Presentation on theme: "DOES TRAINING VOUCHER HELP INCREASE INCOME OF REDUNDANT WORKERS OF RESTRUCTURED SOES IN VIETNAM? Lan Anh Vu Central Institute for Economic Management Vietnam."— Presentation transcript:
DOES TRAINING VOUCHER HELP INCREASE INCOME OF REDUNDANT WORKERS OF RESTRUCTURED SOES IN VIETNAM? Lan Anh Vu Central Institute for Economic Management Vietnam
Key contents: - Abstract - Aims, research questions - Literature Reviews - Policy relevance for Vietnam - General description - Randomized experiment - Modeling and testing - Research team - Partners/collaborators - Task allocation and workplan
ABSTRACT (1) 5,203 Vietnamese State-owned enterprises have been restructured, inclusive of 3,680 equitized About $410 m was disbursed to nearly 200,000 redundant workers A number of schemes have been introduced, inclusive of training voucher. Is training voucher directly improving the human capital (get new jobs, start new businesses...)? Are there other factors influencing the increased earnings of this group?
ABSTRACT (2) A number of redundancies need additional skills to enhance their job-seeking opportunities, especially in the increasing competitive labor market in Vietnam; There have not been any quantitative studies on the actual impact of this program over the increase in income of redundancies? => necessary to assess the impact of this program
AIMS To determine whether the training voucher programs is necessary If yes, which policy adjustments should be made to enhance its efficiency
Core Research Objectives Assess the impact of the training voucher scheme over the employability of redundancies of restructured SOEs. Assess the impact of the training voucher scheme over monthly income of redundancies of restructured SOEs. Make policy recommendations
Research questions Does training voucher scheme help redundancies of restructured SOEs in Vietnam improve their employability? To what extent does training voucher scheme have impacts over the monthly income of redundancies of restructured SOEs in Vietnam? What are policy implications on the training voucher scheme?
LITERATURE REVIEWS (1) Job training programs are quite common in developed countries like the USA and Europe. In the USA, credible randomized evaluations can be found in such programs as the Job Partnership Training Act (Bloom et al, 1997; GAO, 1996; Heckman et al, 1999), the Job Corps (Burghart and Schochet, 2001) and a series of program for welfare recipients (Friedlander et al, 1997). Such studies produce quite different results.
LITERATURE REVIEWS (2) Job Partnership Training Act, the short-run impacts for young disadvantaged women are essentially zero (although the long-run impacts appears to be more positive (GAO, 1996)), while the short-run impacts for young disadvantaged men are negative Job Corps had a significant positive impact on both genders. Lee (2005) proves that Job Corp had about a 12 percent effect on earnings three years after training.
LITERATURE REVIEWS (3) Many non-experimental evaluations: training programs help increase post-program employment rate rather than increase wage rate; training makes a positive contribution on wage growth which translates into a company rate of return of at least of 13%; using data on performance ratings shows that training leads to an improvement in job performance
LITERATURE REVIEWS (4) literature on cases of developing countries remains more limited. Betcherman et al: training impacts in Latin America are more positive than the impacts of programs in the USA and Europe. Card et al (2007), with an experimental design for a job training program in Latin America: the Juventud y Empleo (JE) program in the Dominican Republic had no significant effect on employment. There is evidence of a modest (10%) impact on hourly wages and earnings per month, although the estimated are only marginally significant. Another randomization in Argentina Galasso et al, 2001: Voucher recipients had a significantly higher probability of employment, though their current incomes were no higher, and training had no significant impact
LITERATURE REVIEWS (5): the gaps? Randomization researches on vocational training in developing countries are more limited than in developed countries in which the number of people need vocational training may be higher. No randomization studies on vocational training programs in Vietnam yet. Even non-experimental researches on this issue are quite rare. Most programs in other countries aim at the youth, the poor or some other disadvantaged groups. There are few quantitative studies on impacts of vocational training provision for redundancies due to enterprise restructuring – who may need a special scheme.
LITERATURE REVIEWS (6): Conclusion Different results from different studies in different places show that it is necessary to do a special study for a special case (country) In Vietnam, there is no quantitative study on this issue. There is still the gaps in studying to fill up. To make useful policy advice, we should conduct a serious study.
