Git Product home page Git Product logo

3_cqp's Introduction

README

This repo facilitates the replication of my paper titled "Benefits and Costs of Dual and Informal Apprenticeship in Bénin".

Prerequisites

Stata, SPSS, and R (preferably RStudio) are necessary to fully reproduce paper.

The following default path is used in the scripts below:

"/Volumes/nadel/research/Data/PhDs/Bart 2022/Paper 3 - CQP"

If data is copied into a different directory, path names have to be adjusted where indicated below.

Replication

The paper can be replicated from raw data in three steps, executed in the following order:

  1. open Stata do file "master.do" in the root folder, enter new user path if needed and run
  2. open SPSS syntax "master.sps" in the root folder, enter new user path if needed and run
  3. open R script "master.R" in the root folder, enter new user path if needed and run

The final step generates a PDF called "cnb_apprenticeship.pdf" in the "markdown" folder. This is the replicated paper.

If you have any questions about replicating this paper, please see comments in the scripts above or contact the author at bartlomiej.kudrzycki[at]nadel.ethz.ch

3_cqp's People

Contributors

bkudrzycki avatar elina-smer avatar

Watchers

avatar

3_cqp's Issues

condense literature review

condense to 5-6 pages
page 6-10 (the literature contribution) has to be shortened to max 2 pages

  • part of what is here can go earlier as motivation or later in the discussion
    WHAT IS THE QUESTION THAT COMES OUT OF THIS
    SO WHY DO FIRMS TRAIN

Comment on apprentices being bulk of labor force (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 339 in 6eeb154

If this is the case, why do craftsmen in Bénin train at all? First, apprentices are the cheapest and most abundant source of labor, and can be hired to complete unskilled tasks at little overhead. Second, if trades can be mastered relatively quickly, apprentices will provide skilled labor in the later years of their graduation (and after graduation as well, for as long as they need to work to pay off outstanding training fees or save up for their own workshop). Finally, the apprentices (or, more often, their families) traditionally pay fees to the master craftsmen in return for the provision of training [@velenchikApprenticeshipContractsSmall1995; @frazer2006; @bankole2020]. Aside from the training fee, these include minor fees to cover the provision of equipment and materials, application fees (pertinent for the CQP, as the master trainer must submit paperwork in their apprentices' stead), and initiation and graduation fees. In total, apprentices report paying about 175,000 FCFA (280 \$US) in fees for training, while firm owner report around 160,000 FCFA (255 \$US) in fees per apprentice. This indicates an minor increase in the costs of training in Benin over the past two decades: @walther2007 reports total fees ranging from 50,000 to 150,000 FCFA (96-290 \$US, inflation adjusted). Though generally unregulated, in some cases professional associations and public authorities step in to regulate fees, particularly those levied for initiation and graduation ceremonies.

Tie to theory if possible.

regression analysis of attrition

  1. add regression analysis - "you should run a logit regression on ? characteristics and put "1" for not found anymore -> anything significant?"
  2. "how do you know that apprentices that were not found again were not predominantly leavers?" - no reason to suspect that they were less likely to answer interview requests (we had their phones). apprenticeships generally not too far from parents

Clarify intro to firm benefits section

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 355 in be6645b

If this is the case, why do craftsmen in Bénin train at all? First, apprentices are the cheapest and most abundant source of labor, and can be hired to complete unskilled tasks at little overhead. Second, if trades can be mastered relatively quickly, apprentices will provide skilled labor in the later years of their graduation (and after graduation as well, for as long as they need to work to pay off outstanding training fees or save up for their own workshop). Finally, the apprentices (or, more often, their families) traditionally pay fees to the master craftsmen in return for the provision of training [@velenchikApprenticeshipContractsSmall1995; @frazer2006; @bankole2020].

  • what do we know about fees in Africa and Benin (reference Rubain)
  • what do we know about fees in Africa vs. Europe?

table 3 changes

  1. add notes on how index was calculated. based on how many questions and scale (0-1)
  2. change "group" header to "Measure" or "indicator"
  3. (maybe) add diff column if it fits (probably not)

Explain "External training" (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 276 in 6eeb154

Though dual training is predicated on classroom teaching about once a week, we found that external training (classes or training that took place outside of their master's workshop) was not limited to CQP participants. At baseline, `r mean(as.numeric(df$YS4.43), na.rm = T)*100`% of apprentices reported participating in such training in the preceding three months. However, only `r mean(as.numeric((df %>% filter(SELECTED == 1))$YS4.43), na.rm = T)*100`% of CQP participants reported doing so, despite external training being a constituent and necessary component of the CQP program's dual training structure. Interruptions due to the Covid-19 pandemic may have played an important role in extending the duration of training in the period of observation: `r prop.table(table(df$COV15))[1]*100`% of apprentices reported reduced hours of training, while `r prop.table(table(df$COV15))[2]*100`% reported complete work stoppages at their training firm, due to the pandemic. Covid-19 is also likely to have had an outsized impact on participants in the CQP program, `r prop.table(table(df$COV15))[3]*100`% of whom reported that their training center had suspended classes. Among apprentices who participated in any external training, `r sum(df$YS4.44[!is.na(df$YS4.44)]>6)/length(df$YS4.44[!is.na(df$YS4.44)])*100`% reported working spending at least 10 days in this training in the preceding three months (approximately equal for the CQP subsample), while the average reported training duration was `r mean(df$YS4.46, na.rm = T)` hours. Thus, CQP apprentices reported training externally at only a marginally higher rate than apprentices who applied but were not accepted into the program.

