Open Access

The impact of soft skills training on female youth employment: evidence from a randomized experiment in Jordan

  • Matthew Groh1,
  • Nandini Krishnan1,
  • David McKenzie1, 2, 3, 4Email author and
  • Tara Vishwanath1
IZA Journal of Labor & Development20165:9

DOI: 10.1186/s40175-016-0055-9

Received: 21 September 2015

Accepted: 16 February 2016

Published: 13 May 2016


Employers around the world complain that youth lack the soft skills needed for success in the workplace. In response, a number of employment programs have begun to incorporate soft skills training, but to date there has been little evidence as to the effectiveness of such programs. This paper reports on a randomized experiment in Jordan in which female community college graduates were randomly assigned to a soft skills training program. Despite this program being twice as long in length as the average program in the region, and taught by a well-regarded provider, we find soft skills training does not have any significant employment impact in three rounds of follow-up surveys. We elicit expectations of academics and development professionals and reveal that these findings are novel and unexpected.

JEL codes: O12, O15, J08, J16


Soft skills Youth unemployment Randomized experiment Expectation elicitation

1 Introduction

Firms around the world complain that formal schooling, at best, teaches only the technical skills workers need but that many youth are lacking in the soft skills needed for success in the workplace—such as how to interact with customers, how to work in teams, how to act professionally, and even how to properly represent themselves in job interviews in the first place. For example, in a review of 28 studies using employer surveys in different countries, Cunningham and Villaseñor (2014) find that socio-emotional skills are the first priority of 76.5 % of the studies that rank employer skill preferences. Likewise, several surveys of US employers identify a lack of soft skills as the area where young job seekers have the largest deficiency (White, 2013).

As a result, efforts to improve these skills have become an increasing part of youth employment programs in many developing countries, especially in Latin America (World Bank 2010). For example, the entra 21 program implemented in 18 Latin American countries includes a soft skills training component. Although there is no rigorous evaluation of this intervention, employers report that participants who took part in this program have greater ability to work in teams and take on responsibility than do their other employees (entra 21, 2009). The Dominican Republic’s Juventud y Empleo program also teaches soft skills along with providing work experience; with early evidence from Ibarraran et al.’s (2012) finding, this increased the formality of work but did not lead to any increase in overall employment in the short run. More recently, a series of pilot programs for adolescent girls have combined life skills training with technical or vocational training in Afghanistan, Haiti, Liberia (Adoho et al. 2014), Nepal (Ahmed et al., 2015), and Rwanda.

Yet despite the growing popularity of such programs, there is little empirical evidence on the effectiveness of teaching soft skills. There is a variety of evidence that shows associations between non-cognitive or soft skills and employment and other life outcomes (e.g., Bowles et al., 2001; Heckman et al., 2006; Heckman and Kautz, 2012; Groh et al., 2015), but it is less clear as to whether such skills can be effectively taught, especially in the types of 2- or 3-week courses used in the typical employment program. Moreover, studies in developing countries have typically combined life or soft skills training with vocational or technical training, so that the effect of the soft skills training alone cannot be measured.

Our paper provides the first experimental evidence on the effectiveness of soft skills training for youth in a developing country context.1 The Jordan New Opportunities for Women (Jordan NOW) pilot randomly allocated 1347 female community college graduating in 2010 into a training group that received an invitation to attend a 45-h soft skills training course, and a control group which was not.2 Follow-up surveys then measure employment impacts 6, 14, and 27 months later, allowing us to examine the program impacts at different post-training time horizons. Despite the course being taught by a well-regarded provider and offering twice as many hours of training as the average employability program does in the region, we find that soft skills training did not have any significant impact on any of a range of different employment measures for these young women over any of these time horizons. Examining non-employment outcomes, we find that in the short term, the training does appear to have made the participants more optimistic about the future, and have reduced restrictions on their mobility.

A common refrain following the presentation of experimental results or results from other rigorous impact evaluations is often that the findings “will surprise no one who had opened an economics textbook” (Scheiber, 2007). However, as Watts (2011) discusses, many findings in social science seem obvious once you know the answer, but the opposite conclusions could also often be easily justified ex post. A further contribution of this study is to provide a way to quantify the extent to which the findings are novel or unexpected by means of an audience expectation elicitation exercise. This study and the results from a companion experiment on wage subsidies (Groh et al., forthcoming) were presented in seminars to academics and development professionals and described online on a popular development economics blog; point expectations of the treatment impacts were then elicited before the results were revealed. The results reveal considerable heterogeneity in the expected outcome of these two employment programs, with a general bias towards expecting soft skills training to be more effective than it was in practice, and for there to be a complementarity with wage subsidies which does not exist in practice. In addition to highlighting the unexpected nature of the findings in this study, we believe the approach pioneered here could be useful for other experimental studies to employ.

The remainder of this paper is structured as follows: Section 2 describes the context, our sample, and the details of the intervention; Section 3 the experimental design, data collection, and intervention take-up; Section 4 the results; Section 5 the expectation elicitation exercise; and Section 6 concludes.

2 Context, sample, and the intervention

Jordan’s labor market is similar to others throughout the Middle East in that there is a bulging youth population facing high unemployment rates, particularly among relatively educated youth, with female youth employment rates particularly low. Youth aged 15–29 are the largest demographic group in the country, comprising 30 % of the total population. In 2011, the unemployment rate for 20–24-year-old females was 47.6 % compared to 23.1 % for 20–24-year-old males (Jordanian Department of Statistics, 2011).

