Statistical Analysis on Changing Work Values in the USA

Abstract:

Statistical analysis of the USA General Social Survey (GSS) data shows that work values changed radically after the 2008 recession. What used to be the most important value, “Work important and feel accomplishment” became the least important. What used to be the least important values, “Short working hours”, and ” No danger of being fired” became the most important in 2012.

 

Core Values

 

 

Background:  

Recently, I took an online course from Coursera, Data Analysis and Statistical Inference. It is essentially first year university statistics. It was from Duke University, and taught by:  Dr. Mine Çetinkaya-Rundel.

R Programming Language:

When I took first year statistics in university years ago, we did not use any software. In this course, the R statistical programming language is used.

GSS Project:

In this course, there is also a project. The project was to look at data from the General Social Survey (GSS). The GSS is a large study that has been conducted in the USA since 1972.  The survey data was collected from random samples of US adults.

A subset of the GSS data, Modified for Data Analysis and Statistical Inference course (Duke University) can be found here:
https://d396qusza40orc.cloudfront.net/statistics%2Fproject%2Fgss1.html
A more permanent link to the GSS study can be found here:
http://www.icpsr.umich.edu/icpsrweb/ICPSR/studies/34802/version/1
A 3436 page PDF Codebook document describing the data and the methodologies used can be found here:
http://www.icpsr.umich.edu/cgi-bin/file?comp=none&study=34802&ds=1&file_id=1136502

Each row of data are the responses from an adult individual surveyed during one of the years in question. Please refer to the Codebook for more detail of the methodologies, in particular, Appendix A, which contains a 20 page detail of the Sampling Design and Weighting.

If you want to look at the data yourself, you can run these commands in R, and explore.

load ( url (http://bit.ly/dasi_gss_data ) )
str (gss)

(Use single quotes around the URL.)

 

Variables On Attitudes Toward Work:

There are many variables that are surveyed and many correlations that can be discovered. Some students chose some very interesting studies, which I saw through peer review.

I decide to explore attitudes toward work, and how they have changed over time. The variables were:

jobmeans: Work Important and Feel Accomplishment
jobpromo: Chance of Advancement
jobinc: High Income
jobsec: No Danger of Being Fired
jobhour: Short Working Hours

year: Gss Year For This Respondent

It turns out that the data for these variables on job attitudes are not available for every year of the survey. Of the 29 years surveyed (1972 to 2012), there is data on work attitude for only 17 of the years.

 

table ( gss$year, gss$jobsec )
      
       Most Impt Second Third Fourth Fifth
  1972         0      0     0      0     0
  1973       100    210   290    438   424
  1974       111    159   286    417   481
  1975         0      0     0      0     0
…
  2004         0      0     0      0     0
  2006       170    225   288    403   412
  2008         0      0     0      0     0
  2010         0      0     0      0     0
  2012       261    149   127    246   166

At the end of the years in question, there is data for 2006, then data for 2012. So, the 2008 recession occurred in between these those two survey years.

Analysis:

One of the ways that I looked at the data was to just see which values were the most popular and rank them.

c (  “jobmeans” ,
round (
100 *
table ( na.omit (subset( gss$jobmeans ,  gss$year == 1973 )   )  ) /
length ( na.omit (  subset( gss$jobmeans ,  gss$year == 1973 )   ) ) , 2 )
)

 

            Most Impt     Second      Third     Fourth      Fifth 
"jobmeans"    "51.57"    "16.14"    "12.93"    "11.83"     "7.52" 

I ran similar commands for each value, and put the results into a 5 by 5 matrix.  The results for 1973 look like this.

 

workvalue – 1973	most	second	third	fourth	fifth
jobmeans	51.57	16.14	12.93	11.83	7.52
jobpromo	18.19	35.36	23.39	13.34	9.71
jobinc	18.60	24.08	30.78	19.56	6.98
jobsec	6.84	14.36	19.84	29.96	29.00
jobhour	4.79	10.05	13.06	25.31	46.79

 

If you look at the 1973 data, you will see that for 51.57% of all respondents, the value, jobmeans (Work important and feel accomplishment), was the most important value of all. The next most popular value was jobpromo (chance of advancement), at 35.36%. jobinc (High income) was third. Jobsec (No danger of being fired) was forth. The least important value was jobhour (Short working hours), for 46.79% of respondents.

