Which Skills Are the Most Valuable?

Years of sluggish real wage growth and aggressive technological transformation have convinced many workers that it’s only a matter of time before they are replaced by robots. As a result, workers are asking how they can get paid more and have a sliver of job security in the digital age. The advent of MOOCs and skill-based boot camps has made it easier than ever to acquire new skills, but it has also left workers paralyzed by the sheer volume of career advancement opportunities. Will Excel help them get the job they want? Will Tableau help them earn more money? Will a machine learning bootcamp “future proof” their career?

While this article offers no easy answers, it does offer a framework for thinking about skills and how to value them, along with information on data analyst and data scientist jobs to illustrate this framework. It also offers a lengthy discussion on latte art because, yes, that is a thing and yes, it is relevant.

A Framework for Skills and Skill Premiums

A skill premium, or what compensation professionals call a skill differential, is the difference between how much you are paid and how much someone else is paid when they share your title but do not share a particular skill. For example, if you’re a barista who knows how to make pictures in latte foam, you likely earn more tips than a barista who serves up a latte with a blob of white foam. The extra money you earn because you can make a swan out of foam is the skill premium associated with latte art.

As you can probably guess, not all skills will command a premium. Additionally, which skills are considered most valuable will vary from job to job. At PayScale, for each job we categorize skills into four buckets.

1. Assumed: Skills that are so necessary for doing a job that companies don’t even ask if a candidate has these skills. Assumed skills may be building blocks of essential skills.
2. Essential: Skills that candidates must have to succeed in a job, but cannot be assumed. Essential skills are typically specialized.
3. Useful: Skills that are not required to do a job but help get the job done more efficiently or in some other way that is better.
4. Irrelevant: Skills that don’t improve the efficiency or execution of a job and candidates do not need to have them.

Assumed, essential and irrelevant skills don’t increase pay. The value of essential and assumed skills for a job are already baked into the base salary for that job. Irrelevant skills may be impressive or useful in other areas, but are unrelated to what is needed for that job. As a result, they don’t impact pay.

That leaves us with useful skills. Useful skills can help you earn more than your peers. In other words, they command a premium.

Think back to our barista example. To do their job, baristas must be able to serve lattes, which requires pouring milk into a pitcher. In our framework, pouring milk into a pitcher would be an assumed skill; it is so basic and essential to the job that it goes without saying. Making the latte — heating the milk, pulling the espresso shot, getting the right proportions of steamed and foamed milk into the cup — is an essential skill. You need this skill in order to get the job done, but it is far more complex than simply pouring milk into a pitcher, thus it is not an assumed skill. However, pouring milk into a pitcher and making lattes are similar in that neither of these skills will help you get paid more than your fellow baristas. Their value is already priced into the job.

On the other hand, being able to discuss Shakespeare at length, while impressive, is a skill that is irrelevant to being a barista. Like pouring milk into a pitcher and making lattes, analyzing Shakespeare does not command a premium, but in this case it is because it is irrelevant to the job.

Making latte art is neither a simple skill, nor is it essential to being a barista. It is, however, useful as it makes customers feel special and happy. Consequently, it can help baristas earn more tips. In other words, making latte art is a skill that commands a premium.

Now, let’s use this framework to analyze two of the hottest jobs on the market: data analyst and data scientist.

Data Analysts vs Data Scientists

The era of big data has created considerable demand for data analysts and data scientists. However, the difference between these two titles is not always understood by those who are not data nerds themselves. To sum up the difference in one sentence: Data analysts use standard statistical techniques to analyze data and communicate results to stakeholders, while data scientists leverage advanced statistical techniques and programming to predict findings from large swaths of data. Of course, this difference means data analysts and data scientists get paid differently and need different training.

Let’s start with the numbers. According to our data, median pay for data scientists is $98,000, which is $15,000 more than the median earnings of senior data analysts and $39,000 more than lower level analysts.

This higher median pay is not for nothing. Our data show that data scientists tend to spend more time in school than data analysts. Roughly 57 percent of data analysts have at most a bachelor’s degree. By contrast, only 26 percent of data scientists have at most a bachelor’s degree. Further, 23 percent of data scientists have PhDs, compared to only one percent of data analysts. This difference in formal education is even more stark when we compare senior data analysts and senior data scientists. While close to half of all senior data analysts have at most a bachelor’s degree, 44 percent of all senior data scientists have a PhD.

In terms of what they do, the difference is a bit harder to quantify. Broadly speaking, data analysts are storytellers. Some data analysts help businesses understand how they’re doing by constructing data dashboards that bring key metrics to life. Others build literal narratives with data by writing reports and blog posts. In each of these cases, data analysts use data to construct a narrative that helps their intended audience better understand the millions of data points that are constantly coming at them.

Data scientists, on the other hand, tend to be more of a cross between a computer engineer and a statistician. As the Data Analytics Handbook puts it, “A data scientist is better at statistics than a software engineer and better at software engineering than a statistician.” This emphasis on engineering is driven by the statistical tools data scientists employ. Most of these tools require complex algorithms and rely on advanced mathematical techniques that analysts do not typically utilize.

Now that we have a framework for thinking about skills and understand the differences between data analysts and data scientists we can make sense of their respective skill premiums.

Skill Premiums Depend on Job Title

When we look at the skill premium data² we see the same skills command very different premiums for data analysts and data scientists. Take machine learning for example. For data scientists, this skill only commands a 4 percent premium, while for data analysts the premium is 18 percent.