2.1 The sample

We work with graduates of Jordan’s community colleges. These community colleges offer 2-year programs, similar to community colleges in the USA, and like the USA tend to attract individuals with more limited financial means and/or lower academic scores than youth attending universities. At the request of the Government of Jordan, the World Bank developed the Jordan NOW pilot program to attempt to assist women graduating from these community colleges to find employment.

The pilot worked with the eight public community colleges with the largest female enrolment numbers, together comprising over 85 % of all female public community college enrolment. These consist of four colleges in Central Jordan (Amman University College, Princess Alia University College, Al-Salt College, Zarqa University College) and four located in Northern and Southern Jordan (Al-Huson University College for Engineering, Irbid University College, Ajloun University College, and Al-Karak University College). For ease of exposition, since Amman is the capital of Jordan and two of the four colleges in Central Jordan are located within the city of Amman while the other two are located within a 45-min drive of Amman, we will herein refer to Central Jordan as inside Amman and Northern and Southern Jordan as outside Amman.

Baseline surveys were conducted in July 2010 for all second-year students in these colleges sitting for their final exams, giving data on 1755 female students. In August 2010, the baseline data was merged with administrative data on examination results, which revealed that 1395 had passed their examinations. We randomly selected 1350 of these graduates to be our experimental sample. The majority of students had specialized in administration and finance (43 %), which covers specializations such as accounting, electronic administration, and management information systems; courses in medical assistance (24 %), which covers mainly nursing and pharmacy specializations; or educational programs (10 %), which covers those aiming to be teachers.

2.2 The soft skills training intervention

Interventions that aim to teach soft skills may enhance employment prospects by giving youth better skills and confidence for looking for jobs and by making them more productive in their first months in the job by reducing the amount of time firms need to spend training them on the basics of working in a business environment.

Graduates receiving this intervention were invited to free intensive training on interpersonal skills in areas identified by Jordanian employers as essential yet typically missing in recent graduates. The training course was 45 h over a 9-day period (5 h per day), with a maximum of 30 participants in each training group. Training took place during September and October 2010. The training was provided by Business Development Center (BDC), a Jordanian NGO established in 2005 which has widespread local name recognition and a good reputation for skills training, having implemented USAID, UNCTAD, and a wide variety of local training programs. Training took place in 17 sessions offered throughout 6 governorates to maximize access. Training facilities and training content were identical across all 17 sessions. To minimize the effect of social and cultural restrictions on mobility, sessions were held during daylight hours at locally known and trusted institutions such as the Chambers of Industry and local universities where BDC has satellite training facilities.

BDC designed the course which covered effective communication and business writing skills (e.g., making a presentation, writing business reports, different types of correspondence), team building and team work skills (e.g., characteristics of a successful team, how to work in different roles within a team), time management, positive thinking and how to use this in business situations, excellence in providing customer service, and CV and interviewing skills. Sessions were based on active participation and cooperative learning rather than lectures, with games, visual learning experiences, group exercises, and active demonstrations used to teach and illustrate concepts. Appendix 3 provides some more detail of the training schedule and content.

The World Bank and International Youth Foundation (2013) studied 57 organizations providing life skills or soft skills training in the Middle East. The IYF set benchmarks for such programs, among which is that training lasts a minimum of 30 h, finding that the average program in the region lasted only 20 h, with programs focused on employability having fewer hours and those focused more on civic education having more hours. The program we offered therefore offers more than twice as many hours of soft skills training as the average employability program does in the region.3

The cost of the training was approximately $150,000, which was based on up to 600 graduates attending—leading to a cost of $250 per assigned graduate and, given that only 373 attended, an effective cost of $400 per attendee.

3 Experimental design, data collection, and take-up

Randomization into treatment and control groups was done via a computer. Students were first stratified into 16 strata on the basis of geographic region Amman (Amman, Salt, and Zarqa), and outside Amman (Irbid, Ajloun, and Karak), whether their end of high school (Tawjihi) examination score was above the sample median or not, whether they indicated at baseline that they planned to work full-time and thought it was likely or somewhat likely that they would have a job within 6 months of graduating, and whether or not they are usually permitted to travel to the market alone (a measure of empowerment). These variables were chosen as potential predictors of future employment, for which stratification over would improve power (Bruhn and McKenzie, 2009), as well as variables for which there may be potential heterogeneity in treatment impact. The soft skills training intervention discussed in the paper was cross-randomized with a second intervention which provided wage subsidies (discussed in a companion paper, Groh et al., forthcoming). Appendix 1 shows that there was no significant interaction effect between the two treatments; hence, we can just compare all those who were allocated to soft skills training to all those who were not. Within each stratum, 44.4 % of the students were allocated to receive the soft skills training for a total treatment size of 599 and control group of 748.4

Students were informed that this was a pilot program, with insufficient funding to enable program provision to everyone, and that a lottery was being used as a way to ensure everyone had an equal chance of being selected.

3.1 Baseline information by treatment status

Table 1 provides summary statistics on the experimental sample by treatment status. As one would expect given computerized randomization, the treatment and control groups look similar on observable characteristics, and a test of joint orthogonality cannot reject the null hypothesis that these characteristics are jointly unrelated to treatment status.
Table 1

Comparison of means of baseline characteristics by treatment group


Soft skills


p value

Stratifying variables


 In Amman, Salt, or Zarwa




 Tawjihi score above median




 Low desire to work full-time




 Is allowed to travel to the market alone




Other baseline variables










 Mother currently works




 Father currently works




 Has previously worked




 Has a job set up for after graduation




 Has taken specialized English training




 Household owns car




 Household owns computer




 Household has internet




 Prefers government work to private sector




Sample size




Joint orthogonality test



Note: p values are from regressions which control for randomization strata. As a result, they are not available for the variables which were used for stratification

At baseline, the graduates express high levels of desire to work, with 93 % saying they plan to work after they graduate, and 82 % say they think it is very likely or somewhat likely that they will have a job within 6 months of graduating. We define individuals who do not plan to work, do not think they will find a job, or only expect to work part-time as having less desire to work, with 41 % in both groups falling into this category.