The pattern is seen by looking through the 5 by 5 matrix, from the top left, to the bottom right. Look for the maximum value in each row.

Graphed as a barchart, the 1973 data looks like this:

 

GSS Work Values 1973

 

The legend for the graphs is:

GSS Work Values Legend

 

Long Term Average:

Interestingly, over the period 1973 to 2006, the long term average of all the values stays very similar to what they were in 1973. The data in a table matrix looks like this:

workvalue – LT Avg	most	second	third	fourth	fifth
jobmeans	48.80	8.69	14.31	11.68	6.53
jobpromo	18.68	34.69	23.59	14.73	8.32
jobinc	21.26	24.93	30.43	17.79	5.58
jobsec	7.53	13.41	20.41	31.95	26.70
jobhour	3.72	8.29	11.27	23.85	52.88

 

A graph of the data from 1973-2006 looks like this:

 

GSS Work Values – Long Term Average: 1973 – 2006

 

 

Work Values Changed Radically in 2012:

However, in 2012, the results change dramatically. A barchart of the job values for 2012 looks like this.

GSS Work Values – 2012

 

As you can see, the shape of the 2012 graph is radically different from the other two.

In 1973, and for the long term average of 1973 to 2006, the values of jobmeans (Work important and feel accomplishment), and jobhour (Short working hours) dominate the chart in the corners.

However, in 2012, the shape of the graph is different. Jobmeans still takes up a large percentage, but it is the third most important value. Jobinc (High income) looks to be the only other value with a large percentage.

 

A table matrix of the work values also shows how the values toward work changed.  Previously the matrix ranking was:  (jobmeans, jobpromo, jobinc, jobsec, jobhour).  However, in 2012, the ranking became: (jobsec, jobpromo, jobmeans, jobhour, jobinc)

workvalue – 2012	most	second	third	fourth	fifth
jobsec	27.50	15.70	13.38	25.92	17.49
jobpromo	15.24	29.10	20.63	20.53	14.50
jobmeans	6.79	24.60	46.34	16.76	5.51
jobhour	23.36	22.09	13.00	24.63	16.91
jobinc	27.03	8.55	6.65	12.14	45.62

 

No Danger Of Being Fired Became The Most Important Value In 2012:

Using the matrix ranking in the same manner, in 2012, jobsec (No danger of being fired) moves from the fourth most important value, to the most important value.

High Income Becomes Either The Least, Or Most Important In 2012:

Looking at the 2012 table data, what’s very interesting is that in 2012, attitudes toward High Income (jobinc) change drastically. The matrix ranking moves from third, to fifth place. At the same time, the attitudes toward High Income move to either the most important (27%), or the very least important (45.6%). All the other values in the row are either single digits (8.5%, 6.6%), or close to single digits (12%). This indicates that the attitudes have become very polarized.

Jobmeans Moves From First to Third Position in 2012:

While Work important and feel accomplishment (Jobmeans) dominated all the results in 1973, and over the long term (1973-2006) as the most important value, in 2012, jobmeans moved into third place, where 46.34% rated it their third most important value.

Free Time Becomes Much More Important In 2012:

Over the long term average, 52.88% of respondents said that Short Working Hours were the very least important work value. However, in 2012, this value in last place dropped to only 16.91%. At the same time, 23.36% of respondents rated Short Working Hours as the most important, and 22.09% rated it as the second most important value.

———–

Analyzing the Most Important Value Only:

Looking at the just the most important value only, we can also see some very interesting results.  Here, the most important column, is extracted from the two 5 by 5 matrices for 1973 and the long term average. The bold indicates the value from the 5 by 5 matrix.