The study should be a randomization: To assess the policy impacts, we should realize difference between supported and non-supported redundancies, so the random experimental method is suitable for the study Econometrics methods and some software programs such as STATA, Eviews,… can help us to estimate that difference.
POLICY RELEVANCE FOR VIETNAM A vocational training scheme for redundancies due to SOE restructuring – a start for other national similar schemes A scientific study on this issue as randomized experiment on training voucher scheme for redundancies may help identify the actual impacts of the scheme, which provides quantitative input for policy recommendations. There should be programs for redundancies of private companies or programs for the youth, so the results from this study may shed some light for such future programs in the future.
GENERAL DESCRIPTION (1) Conduct a randomization project on the targeted population 1,500 SOEs to be restructured, the estimated population of redundancies may be around 100,000. About half of them are under 45 and need to get new jobs for their living. A sample will be drawn by selecting redundancies in SOEs starting their restructuring in 2007, between 18 and 45, who have completed secondary schooling level at least and have not taken any vocational training courses after being made redundant
GENERAL DESCRIPTION (2) The sample will be divided equally into control and treated groups. The treated group will receive free vocational training. Both groups will also receive job search assistance. After twelve months from the completion of training, both groups will be surveyed. Expected outcomes: Those in the treated group can find jobs Their monthly earning is higher The increased income is economically significant
GENERAL DESCRIPTION (3) Timing: The twelve-month period may be deemed sufficient for the evaluation of impacts. Randomization studies on training programs in the US also have similar timing of effects. Data: Both existing and new data will be used. Existing data from Ministry of Finance to get the overall information of redundancies. New data from the baseline and follow-up surveys.
RANDOMIZED EXPERIMENT Roughly select eligible candidates from the lists of redundancies of restructured SOEs (high number of redundancies, operating in a limited number of industries and concentrated in certain provinces) A baseline survey will be conducted to identify characteristics of redundancies Randomly and equally divide them into control and treatment groups. The treatments will receive free vocational training, while the controls will not and also not take any vocational training courses funded by the Government. Both groups will receive job search assistance.
RANDOMIZED EXPERIMENT Training courses may last for three months and redundancies then need at least six to twelve months to search for new jobs or start new businesses; The survey will be conducted to assess the impact of the scheme.
Selection of Training Providers (1) How to choose the best training institutions for participants? via a competitive process (regardless of their ownership structure). Proposals from potential training providers were required to include written commitments from one or more firms to offer a two-month internship to all trainees from the provider’s program (if any). They will be also required to provide counseling and technical assistance
Selection of Training Providers (2) Several training institutions may be selected depending on types of training courses and location of training to reduce training cost of the experiment. However, there will be only one training curriculum for each training topic. The Department of Vocational Training of Ministry of Labor, War Invalids and Social Affairs may help recommend bidders and evaluate training proposals.
Selection of Training Providers (3) Job search assistance: Training providers also provide information on local labor market demands, job vacancies and job brokerage centers. Such information will be circulated to participants by the research team to ensure homogeneous information supply.
Choosing Participants In selected enterprise/province, a group of at least 40 eligible participants are identified 20 treated participants selected by lottery among eligible candidates; the other 20 will be considered as the control group. Up to 10 people in the control group may be reassigned to the treatment group by lottery if one or more of the original treatments fail to show up for training (“no-show”) or drop out within the first two weeks of the course (“drop-out”). Certain demographic characteristics of redundancies (such as age, gender) and/or initial income level may be taken into account to ensure the initial similarity of the control and treated groups
Training Course The training course will be designed to meet the labor demand of local employers. Trainees can enjoy technical counseling from training providers and internship scheme, as well as job search assistance as mentioned above. The control group may get job search assistance as a benefit of participation in the experiment
Estimated Training Cost Tentatively the training cost per participant will be USD 200. In addition, daily stipend to support for their lunch and traveling cost. Some participants who live far from the training venue may receive additional allowance for hotel accommodation.