The average age of the youth in our study is 21, with only 14 % married at baseline. Only 16 % have any previous work experience, and only 7 % have a mother who currently works. At the time of the baseline survey, which took place only weeks before final examination results were available, only 8 % of students had already found a full-time job for work after graduation. As a result, the majority of our sample consists of young women who have expressed a desire to work, have just graduated, and have not yet found work or worked before.

3.2 Data collection

Data collection consisted of three follow-up surveys, conducted in April 2011, December 2011, and January 2013. This timing allows us to test both whether the soft skills training helps graduates find their first job faster, as well as to address the possibility that in an environment where it takes graduates some time to find a job, it may take more than 1 year to see impacts. All rounds collected data on employment outcomes, and the second follow-up survey also collected information on several well-being measures, including mental health, subjective well-being, and empowerment. The response rates were 92, 96, and 92 % for the first, second, and third follow-ups, respectively, including proxy reports from relatives for 3 and 9 % of the subjects in the second and third rounds. These attrition rates are low in absolute terms and are balanced by treatment status (see Appendix 4), and so we treat attrition as at random in our estimation.

In addition to the survey data, we obtained administrative data from the Social Security Corporation of Jordan. In March 2012, the Social Security Corporation sent us employment data for all graduates who provided their social security numbers in the follow-up surveys, which is 95 % of the sample.

3.3 Take-up of the soft skills training

The soft skills training course was completed by 373 of the 599 graduates (62 %) assigned to treatment. This take-up rate is similar to that of other training courses in developing countries. For example, McKenzie and Woodruff (2014) find an average take-up rate of 65 % for business training programs, with this average including a number of studies which made training a required part of obtaining a loan. Only five students attended a part of the course but did not finish. Participants were asked to give their assessment on a five-point scale of each of seven topics in the course at the end of the course: their rating were extremely positive, with 98 % of participants rating each component as “excellent,” the top of the scale, and the remainder typically rating it as good (Business Development Center BDC 2010). Eighty-one percent of women who participated in the training say that they would have paid 150 JD to take the course had they not had the opportunity to take it for free. In qualitative comments, many participants said the course had given them more confidence and positivity, as well as noting their appreciation for learning practical topics not taught in college.

Table 2 examines the correlates of training take-up. Column 1 considers just the stratification variables, while column 2 adds a number of additional controls. We see training attendance is lower for those who have less desire to work full-time, and for individuals who are married. It is also lower for those with higher Tawjihi scores, perhaps reflecting that these higher-skilled graduates believe they have less need for these additional skills. Attendance is higher for those with brothers (who can perhaps oversee them traveling to training), for those with email, and for individuals studying administrative and business subjects.
Table 2

Correlates of soft skills training take-up

Marginal effects from probit estimation


Amman, Salt, or Zarqa

0.067 (0.108)

−0.057 (0.114)

Tawjihi score above median

−0.232** (0.107)

−0.220** (0.109)

No desire to work full-time

−0.265** (0.107)

−0.270** (0.110)

Is allowed to travel to market alone

−0.083 (0.105)

−0.064 (0.107)

Also given wage subsidy

0.158 (0.105)

0.137 (0.107)

Married at baseline


−0.514*** (0.155)

Wealth index


0.008 (0.032)

Number of brothers


0.056** (0.028)

Number of sisters


−0.019 (0.024)

Has email


0.340** (0.165)

In admin/finance program


0.319*** (0.111)

Sample size



Note: robust standard errors in parentheses

*, **, and *** denote significance at the 10, 5, and 1 % levels, respectively

Unfortunately, neither administrative data from the course itself nor our surveys provide any direct measure of soft skills. Soft skills are difficult to measure in a survey setting, and there are concerns about self-reported Likert scale measures for questions like “I can work with someone who has different opinions from mine”. In follow-up work (Groh et al., 2015), we worked with an organizational psychologist to develop several measures of soft skills in a different Jordanian youth population and show these soft skills are predictive of future employment outcomes. However, this approach required individuals to come in person to a testing center, where they were tested in both individual and group settings, with the tests taking an hour to complete. As a result, this approach would be costly and likely subject to high attrition in a follow-up survey setting. The consequence is that we can only report the above information that participants expressed a high degree of satisfaction with the course content but are not able to distinguish how much they learned during training.