 

Most Important workvalue	1973	1973-2006
jobmeans: Work important and feel accomplishment	51.57	48.80
jobinc: High income	18.60	21.26
jobpromo: Chance of advancement	18.19	18.68
jobsec: No danger of being fired	6.84	7.53
jobhour: Short working hours	4.79	3.72

 

In 1973, and over the long term average, the ranking of the most important column, is quite similar to the matrix ranking, with two columns exchanging places.

Jobmeans (first) jobsec (fourth) and jobhour (last) are all ranked in the same positions. However, whereas in the matrix ranking, Jobinc (High income) was ranked third, looking at the most important column only, it becomes second. And jobpromo (Chance of advancement) moves from second place in the matrix rank to third place in the most important column.

However, in 1973, looking at the actual percentages for these two values (18.60%, and 18.19%), they are practically equal, as indicated by the italics.

When you consider that a promotion is usually accompanied by a raise in pay, how exactly are the respondents distinguishing the two values?

Over the long term average (1973 to 2006), the ranking is the same as in 1973. However, jobinc/High income (21.26%) edges out jobpromo/Chance of advancement (18.68%) by a few percentage points.

 

Most Important Values Changed Radically in 2012:

 

Most Important workvalue	2012
jobsec: No danger of being fired	27.50
jobinc: High income	27.03
jobhour: Short working hours	23.36
jobpromo: Chance of advancement	15.24
jobmeans: Work important and feel accomplishment	6.79

 

Looking at the most important values, they all changed radically in 2012.

No danger of being fired (jobsec) moves from forth position (7.53%), into the first position (27.5%). Short working hours (jobhour) moves from last (3.72%) to third position (23.36%). And Work important and feel accomplishment (jobmeans) moves from the most important (48.8%) to the least important (6.79%).

High income remains in the second position. However, the actual percentage for High income (27.03) is within one percentage point of No danger of being fired (27.50), practically equal (as indicated in italics). And Short working hours (23.36) are only four percentage points behind.

Looking at the most important work value, the respondents’ values are practically reversed from the long term average!

A graph of the most important values over the three time periods is here:

 

GSS Most Important Work Value Only

 

———–

Analyzing the Two Most Important Values:

If we combine the two most important columns, there are even more interesting results.

For 1973, and for the long term average, combining the top two columns gives results very similar to the 5 by 5 matrix for the same years.  The ranking of the values is exactly the same as in the matrix model.

workvalue – Sum of Top Two Columns	1973	1973-2006
jobmeans: Work important and feel accomplishment	67.72	67.49
jobpromo: Chance of advancement	53.56	53.37
jobinc: High income	42.68	46.19
jobsec: No danger of being fired	21.20	20.94
jobhour: Short working hours	14.84	12.01

The range of difference between the most important value jobmeans (67.49%) and the least important, jobhour (12.01%) is 55.48%.

 

However, in 2012,  jobmeans (Work important and feel accomplishment) falls from the most important, the least important value. And jobhour (Short working hours) rises from the least important value, to the most important.

workvalue – Sum of Top Two Columns	2012
jobhour: Short working hours	45.45
jobpromo: Chance of advancement	44.34
jobsec: No danger of being fired	43.20
jobinc: High income	35.58
jobmeans: Work important and feel accomplishment	31.39

 

What’s also interesting to note is that the top three values, are all valued practically equal, from 43.20% to 45.45%. The differences in the response rate of these three values do not seem to be statistically significant.

And, the difference between the most important value jobhour (45.45%) and the least important, jobmeans (31.39%) is now only  14.06%.

A graph of the top two work values is here:

GSS Top Two Work Values

 

———–

Interpretation:

How can we interpret or explain these dramatic changes in work values?

Before the 2008 recession, values used to be very clear cut. People wanted their work to be very meaningful (jobmeans). The last thing they cared about was free time (jobhour), they wanted to put a lot effort into their work, and get promoted (jobpromo).

However, in 2012, people have multiple values, and there is not much distinction between them. Looking at the most important value, in 2012, No danger of being fired (jobsec), and High income (jobinc) are essentially equal in weight.