To put this into dollar terms, a typical data analyst could earn $10,500 more a year by acquiring machine learning as a skill, while a typical data scientist would earn only $4,300 for getting this skill if they didn’t already have it.

What gives? Why would machine learning command a high premium for data analysts but not so much for data scientists? And if it doesn’t command a premium for data scientists, why do nearly two-thirds of all data scientists have this skill?

In coffee terms: For data scientists, machine learning is making a latte. For data analysts, it is making latte art.

In other words, machine learning is essential for data scientists and merely useful for data analysts. The value of machine learning for data scientists is, for the most part, already priced into their median salary. For data analysts, the value of machine learning is not already part of their salary, so it commands a premium. Knowing machine learning techniques and being able to use them is one of the reasons why a data scientist earns $15,000 more than a senior data analyst.

Machine learning is not the only skill like this. The skill of statistical analysis increases pay by 7 percent for data analysts and only 1 percent for data scientists. Python increases pay by an average of 13 percent for data analysts and only 4 percent for data scientists. These are skills that are, for the most part, considered essential to being a data scientist but are “nice to haves” for a data analyst.

What’s the Deal with Microsoft Office?!?

You probably have noticed that Microsoft skills — Microsoft Excel, Microsoft Word, Microsoft Access and Microsoft Office — command negative premiums. Putting them down as a skill in PayScale’s salary survey is associated with lower pay for both data analysts and data scientists.

Does this mean that data analysts and data scientists should forget about Microsoft Office? Not quite.

In our skills framework, these are assumed skills for data analysts and data scientists. While data analysts and data scientists may spend most of their day plugging away in SQL or Python, they are still expected to know how to make a pivot table in Excel. In short, Excel is to data analysts and data scientists what pouring milk into a pitcher is to baristas.

So, why do they command negative premiums? Do companies ding their workers for having skills that they need?

Not exactly. The first rule of data analysis is that correlation does not imply causation, and in this case we are strictly talking about correlation.

One of the features of the PayScale model is that it produces a salary range for each individual respondent to our salary survey based on, among other things, the skills that they report. However, we know that not all skills are equally important for doing a job. The PayScale model is smart and also knows this. The model uses statistical techniques to determine which skills are the most important for determining pay for a given job title and then produces a salary range based only on those top skills. Of course, if a respondent to our salary survey only reports lower-level skills, those are the only ones that the model can use to construct a salary report. Unsurprisingly, respondents who report only lower-level skills tend to be less skilled and lower paid workers. Therefore, the resulting premium shows up as negative.

This is how the skill premium for Excel is negative for both data analysts and data scientists. It is not that Excel is not useful for them. It is not that Excel is not needed by them. It is that high-skilled data analysts and data scientists report other, higher paying skills. Lower skilled workers can be expected to make lower salaries, and the negative premium reflects that reality.

We can use this logic to make sense of why the skill premium for Tableau is 9 percent for data analysts and -8 percent for data scientists. Being able to deliver insightful and engaging data visualizations is an in-demand skill for data analysts, and Tableau is a great tool for building these visualizations. The negative premium for data scientists is likely because they do not need to use Tableau to succeed at their job, and highly skilled data scientists have more specialized skills that do help them accomplish their tasks. Tableau is not an essential skill, and may not even be a useful one for many data scientists. As such, it does not offer a skill premium, and data scientists who list it as a skill will tend to be those with fewer of the specialized skills that boost their pay.

Which Skills Pay More?

We have focused on two technical job titles here, so naturally the skills we looked at were also technical. However, in previous research, we found public speaking and critical thinking are in short supply. These and other soft skills may be the ones a given worker needs to earn more money or to secure their job for years to come. It all depends on which skills are assumed, essential, irrelevant or useful for their current positions and the positions they hope to secure in the future.

Methodology

This report is based on responses to PayScale’s salary survey between January 1, 2018 and December 31, 2018.

In addition to being able to select skills that are populated based on their job title, respondents can report skills in an open-response box. Respondents can report more than one skill, but are not required to report skills in order to complete the survey. We have reported skills that only have 40 or more observations.

Number of Observations

For Pay

For Degrees

For Skills

Definitions

Degrees – The degrees reported are the highest degree reported. If someone reported they had a bachelor’s degree and a master’s degree, they are counted as having a master’s degree.

Job Titles

• Data Analyst – Includes anyone who reports their job title as “junior data analyst” or “data analyst.”
• Data Scientist – Includes anyone who reports their job title as “associate data scientist,” “data scientist” or “data scientist, IT.”
• Senior Data Analyst – Includes anyone who reports their job title as “senior data analyst.”
• Senior Data Scientist – Includes anyone who reports their job title as “senior data scientist” or “senior data scientist, IT.”

Median Pay: Median pay is the national median (50th percentile) annual total cash compensation. Half the people doing these jobs earn more than the median, while half earn less.

Skill Premium: The percent increase/decrease in pay associated with a given skill. The skill premium is measured using PayScale’s proprietary compensation model.

Total Cash Compensation (TCC): TCC combines base annual salary or hourly wage, bonuses, profit sharing, tips, commissions, and other forms of cash earnings, as applicable. It does not include equity (stock) compensation, cash value of retirement benefits, or value of other non-cash benefits (e.g., healthcare).

1. We only report data for degrees that have more than 10 observations for a given title, therefore some of the columns may not sum to 100%.
2. We measure skill premiums using our proprietary compensation model. For each job title, the model is able to estimate the marginal impact the skill has on pay, which is what we report as that skill’s premium for that job.