4 Results

To evaluate the impact of assignment to the soft skills training treatment, we estimate the following equation for graduate i via OLS over the three follow-up periods t = 1, 2, 3:
$$ \begin{array}{l}{\mathrm{Outcome}}_{i,t}=\alpha +{\beta}_1{\mathrm{Soft}}_i+{\beta}_2\mathrm{S}\mathrm{o}\mathrm{f}{t}_i\cdot \mathrm{Time}{2}_t+{\beta}_3{\mathrm{Soft}}_i\cdot \mathrm{Time}{3}_t\hfill \\ {}+{\theta}_2\mathrm{Time}{2}_t+{\theta}_3\mathrm{Time}{3}_t+{\displaystyle {\sum}_{s=1}^{16}{\delta}_s{d}_{i,s}+{\varepsilon}_{i,t}}\hfill \end{array} $$
where Soft i is a dummy variable taking the value 1 if graduate i was assigned to the soft skills training treatment, Time2 and Time3 are dummy variables for the second and third follow-up surveys, respectively, and the d i,s are the randomization strata dummies (Bruhn and McKenzie, 2009). The coefficient β 1 then measures the intention-to-treat (ITT) effect of soft skills training in the first follow-up period, a time period 6 to 7 months post training. β 2 and β 3 then enable us to examine whether this effect changes over longer time horizons. The total effect of the treatment in follow-ups 2 and 3 are then given by β 1 + β 2 and β 1 + β 3, respectively, and we test whether these effects are 0. The standard errors are clustered at the individual level. We focus on ITT impacts which give the average effect of being offered training. We choose not to estimate the treatment effect on the treated because of the possibility that the offer of training may affect job search behavior even if the treatments are not taken up. Nonetheless, late estimates for the training effect can be easily approximated by dividing the ITT estimates by the take-up rate (0.62).

4.1 Average impacts on labor force participation and employment

Table 3 reports the results of estimating Eq. (1) for different employment outcomes, defined in detail in Appendix 2. Column 1 examines labor force participation, defined as either working or actively looking for work. At the time of the first follow-up survey, labor force participation was relatively high at 78 % in the control group. Soft skills training has no significant impact on this rate. Labor force participation then falls by over 20 percentage points in the two subsequent surveys with the training having no impact on reducing this drop. Note that the labor force participation variable captures job search activity, which is high to begin with, and falls over time. Soft skills training does not result in more job search.
Table 3

Treatment impacts on employment


Labor force participation


Formal contract

Employed and registered for social security (survey)

Employed and registered for social security (admin data)

Hours worked last week

Work income (not conditional on working)

Work income (conditional on working)










Assigned to soft skills training

0.026 (0.023)

0.012 (0.028)

0.005 (0.023)

0.004 (0.017)

−0.007 (0.015)

0.748 (1.062)

2.262 (4.715)

−0.019 (4.198)

Assigned to soft skills*follow-up 2 (December 2011)

−0.001 (0.031)

−0.015 (0.031)

0.006 (0.026)

−0.001 (0.020)

−0.003 (0.015)

−0.279 (1.249)

−2.971 (5.253)

−2.908 (7.109)

Assigned to soft skills*follow-up 3 (January 2013)

−0.028 (0.033)

0.021 (0.034)

0.017 (0.028)

0.013 (0.025)



8.177 (6.635)

6.435 (9.720)

Follow-up 2 (December 2011)

−0.256*** (0.021)

−0.155*** (0.021)

−0.068*** (0.018)

0.040*** (0.014)

0.066*** (0.010)

−3.839*** (0.827)

−17.296*** (3.637)

17.652*** (4.992)

Follow-up 3 (January 2013)

−0.214*** (0.023)

−0.124*** (0.023)

−0.027 (0.019)

0.070*** (0.017)


−3.996*** (0.889)

−9.125** (4.430)

54.463*** (6.912)

Control mean in first follow-up









Sample size









p value: no effect at follow-up 2









p value: no effect at follow-up 3









p value: equality of follow-up 2 and 3 effects









Note: Huber-White standard errors in parentheses, clustered at the student level. All regressions also control for stratification dummies

n.a. denotes Administrative Data on Social Security Registration not available for the third follow-up period

*, **, and *** indicate significance at the 10, 5, and 1 % levels, respectively

Columns 2 through 4 examine employment. Column 2 looks at whether individuals are currently employed or have worked for cash in the last month. Columns 3 and 4 look at formal employment, defined by whether individuals report being employed and registered for social security in the survey (column 3) and in the social security administrative data (column 4). We see that the soft skills training has no significant impact on any of these measures, neither in the short run nor in either of the two longer-term follow-up surveys. The point estimates are close to 0, especially for formal work, with a 95 % confidence interval for the ITT effect of training on employment 6 months later of (−0.043, +0.067) and for formal employment registered with social security of (−0.029, +0.037). As Hoenig and Heisey (2001) argue, these confidence intervals, rather than post-experiment power calculations, are the appropriate guide for interpreting statistically insignificant results and show the range of possible effect sizes not refuted by our data enables us to rule out large impacts of the treatment.

The remaining columns of Table 3 consider dimensions of employment at the intensive margin hours worked and income earned. Columns 6 and 7 are unconditional estimates, which code the outcome as 0 for those not working, while column 8 looks descriptively at income conditional on work. We find no significant impact of the soft skills training treatment on either hours worked or income earned.

The point estimate for the short-term increase in income is 2.3 JD per month, or approximately US$3. The training cost is $250 per individual invited, so these gains (which were not statistically significant) would have to last for 82 months to recoup the cost of training. It is important to note that these results do not necessarily imply that soft skills are not important determinants of income. Since we do not observe soft skills directly, we cannot say whether the lack of effect is due to the training not being very effective at increasing soft skills or to there being little return to these skills. In either case, the results here imply that the training does not appear cost effective.