Looking at the top two most important values, in 2012, Short working hours (jobhour), Chance of advancement (jobpromo) and No danger of being fired (jobsec) are practically equal in weight.

In both cases, Work important and feel accomplishment (jobmeans) is now last on workers’ priority.

I would conclude from the data, that the 2008 recession deeply affected the psyche of Americans.

 

Maslow’s Hierarchy Of Needs:

The results reminded me of Maslow’s Hierarchy Of Needs.

Maslow’s Hierarchy Of Needs

Rising up the pyramid, the Maslow defines everyone’s needs as: Physiological, Safety, Love, Esteem, and Self Actualization.

Before the 2008 recession, people’s awareness, hopes, and aspirations seemed to be focused on self esteem and self actualization (Work important and feel accomplishment, Chance of advancement).

Since the 2008 recession, in 2012, people now seem to be very focused on survival (No danger of being fired, High income).

 

Potential Explanations:

Why have the values toward work changed so much?  The US people have certainly seen a lot since the 2008 recession. In the news, they have seen:

Stock market crash and the savings of many wiped out.
People lost their homes in the MBS and rising interest rates fiasco.
Gas and energy prices have risen much higher.
Companies have done massive layoffs, or have gone bankrupt.
Outsourcing of jobs overseas
Even before the 2008 recession, there were scandals like the one at Enron.

On a personal basis, many people have:

Carried the pager and worked over the weekend.
Worked two jobs (no wonder Free time is becoming important).
Got short term contracts rather than long term employment.
Taken cuts in pay, while prices still rise.
Seen rapidly changing skill sets, making existing skills, built over a long period of time, obsolete.

Students have graduated with huge amounts of student loan debt that can never be forgiven. Even some law students that have graduated cannot find a job.

No wonder values have changed.

 

Further Research:

In the GSS study, there are lots of demographic variables, such as their work status (gss$wrkstat and gss$spwrksta), whether or not the respondent is unemployed (gss$unemp) and their income (gss$income06) There are also many other variables on attitudes, such as how easy it is to find a job (gss$jobfind), how likely they feel it is to lose their job (gss$joblose), and how do people get ahead (gss$getahead), and so on.

A potential study for further research could be to look for correlations with the many other variables that are found in the GSS.

 

Resources:

A PDF spreadsheet of the results can be found here:
Changing Work Attitudes GSS Data – PDF

A PDF of some R code can be found here.  It can be modified to get the results for the other years.
R code for workvalues 1973

By Rodger Lepinsky

 

Notes on the Statistical Tests:

Originally, the idea was to make these results a line graph, with time on horizonal axis, and the percent of importance on the vertical axis. In the graph, there would be five lines, one for each value, rising or falling over time.

However, this was a course in statistics, and part of the project required doing a statistical test.  The test for statistical significance of all these five variables, plus time, would be quite complex. I visited a PHD in Math, and later two PHDs in Statistics. Some of the potential ways to test for statistical significance might be:
KruskalWalliss Analysis of Variance by Rank
Friedman Two Way Analysis of Variance By Rank

Non parametric measures of Correlation included:
Spearman Rank Correlation
Kolmogorov Smirnow One Sample Test
Kolmogorov Smirnow Two Sample Test

All of these tests are well beyond this first year statistics course, and not able to be understood by peer (fellow student) reviewers. Even the PHDs had to pull out their statistics books to determine what the appropriate test might be.

So my project could be peer reviewed, in the end, I ran the Chi Squared test on a single variable, how it changed from the long term average, to 2012. The result was highly statistically significant.

 

Notes On Plagarizing:

For those of you who are taking online courses, you should not need to be told to not plagiarize this work. For one thing, it violates the honor code.  For another reason, the big data capabilities of Coursera and other companies will most certainly discover your copying soon enough, and you will fail your project. Finally, this analysis is well beyond first year statistics. It’s better to choose a simpler project that can actually be understood by fellow students and peer reviewed.