How soft skills could increase productivity

There’s a lot of talk about soft skills and how they might help boost productivity and earnings. Into this literature comes a neat new paper by Achyuta Adhvaryu, Namrata Kala, and Anant Nyshadham which looks at the returns to providing training on these skills for factory workers in India. They provide a convincing case that it might make economic sense for firms to provide trainings for these skills.

The setting: Off we go to southern India, where Adhvaryu and co. are partnering with the largest ready-made garment export firm in the country. Workers (all of them female) work in lines of about 70-100 workers, where everyone has a given task to produce a single style of a garment at a time (the paper has a neat description of the process which may change how you look at your shirts). Working with the company, Adhvaryu and co. randomize lines, and then workers within lines, to receive a soft skills training.

Soft skills are many things to many people; what are they in this case? This program is called Personal Advancement and Career Enhancement (PACE) and it covers a range of topics including: time management, effective communication, problem solving, and financial literacy. The whole course took 80 hours over a period of 11 months, with half of those hours coming from the workers’ time and half from working hours.

Adhvaryu and co. collect a range of data. Most important is production data, where they put tablet computers on the factory floor with an employee collecting indicators of the workers in the production line (these production data collecting folks seem to have existed before – but the tablets were new). One interesting data collection constraint fact: the tablets eventually gave out (many of them simultaneously) which provides a time limit on the post-treatment effects that could be observed. Adhvaryu and co. complement this production data with HR data on attendance and salaries, plus a worker survey a month after the training was completed. Among other things, this survey lets them measure some key personality characteristics.

So what do they find? Let’s start with retention. Attrition from these jobs is high overall – after 26 months of observations, Adhvaryu and co. are left with only about 25-30% of their initial sample. The PACE training improves retention, but only marginally so, and only during the training – post-training the two rates are quite similar. This could potentially create issues for the estimation (e.g. if the nature of attrition was affected by the program), but Adhvaryu and co. do a range of checks (e.g. balance tests of baseline characteristics at different points in time) to show us we shouldn’t be worried. Attrition aside, the program does have a positive and significant effect on attendance – both during the training and after.

Are the trained workers more productive? Here Adhvaryu and co. use three different measures. The first is the number of pieces (garments) produced. The second is efficiency, which is measured by the number of pieces produced divided by the target. The target comes from the wonderfully named “standard allowable minute” (SAM). This is a global industry standard, courtesy of industrial engineers, which tells us how many minutes it should take to produce a garment. (Important side note: when Adhvaryu and co. estimate efficiency they use garment style fixed effects to control for complexity). The third production measure is to actually look at the SAM to see if treatment workers are doing more complex tasks.

Production of the workers who get the training goes up, as they produce about 6 more garments per hours, or 10% more relative to the control group. Efficiency also goes up by 7 percentage points (or 12% relative to controls). Interestingly, both of these measures go up after the training, not during (note that all the productivity measures control for time on the job). The complexity of the tasks the workers are assigned to actually goes up during the training, and remains higher (at 10 percent significance) post-training as well.

Before we turn to wages, let’s look at potential mechanisms. Here Adhvaryu and co. provide some insight from their survey data. In terms of personality characteristics, the big change comes from extraversion, which increases substantially relative to the control group. The folks who got the training also show some signs of increasing their forward-looking behavior – they’re more likely to be saving for education and have higher aspirations for their children’s education. They are also substantially more likely to be availing themselves of government pensions and subsidized healthcare (which could be through the financial literacy part of the training). One other interesting effect is the those who were trained in the soft skills are about 15 percentage points more likely to request training in technical (hard) skills. So when we see the productivity and complexity improvements, this needs to be kept in mind.

Another way Adhvaryu and co. get at the potential channels is to look at spillovers. As I mentioned above, not all workers on a given line get the treatment, and the treatment was also randomized across lines.This lets Adhvaryu and co. look at the effect of the treatment on co-workers (on the same line) who did not get the treatment. Interestingly, these folks show similar productivity and complexity increases but no changes in personality.

So does this boost in skills and increased productivity translate into higher wages? Not so much. For those workers who get the training wages go up by a measly 0.5 percent post-training (significant at 10 percent). Starting from the wages and the productivity gains, Adhvaryu and co. can compute the net rate of return for the firm from the training. They conclude that the rate of return at the end of the training period is 12 percent (with this mostly coming through the reduced attrition). So this program makes sense for the firm even when it is going on. After the program though, productivity takes off and the net rate of return 20 months after the program climbs to a whopping 256 percent.

This is a neat paper. It shows us that some of the personality traits are still significantly malleable in adulthood for this group of workers. And when these skills change, we see productivity gains. What is also striking is what it tells us about the labor market. Almost all of these gains are captured by the firm – possibly due to issues of signaling or firms’ inability (or unwillingness?) to screen for these skills – which means that this kind of training isn’t immediately portable. Understanding this piece of the puzzle is a key next step.