4.2 Heterogeneity in employment impact

The lack of overall effectiveness of the soft skills training raises the question of whether training may be more effective if targeted at specific subgroups. To do this, we pool together follow-up rounds 2 and 3 to increase power and ease in interpretation of coefficients and estimate the following equation:
$$ \begin{array}{l}{\mathrm{Employed}}_{i,t}=\alpha +{\beta}_1{\mathrm{Soft}}_i+{\beta}_2{\mathrm{Soft}}_i\cdot \mathrm{Time}2 or{3}_t+{\gamma}_1{\mathrm{Soft}}_i\cdot {X}_i\hfill \\ {}+{\gamma}_2{\mathrm{Soft}}_i\cdot {X}_i\cdot \mathrm{Time}2\mathrm{or}{3}_t+{\theta}_2\mathrm{Time}{2}_t+{\theta}_3\mathrm{Time}{3}_t+{\delta}_2\mathrm{Time}{2}_t\cdot {X}_i\hfill \\ {}+{\delta}_3\mathrm{Time}{3}_t\cdot {X}_i+{\displaystyle {\sum}_{s=1}^{16}{\delta}_s{d}_{i,s}+\mu {X}_i+{\varepsilon}_{i,t}}\hfill \end{array} $$
where X is the baseline characteristic of interest for heterogeneity testing and Time2or3 indicates the interaction with being in the second or third follow-up round. We examine heterogeneity with respect to the four variables used to stratify the randomization (and which can therefore be considered as pre-specified dimensions of heterogeneity to examine), as well as with respect to baseline marital status since this was a strong predictor of training take-up and of subsequent employment status.
Table 4 reports the results of estimating Eq. (2) to examine how the impact of soft skills training on employment varies with individual characteristics. The foot of the table then tests whether there is a significant impact of the training in the pooled round 2 and round 3 surveys for the case when the interaction variable is 0, and for when it is 1. We see that in no case can we reject the null hypothesis of no impact of treatment. That is, it is not the case that the soft skills training works particularly well for those in Amman, or those with higher ability, or those who are not married to begin with, etc. This is despite these variables having predictive power for employment outcomes. For example, we see that in the first follow-up only 22 % of those married are employed, compared to 36.5 % of those who were unmarried at baseline, but that we cannot reject that training has the same effect for both groups and that this effect is 0.
Table 4

Heterogeneity in treatment effects

Dependent variable: employed


Interaction is between treatment and:



High aptitude

Low desire to work

Allowed to travel


Assigned to soft skills

0.023 (0.038)

0.034 (0.042)

0.006 (0.037)

0.057 (0.041)

0.020 (0.031)

Assigned to soft skills*follow-up 2 or 3

0.017 (0.039)

−0.002 (0.044)

−0.005 (0.040)

−0.051 (0.042)

−0.007 (0.033)

Assigned to soft skills*interaction

−0.021 (0.057)

−0.040 (0.057)

0.017 (0.058)

−0.087 (0.057)

−0.060 (0.075)

Assigned to soft skills*follow-up 2 or 3*interaction

−0.035 (0.060)

0.008 (0.060)

0.018 (0.060)

0.105* (0.060)

0.069 (0.074)

Follow-up 2

−0.221*** (0.026)

−0.172*** (0.031)

−0.160*** (0.028)

−0.131*** (0.028)

−0.159*** (0.023)

Follow-up 3

−0.161*** (0.028)

−0.120*** (0.033)

−0.106*** (0.030)

−0.090*** (0.030)

−0.114*** (0.024)

Follow-up 2*interaction

0.134*** (0.042)

0.015 (0.042)

−0.009 (0.041)

−0.063 (0.041)

−0.029 (0.050)

Follow-up 3*interaction

0.103** (0.044)

0.007 (0.044)

−0.027 (0.044)

−0.053 (0.044)

−0.019 (0.056)

Control mean in first follow-up when interaction equals 0






Control mean in first follow-up when interaction equals 1






Sample size






p value: no effect when interaction = 0 at follow-ups 2 and 3






p value: no effect for interaction = 1 at follow-ups 2 and 3






p value: equality of effect by interaction at follow-ups 2 and 3






Notes: Each column shows results from interacting treatment effect with the variable listed at the top of the column. All regressions also control for randomization strata dummies, while the last column also controls for marriage at baseline. Huber-White standard errors in parentheses, clustered at the student level

*, **, and *** indicate significance at the 10, 5, and 1 % levels, respectively

4.3 Impact on well-being, empowerment, and attitudes

As noted above, the training was rated extremely highly by those who attended it, with the participants saying that it had given them more confidence and positivity. This raises the possibility that it may have had impacts in other dimensions of life, even if it did not affect employment. We investigate this in Table 5.
Table 5

Impacts on well-being, empowerment, attitudes, and marriage


Life ladder

Life ladder future


Severely poor mental health

Mobility index

Empowerment index

Married at 2nd follow-up

Assigned to soft skills
















Control mean








Sample size








Note: Robust standard errors in parentheses. All regressions include stratification dummies. Severely poor mental health denotes the MHI-5 index is below 17

*, **, and *** denote significance at the 10, 5, and 1 % levels, respectively

Column 1 examines the impact on current subjective well-being, measured on the Cantril self-anchoring striving scale (Cantril, 1965), a measure that has been used by Gallup around the world. Respondents are asked to imagine a ladder with 11 rungs, number from 0 at the bottom to 10 at the top, where the top represents the best possible life for them, and the bottom the worst possible life. Kahneman and Deaton (2010) refer to this as “life evaluation.” The control group mean (s.d.) is 5.2 on this scale, and being assigned to soft skills training lowers this by 0.23 rungs, significant at the 10 % level. Column 2 examines which step on the Cantril ladder graduates believe they will be on in 5 years’ time, with this forward-looking measure reflecting the degree of optimism they have about their futures. Overall, the graduates show a high degree of optimism, with the mean of 8.0 for the control group three rungs higher than they assess their current position to be. The training intervention leads to additional optimism: graduates assigned to this group think they will be 0.23 steps higher still, significant at the 5 % level, and equivalent to a 0.1 standard deviation increase. So, training participants are less satisfied with their current status but more optimistic about the future. Such a result could arise if the training causes them to think about a better possible future life for themselves than they had previously imagined, making their current status seem less desirable.