Data Collection Method Information on restructured SOEs between 2007 and 2009 got from the State Capital Investment Corporation (under Ministry of Finance) who is assigned to manage and disburse the Redundancy Fund will help save costs. Dealing with the enterprises’ managers who help circulate invitations to the experiment among eligible redundancies of these enterprises will help get information from applicants (a survey can gathers more detailed information on their age, education, employment status, income, training need, etc)
Data Collection Method Designing questionnaires for the surveys: to get initial /basic information of redundancies to get information from them after the experiment, different for the treatment and the control Twelve months after the training, the treatment group will be interviewed. The control group will be also interviewed (maybe earlier) to ensure that both groups will be assessed after twelve months of participation.
Sample size Since there is no other related studies except for the survey of the Ministry of Finance which does not provide sufficient information, we has to make their own guess. To calculate the sample size under the formula proposed by Bloom (2006) Empirically, our sample size of 350 is sufficiently credible (n treated =174).
Calculating sample size by Bloom’s method Consider a balanced allocation of sample size, i.e. half of the sample is randomized to a treatment group and half is randomized to a control group, and everyone adheres to their assigned treatment. Follow-up data are obtained for all sample members and the treatment effect is estimated by the difference in mean outcomes for the two groups. This difference provides an unbiased estimate of the average treatment effect (ATE) for the study sample, because the mean outcome for control individuals is an unbiased estimate of what the mean outcome would have been for treatment individuals had they not been offered the treatment (their counterfactual) However any given sample can yield a treatment group and control group with pre-existing differences that occur solely by chance and can overestimate or underestimate the ATE
Calculating sample size by Bloom’s method The standard error of the impact estimator accounts for this random error is: Where: P is the proportion of the sample that is randomized to treatment N is the sample size is the outcome variance across subjects in the experiment group And Where: MDE (minimum detectable effect) is the smallest true treatment effect that the research design can detect with confidence. M n-2 is the multiplier given by a t-table n = [σ 2 (M n-2 /MDE) 2 ]/[P*(1-P)], so if we want to have confidence at least 90%, MDE = 0.3, while P = 50%, M n-2 = 2.8, then sample size should be 350
Questionnaires – the baseline The baseline questionnaires will be the same for both control and treated groups. Key data include: Classification of nominal data to identify control and treatment individuals Contact addresses of the redundancy and one his/her next of kin Demographic characteristics (age, gender…) Education level Living location Income level and sources of income at the time of redundancy Level of working skills, reasons of redundancy Training needs Expectation of new jobs (type of job, expected earnings, whether willing to move to other regions to search jobs, etc). Commitment to experiment participation Any suggestion to experiment organizers
Questionnaires – the post-experiment Key data include: Assessment of training courses taken (suitability, usefulness, quality… this question only for the treated participants) Do they need to take any vocational training other than the one provided under the experiment? Monthly earnings after twelve months of experiment Income level and sources of income after twelve months of experiment How is their new job finding? Types of new jobs or new business they have found? How long can they find a new job? How easy to find a job? Why? Is it a long-term or short-term job? Are they entitled to social insurance scheme? Are training skills useful and/or suitable with their new jobs? (Note: this question is only for the treated participants) Other reasons for income gain Any suggestions to government policy …
Which troubles may occur and solutions (1) Those in the control group may also receive training vouchers provided by the Government; a survey of the Ministry of Finance shows that only 3.7% of redundancies have taken training courses so far, it shows the hesitation of redundancies on taking vocational training courses the training should be conducted right after the redundancies and their training needs are identified.
Which troubles may occur and solutions (2) May be difficult to identify the whole population within several months since SOEs who may conduct their restructuring process at different times in the year. To solve this, the research team will try their best to identify as many redundancies within several months as possible. In addition the team intends to extend the selection to redundancies of SOEs started their restructuring since 2007.
Which troubles may occur and solutions (3) The treatments are either dropout or no-shows, and/or the control may move to the treatment or leave the outcome of the analysis may not show the true picture of the average value of the training. To prevent these: a competitive bidding process to select the best training providers to ensure high training quality. participants will be informed that they will get job search assistance after completing the courses. a cash allowance and regularly trace them at least once a quarter
Which troubles may occur and solutions (4) Difficult to trace participants after twelve months. The individuals who are most successful in the market might be harder to interview because they have moved elsewhere or because they are so busy working that they do not have time to be interviewed. Alternatively, those who are least successful might be easier to interview- because they aren’t working-or might be harder to find and interview, for example if they migrated to another city for working.
