or achievement levels get dropped
> out in university and grad school.
>
> Good message...gotta run.
>
>
>
> Visit the work911.com supersite at http://www.work911.com
> for work related articles, or to find almost anything including
> book reviews and suggestions, discussion lists and more.
>

Unfortunatley piecework systems (which if I am thinking the same as you
is the system by which workers are rewarded for the amount of product
they produce in a certain timeframe) are completely outdated. One
reason they don't work is that the motivation to complete the work is
contingent on the presence of the manager - employees will work faster
when their manager walks in the room than when the manager is out of the
room.

The argument about normality is an interesting one. In any highly
select sample there will be problems with range restriction. For
example, job incumbents consist of those who have scored the highest on
selection tests. (Or the tallest players are selected for the NBA).
Therefore, the range of scores (or heights) in the sample is restricted
and the estimated population correlation will be reduced. Corrections
for range restriction (see Gulliksen 1987 Theory of Mental Tests. NY:
John Wiley & Sons)exist, allowing select samples to be studied with
meaningful conclusions. However, I agree with Robert Bacal on this one,
due to the range restriction present - you can't apply a normal
distribution to a select sample.



Charlotte Bowden
MSc, Psychology
Specialising in personality
questionnaires as selection tools

On 30 Apr 00, at 15:49, bowdens wrote:

> The argument about normality is an interesting one. In any highly
> select sample there will be problems with range restriction. For
> example, job incumbents consist of those who have scored the highest on
> selection tests. (Or the tallest players are selected for the NBA).
> Therefore, the range of scores (or heights) in the sample is restricted
> and the estimated population correlation will be reduced. Corrections for
> range restriction (see Gulliksen 1987 Theory of Mental Tests. NY: John
> Wiley & Sons)exist, allowing select samples to be studied with meaningful
> conclusions. However, I agree with Robert Bacal on this one, due to the
> range restriction present - you can't apply a normal distribution to a
> select sample.

So we don't lose the point here - (there's actually two separate
ones). Jack Ring said:

"First, presume that any organizational
component with 15 or more employees should exhibit a
performance profile
approximating a normal distribution. Any group skewed
higher simply is not
being sufficiently challenged and any skewed lower are not
getting the
competency development (quality of management) they
deserve."

I challenged this by indicating one can't PRESUME normality for a
selected group. (Jack seems to think that there is no reason to
test this assumption). However, if the presumption is wrong, then
all the conclusions are probably going to be wrong.

Second, the normal distribution idea is used sometimes in forced
ranking appraisals and compensation. While forced rankings have
huge problems, again there is no justification for assuming that
performance should follow a normal distribution. In fact, if I recall
Dr. Sullivan, at the Performance Management Conference in Texas,
indicated that IF you have a normal performance curve, you should
fire at least half of your employees, since that's not what you want
anyway.

The reason I'm being a nuisance about this is that this list is for
people involved in teaching and developing managers. It seems to
me there is a higher responsibilty here to teach what is correct and
accurate, and not teach half-truths. If the professors and
management developers can't get this right, what hope do we have
of teaching managers how all this works?


Visit the work911.com supersite at http://www.work911.com
for work related articles, or to find almost anything including
book reviews and suggestions, discussion lists and more.

Charlotte Bowden wrote:

>Unfortunatley piecework systems (which if I am thinking the same as you
>is the system by which workers are rewarded for the amount of product
>they produce in a certain timeframe) are completely outdated. One
>reason they don't work is that the motivation to complete the work is
>contingent on the presence of the manager - employees will work faster
>when their manager walks in the room than when the manager is out of the
>room.

I'm not sure what you mean by "they don't work" in the paragraph above. If
you mean piecework systems offer no guarantee that employees will work at
the maximum rate, I agree. But is that the aim? If it is, the managers
involved have failed to profit from all that has been learned in the last
100 years. I don't know if people on piecework systems work faster when
the manager is present or not but it wouldn't surprise me if they
did. People tend to "manage" lots of things, two of the more important
ones being their income and their image. People on piecework (and those on
commissions as well) tend to do what they have to do to earn the amount of
money they want to earn. They are managing their income, not responding to
an incentive system. That's one of the fundamental problems with all
incentive systems; they tend to overlook the fact that the people subject
to it will manage it to their own ends, not those of management. There is
also abundant evidence that employees will manage their own production
under piecework systems because management has a long history of using
piece rates to step up the pace of work and then shift from piecework to
production quotas at hourly rates that are less than the workers could earn
on the piecework rate.

>The argument about normality is an interesting one. In any highly
>select sample there will be problems with range restriction. For
>example, job incumbents consist of those who have scored the highest on
>selection tests. (Or the tallest players are selected for the NBA).
>Therefore, the range of scores (or heights) in the sample is restricted
>and the estimated population correlation will be reduced. Corrections
>for range restriction (see Gulliksen 1987 Theory of Mental Tests. NY:
>John Wiley & Sons)exist, allowing select samples to be studied with
>meaningful conclusions. However, I agree with Robert Bacal on this one,
>due to the range restriction present - you can't apply a normal
>distribution to a select sample.

My participation in this discussion ties to what I see as a requirement to
be clear about your reference point. The distribution of height among NBA
players is doubtless quite different from the distribution of height in the
general population. That said, it is entirely conceivable that the
distribution of height among the NBA players could in fact be characterized
by a normal distribution. In other words, the NBA players can be viewed as
a subset of the general population and as a population unto itself and it
helps to be clear about which is which.
--


Fred Nickols
The Distance Consulting Company
"Assistance at A Distance"
http://home.att.net/~nickols/distance.htm
nickols@worldnet.att.net
(609) 490-0095