Mental health is a distinct concept of well-being from happiness and has been shown to have different associations with individual characteristics and with life events (Das et al., 2008). We measure mental health using the Mental Health Inventory (MHI-5) of Veit and Ware (1983). This is a five-item scale with a maximum score of 25 and minimum score of 5, with higher scores indicating better mental health in terms of the experience of psychological well-being and the absence of psychological distress in the past month. While there is no universal cutoff, several studies have used a cutoff of less than 17 as an indicator of major depression (e.g., Urban Institute, 1999; Yamazaki et al., 2005). Twenty percent of the control group has scores below this threshold. Columns 3 and 4 look at the impact on the overall index, and on the binary classification of depression, respectively. We find that the training intervention results in a small increase in the MHI-5 and reduction in the likelihood of major depression, but neither is statistically significant.

Columns 5 and 6 of Table 5 examine whether there were changes in two measures of empowerment: the ability to go to different places themselves, and attitudes towards women in society. We find a highly significant increase in the ability to travel to different places: the mean in the control group is being able to travel to 5.0 out of 6 places by herself, with the training increasing this by 0.38 places. In contrast, there is no significant change in attitudes towards women in society. Finally, we note that while at baseline (graduation from community college) only 13.7 % of the sample was married, this had increased to 31.6 % by the second follow-up 18 months later. Column 7 shows that training had no impact on the likelihood of marriage.

All together, there are seven different outcomes in Table 5, of which three are significant at the 10 % level. If we correct for multiple hypothesis testing using a Bonferroni or a step-down procedure, only the impact on mobility remains significant. Therefore, it appears that overall the impacts on other life outcomes are also relatively modest.

5 Are these impacts unexpected?

Ex post, it is relatively easy for policymakers and academics to view the result of almost any impact evaluation and claim that the results are exactly what they would have expected. In a classic article, Lazarsfeld (1949) sets out six findings from a study of American soldiers (e.g., Southern soldiers were better able to stand the climate in the hot South Sea Islands than Northern soldiers), and then for each, a probable reader reaction (e.g., of course, Southerners are more accustomed to hot weather). He then reveals that each of these statements was false (e.g., Southerners had no greater ability than Northerners to adjust to a tropical climate). Watts (2011) refers to this as the “myth of common sense” and means that ex post, either a large positive impact, a zero impact, or a negative impact may all be easily explained away as obvious.

To understand whether this study merely confirms conventional wisdom or generates new and unexpected results, we undertook a novel audience expectation elicitation exercise. During the first four presentations of this research (two in academic departments, and two in international organizations),5 no paper was distributed or available in advance. Instead, the audience was presented with 25 slides detailing the motivation, existing evidence, context, design, and implementation of this study and of the cross-randomized wage subsidy experiment. We then distributed paper surveys throughout the audience and asked them to provide their point estimates of the first and second follow-up ITT impacts for each treatment group: soft skills training, wage subsidy only, and both combined. To complement the audience expectation elicitation exercises, we posted a description of the study on the World Bank’s Development Impact blog6 with a link to an online survey to capture reader expectations.7 The advantage of seminar presentations is that they offer time to answer any questions, explain the intervention in detail, and get high response rates (very few seminar attendees refused to fill in the one-page expectations sheet). In contrast, online elicitation opens up the process to a wider audience but offers less of a chance to explain the intervention, and has low and self-selected response rates.8 In both cases, we explicitly asked the audience to give the ITT impact but gave them the take-up rate for the soft skills training so that the results would reflect beliefs about treatment impact and not treatment take-up.

Figure 1 provides histograms of the resulting expectations about the soft skills training impact, while Table 6 summarizes the resulting expectations and compares them to their expectations of the wage subsidy effects. We see first that there is tremendous heterogeneity in expectations about the likely effect of the soft skills training: the standard deviation is 10 percentage points for the first follow-up impact and 11 for the second. Second, on average, people are biased towards thinking soft skills training will be more effective than it was in practice. The mean expected impact of 9 to 10 percentage points is three to five times as large as the estimated ITTs. Only 63 % of the expectations lie within a 95 confidence interval of the treatment effect.
Fig. 1

Histograms of distribution of audience expectations of ITT impacts of soft skills training on employment

Table 6

Expected impact of the interventions from seminar participants and online respondents



Actual treatment effect

μ expected

σ expected


% of expectations within 95 % CI

Expected impact at first follow-up


 Wage subsidies alone







 Soft skills training alone







 Wage subsidy and soft skills







Expected impact at second follow-up


 Wage subsidies alone







 Soft skills training alone







 Wage subsidy and soft skills







We also see that expectations were inaccurate for the wage subsidy program, with respondents dramatically underestimating the short-term effect of the subsidy and overestimating the longer-term effect. Only 4 % of the 136 respondents gave an expected value of the voucher impact at midline that lies within the 95 confidence interval of the treatment effect, with the median expected impact of 8 percentage points less than one quarter of the actual impact of 39.5 percentage points. Audiences believed the combination of the wage subsidy voucher and soft skills training was likely to have larger impact than either treatment alone, but this complementarity did not occur in practice. None of the respondents had all six of their expectations lie within the 95 % confidence intervals of the actual impacts. Furthermore, there was considerable heterogeneity among respondents in their relative rankings of the interventions: at the first (second) follow-up, 59 (47) % thought the wage subsidy would have a larger impact than the soft skills training, 11 (14) % that the impact would be the same, and 30 (39) % that the wage subsidy would have less impact than the soft skills training.