Which troubles may occur and solutions (5) To solve this problem : job search assistance and certain allowance for the controls to encourage their full participation. Meanwhile the treatments will not receive the original of the training certificate. participants have to provide contact information of themselves and one of their closest relatives. When they move to other provinces to find jobs, the team may contact their next of kin for information. the team will try to trace them every quarter by phones and/or emails to timely include substitutes if some of them cannot be traced after six months. if the attrition rate remains high, the team may use statistical techniques to solve the attrition bias problem
What does the research team do? be in charge of designing and monitoring the experiment. deal with SOEs, redundancies, training providers and related government agencies, conduct the selection of training providers and experiment participants. cooperate with other organizations to conduct the data collection, for example provincial Department of Labor, War Invalids and Social Affairs, local Statistic Offices conduct data analysis and report writing, once the experiment completes.
MODELING AND TESTING (1) Assume that we have a large number of redundancies. Some take training courses and others do not. We can take the average of both groups and examine the difference between average monthly earnings of those taking training courses and those not taking training courses. In a large sample, this will converge to D = E[Y i T |redundance with training] – E[Y i C |redundance without training] = E[Y i T |T] – E[Y i C |C] = E[Y i T |T] – E[Y i C |T] - E[Y i C |C] + E[Y i C |T] = E[Y i T – Y i C |T] + E[Y i C |T] – E[Y i C |C]
MODELING AND TESTING (2) When we conduct the randomization with treatment and control groups, the average treatment effect can then be estimated as the difference in empirical means of Y between the two groups. Where Ê denotes the sample average. As the sample size increases, this difference converges to D = E[Y i T |T] – E[Y i C |C] Since the treatment has been randomly assigned, individuals assigned to the treatment and control groups differ in expectation only through their exposure to the treatment. Had neither received the treatment, their outcomes would have been in expectation the same. This implies that the selection bias, E[Y i C |T] – E[Y i C |C], is equal to zero. If, in addition, the potential outcomes of a redundance are unrelated to the treatment status of any other redundancies, we have E[Y i |T] – E[Y i |C] = E[Y i T – Y i C |T] = E[Y i T – Y i C],
MODELING AND TESTING (3) The estimation equation is Y = ß 0 + ß 1 T + ε Where: Y is the average monthly income of redundancies twelve months after the training course T is dummy variable (T=1 if the redundant receives training voucher, and T=0, if not) Hence ß 1 can be considered as the impact of the scheme over income Y. It is expected that ß 1 >0 which means the training voucher scheme helps increase the income of redundancies.
MODELING AND TESTING (4) Similarly the team will assess the employability of participants between the control and treated groups. Depending on the availability of data, the team may assess other outcomes like mean time to find a job, types of job, income sustainability, etc. For testing, the null hypothesis is: H 0 : µ treatment =µ control. T-test will be used to examine the differences between treatment and control groups twelve months after the training course.
MODELING AND TESTING (5) One important assumption for the above regression is that T does not correlate with ε or E(ε|T)=0. Since at this moment it cannot be assured that such assumption can be met, another estimation may be also utilized. Specifically Y = α 0 + α 1 T + γ 1 X 1 +…. + γ n X n + ε Where X 1,…., X n are such variables as initial income, professional skill level, experience, region, gender and so on. This will help identify whether these factors have additional impacts on the income of redundancies. However to ensure the statistical power of the sample, the number of factors will be limited to less than ten.
THE RESEARCH TEAM Dr. Xuan Ba Le MBA Lan Anh Vu MPP Huy The Nguyen MPP Minh Thao Ta
PARTNERS/COLLABORATORS CIEM departments (MPI) State Capital Investment Corporation (MOF) Department of Vocational Training (MOLISA) Provincial Department of Labor, War Invalids and Social Affairs (DOLISA) Training providers SOEs Dr. Tri Thanh Vo