While the sample size is relatively small to examine in detail the heterogeneity of expectations across groups, we can examine the extent to which expectations differ among the five different audiences (four seminars and the blog). There are three results of interest here. First, the results are not driven by any one audience. Second, overall, the most accurate responses were from blog readers, although these responses were not that different from participants in a World Bank seminar. Third, the least accurate responses were from the Paris School of Economics audience, who were most optimistic about the soft skills training having longer-term effects. We cannot reject that expectations of the soft skills training impact were the same among the other four groups (p = 0.19 short-run, p = 0.52 long-run). We speculate that this might reflect a European view that is more positive about active labor-market policies generating employment than is the belief in the USA, which could be interesting to test in the future such exercises.9

Finally, a simplified version of the expectations was also elicited from nine Jordanian policymakers in a presentation of the results made in Jordan, who were just asked expectations in five bin ranges (<0, 0, 0 to 5 %, 5 % to 10 %, more than 10 %). The policymakers’ responses also show considerable heterogeneity in responses, and the majority of policymakers underestimated the short-term impact of the wage subsidy voucher.

Overall, these results therefore show that there is considerable uncertainty about what the likely impact of soft skills training is likely to be but that people are on average over-optimistic about its impacts and over-optimistic about the possibility of complementarities between policies.

6 Conclusions

Soft skills training programs have become an increasingly common type of youth employment policy in developing countries, but to date, there has been little evidence as to their impact. Our randomized experiment finds that even though our soft skills training program was provided by a well-regarded provider, was of twice the length as the regional average, and had very positive post-training participant assessments, it did not have any significant impact on employment outcomes of young women over either the short or the medium term. These findings are in contrast to the expected impact of such programs from academic audiences and policymakers. The expectation elicitation exercise conducted here both serves as an illustration of an approach that could be used in other studies to reveal the extent to which impact evaluations confirm or contrast with existing priors, as well as serving to show that the results generated here are different from what many people would expect.

While these results are disappointing, they are consistent with findings from recent overviews of vocational training programs (Blattman and Ralston, 2015) and business training programs (McKenzie and Woodruff, 2014) which have noted that these other forms of training programs have also often struggled to show significant employment impacts. This other literature suggests two possible reasons for the lack of effectiveness of soft skills training here. The first is that relatively short courses may only result in only small changes in the skills that they aim to teach. Measuring soft skills well is difficult, and we were not able to observe the extent to which the training actually improved soft skills for the participants. Devoting more attention to how to measure changes in soft skills therefore seems an important area for future work, with Groh et al. (2015) providing one example of how to do this on a different sample of Jordanian youth. It would also be useful to examine the effectiveness of such programs for young men, in addition to the young women studied here. The second possible reason for a lack of effect is that the diagnosis that a lack of skills is the key constraint to employment may be incorrect. The more binding constraint may be more general constraints that prevent productive firms from growing and employing more workers, in which case policy efforts may need to be directed more towards helping firms than helping workers.


A recent working paper by Honorati (2015) provides a second example that helps confirm our results. She tests a 2-week program in Kenya and finds no significant impact on employment outcomes.


The program also cross-randomized a wage subsidy component which is covered in a companion paper (Groh et al., forthcoming). Appendix I shows this program had no interaction effect with the soft skills training.


Our course is also similar in length to the 40-h life skills training in Nepal (Ahmed et al., 2015) and the 2-week training in Kenya (Honorati, 2015).


One individual in each of the treatment and control groups had his/her gender misidentified in the data and was dropped from the study.


This comprises presentations in May and June 2012 at the Applied Micro workshop at the University of Virginia; the development seminar at the Paris School of Economics; the applied micro seminar at the World Bank; and a seminar with the labor group of the InterAmerican Development Bank. The World Bank and IADB seminars contained a mix of researchers and development professionals engaged in the implementation of programs. We thank the seminar audiences for their participation in this exercise.


To our knowledge, this is the first paper to systematically collect audience expectations for an impact evaluation. Dean Karlan and Annie Duflo concurrently collected qualitative expectations during an online poll at the Stanford Social Innovation Review


The Development Impact blog post had 900 page views, along with approximately 2000 RSS and Email subscribers (of which only some would have actually read the post). From this, we received 48 responses, a click-through rate consistent with click-through from blog post links to academic papers (McKenzie and Özler, 2014).


A breakdown of results by these five groups can be found on the Development Impact blog [accessed February 6, 2016].




We would like to thank the editor and the anonymous referees for the useful remarks.

We thank the World Bank’s Gender Action Plan, TFESSD, and PSIA trust funds for funding this project. We thank Nithin Umpathi and Abdalwahab Khatib for their assistance in implementing the project, and the editor, an anonymous referee, and seminar audiences in Jordan, Qatar, at Bar-Ilan, BREAD, the IADB, the IZA/WB Employment Conference, the Paris School of Economics, the University of Virginia, and the World Bank for useful comments. An earlier version of this paper is circulated under the title “Soft Skills or Hard Cash? The impact of training and wage subsidy programs on female youth employment in Jordan”—this paper has subsequently been split into two separate papers. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors and should not be attributed to the World Bank or any affiliated organization.

Responsible editor: David Lam

Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (, which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made.

Authors’ Affiliations

World Bank


  1. Adoho F, Chakravarty S, Korkoyah Jr. D, Lundberg M, Tasneem A (2014) The impact of an adolescent girls program: the EPAG Project in Liberia. World Bank Policy Research Working Paper No. 6832.
  2. Ahmed A, Chakravarty S, Lundberg M, Nikolov P (2015) The role of training programs for youth employment in Nepal: Impact Evaluation Report on the Employment Fund, AGI Results Brief., Google Scholar
  3. Blattman C, Ralston L (2015) Generating employment in poor and fragile states: evidence from labor market and entrepreneurship programs. Columbia University, MimeoGoogle Scholar
  4. Bowles S, Gintis H, Osborne M (2001) Incentive-enhancing preferences: personality, behavior and earnings. Am Econ Rev 91(2):155–158View ArticleGoogle Scholar
  5. Bruhn M, McKenzie D (2009) In pursuit of balance: randomization in practice in development field experiments. Am Econ J: Appl Econ 1(4):200–232Google Scholar
  6. Business Development Center (BDC) (2010) “Jordan Now Training Program: Training report 25th September – 2nd November, 2010”, Report delivered to the World Bank.
  7. Cantril H (1965) The pattern of human concerns. Rutgers University Press, New Brunswick, NJGoogle Scholar
  8. Cunningham W, Villaseñor P (2014) Employer voices, employer demands, and the implications for public skills development policy, World Bank Policy Research Working Paper no. 6853.
  9. Das J, Do Q-T, Friedman J, McKenzie D (2008) Mental health patterns and consequences: results from survey data in five developing countries. World Bank Econ Rev 23(1):31–55View ArticleGoogle Scholar
  10. Entra 21 (2009) Final Report on the entra21 Program Phase I 2001–2007. International Youth Foundation.
  11. Groh M, McKenzie D, Vishwanath T (2015) Reducing information asymmetries in the youth labor market of Jordan with psychometrics and skill based tests. World Bank Econ Rev Papers Proc 29(suppl 1):S106–S117View ArticleGoogle Scholar
  12. Groh M, Krishnan N, McKenzie D, Vishwanath T (forthcoming) Do wage subsidies provide a stepping stone to employment for recent college graduates? Evidence from a randomized experiment in Jordan, Review of Economics and Statistics.
  13. Heckman J, Kautz T (2012) Hard evidence on soft skills, IZA Discussion Paper no. 6580Google Scholar
  14. Heckman J, Stixrud J, Urzua S (2006) The effects of cognitive and noncognitive abilities on labor market outcomes and social behavior. J Labor Econ 24(3):411–482View ArticleGoogle Scholar
  15. Hoenig J, Heisey D (2001) The abuse of power: the pervasive fallacy of power calculations for data analysis. Am Stat 55(1):1–6View ArticleGoogle Scholar
  16. Honorati M (2015) The impact of private sector internship and training on urban youth in Kenya, World Bank Policy Research Working Paper no. 7404.
  17. Ibarraran P, Ripani L, Tapoada B, Villa JM, Garcia B (2012) Life skills, employability and training for disadvantaged youth: evidence from a randomized evaluation design, IZA Discussion Paper no. 6617.
  18. Jordanian Department of Statistics “Table 2.4: Jordanian Population Age 15+ Years by Activity Status, Sex and Broad Age Groups (Percentage Distribution) – 2011”, [accessed 24 April, 2011].
  19. Kahneman D, Deaton A (2010) High income improves evaluation of life, but not emotional well-being. PNAS 107(38):16489–16493View ArticleGoogle Scholar
  20. Lazarsfeld P (1949) The american soldier—an expository review. Public Opin Q 13(3):377–404View ArticleGoogle Scholar
  21. McKenzie D, Özler B (2014) Quantifying some of the impacts of economics blogs. Econ Dev Cult Chang 62(3):567–597View ArticleGoogle Scholar
  22. McKenzie D, Woodruff C (2014) What are we learning from business training and entrepreneurship evaluations around the developing world? World Bank Res Obs 29(1):48–82View ArticleGoogle Scholar
  23. Scheiber N (2007) Freaks and Geeks; How Freakonomics is ruining the dismal science, The New Republic, April 2. [accessed July 9, 2012].
  24. Urban Institute (1999). Snapshots of America’s Families: Appendix, Accessed March 2, 2016.
  25. Veit C, Ware J (1983) The structure of psychological distress and well-being in general populations. J Consult Clin Psychol 51:730–742View ArticleGoogle Scholar
  26. Watts D (2011) Everything is obvious (*once you know the answer): how common sense fails Us. Crown Business, New YorkGoogle Scholar
  27. White M (2013) The real reason new college grads can’t get hired, Time, November 10,
  28. World Bank (2010) “Active labor market programs for youth: a framework to guide youth employment interventions”. World Bank Employment Policy Primer No. 16.
  29. World Bank and International Youth Foundation (2013) Preparing youth for success: an analysis of life skills training in the MENA region. World Bank, Washington, D.CGoogle Scholar
  30. Yamazaki S, Fukuhara S, Green J (2005) Usefulness of five-item and three-item Mental Health Inventories to screen for depressive symptoms in the general population of Japan, Health and Quality of Life Outcomes, 3/1/48.


© Groh et al. 2016