External training is not narrowly defined
BUT
We would expect more CQPs to be doing it relative to others. Also cite Rubain's new papers regarding "bunching", Gnahoui and Davodoun perhaps as well

table 5 changes

  1. remove endline
  2. add results for models III and IV
  3. add note: "is this per year, month, or totalled over 4 years?"

"knowledge questions too easy"

address in results for individuals

  • why not. also, there were very few, and some were not entirely trade specific, in contrast to tasks

Clean up firm descriptive statistic section

  1. "Would they still have to do an additional three years" if the sign up for CQP? answer: yes (not sure where to confirm, seems obvious. check CQP documentation maybe).
  2. "huge drop off in apprentices - WHY" -> unable to track down after multiple attempts. not sure why
  3. selection by firm. how to show that mix of selected/not selected/dna is heterogenous by firm?
  4. "what are the numbers for example in small workshops in switzerland" - re: number of apprentices in the average firm

Move details of productivity estimates to Appendix

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 474 in b1b717e

Second, we add to Model I an additional estimated cost of training: foregone trainer productivity. Firms estimated the hours trained on the last day the firm stopped all productive activities to train apprentices, as well as the number of days per week that such training occurred. We use this information as the basis for our estimation. We extrapolate weekly hours of training (hours trained on previous day of training x days trained per week) to annual hours by assuming four work weeks per month and multiplying by the number of months the firm owner reported being open in the previous year. As when estimating apprentice productivity, we set reported monthly wages for skilled workers (wage employees who had trained with the current firm) equal to trainer productivity and divide by the approximate hours worked in the past month to arrive at approximate hourly wage per trainer (assuming four work weeks per month and using firm-reported days open last week and hours worked on the last day). Finally, we multiply by the number of trainers and divide by the number apprentices per firm to arrive at an estimated cost, per apprentice, in terms of total foregone employee productivity resulting from training activities.

table 4 changes

  • 1. remove endline

  • 2. stat. significance between apprentice and firm

  • 3. difference between CQP and non-CQP

  • 4. note: shown are total fees paid for entirety of apprenticeship (typically four years)

overall selected non selected dna
fee apprentice|firm apprentice|firm apprentice|firm apprentice|firm

Focus on pooled analysis (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 314 in 6eeb154

In the above analysis, and in Table \@ref(tab:skills) in the Appendix, scores are pooled at the trade level in each wave. Alternatively, we can observe the change in scores for individuals before taking means. Individual changes in score averaged across trades are shown in Figure \@ref(fig:improvement2). Viewed thus, improvements in competence and experience scores are much more pronounced, with apprentices in masonry doubling their experience and nearly doubling their competence (on the other hand, improvements for electrical installation appear lower compared to the case where data is pooled). However, due to attrition between the two waves, this subgroup analysis suffers from small group size, with $N=34$ for masonry and $N=20$ for carpentry.

Move figure 2 to appendix.

Add permutation table for apprentice productivity estimates (appendix)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 472 in b1b717e

In a second approach, we keep all components of Model I and add two additional factors. First, we estimate an additional benefit of training to the firm: apprentices' net productive value to the firm. In the competitive model of labor markets (with heterogeneous wages), workers are paid their marginal productivity. We assume competitive labor markets and use detailed wage information elicited from each firm to estimate the total productive output of apprentices. Namely, we assume apprentice productivity is equal to that of an untrained employee with no more than a primary education for the first two years of training, and increases to that of trained employee (who had trained at the training firm) for the final two years. Under these assumptions, the annual productive value generated by apprentice work amounts to the average of these two wages\footnote{A popular alternative to this admittedly unpolished approach involves eliciting specific tasks performed by apprentices and estimating costs savings based on the wages paid to workers who would otherwise be responsible for said tasks [@hauschildt2018]. Our firm-apprentice data did not cover specific tasks and is thus not equipped to carry out such an analysis.}.

Combine Table 1 and Table 2 (Isa)

"Table 2 should be integrated into table 1. This is really interesting and is more data on what is different between firms and trainings in low-income settings. -> first section of the results section"

Alternative scenarios

One add. model with only fees and allowances
One add. model II without trainer productivity

Why "first two years" (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 472 in b1b717e

In a second approach, we keep all components of Model I and add two additional factors. First, we estimate an additional benefit of training to the firm: apprentices' net productive value to the firm. In the competitive model of labor markets (with heterogeneous wages), workers are paid their marginal productivity. We assume competitive labor markets and use detailed wage information elicited from each firm to estimate the total productive output of apprentices. Namely, we assume apprentice productivity is equal to that of an untrained employee with no more than a primary education for the first two years of training, and increases to that of trained employee (who had trained at the training firm) for the final two years. Under these assumptions, the annual productive value generated by apprentice work amounts to the average of these two wages\footnote{A popular alternative to this admittedly unpolished approach involves eliciting specific tasks performed by apprentices and estimating costs savings based on the wages paid to workers who would otherwise be responsible for said tasks [@hauschildt2018]. Our firm-apprentice data did not cover specific tasks and is thus not equipped to carry out such an analysis.}.

Perhaps reference Bolli paper with increasing productivity? Otherwise idk, it's just an assumption

Explain how not all apprentices were re-referenced in endline (Laura)

3_cqp/markdown/cnb_apprenticeship.md

Line 200 in 2306971

Summary statistics for the apprentice sample are shown in Table \@ref(tab:appdesc). Data on 432 apprentices working for 199 unique firms was collected at baseline. Of these firms, 155 were available for an endline survey. Data was collected for only 245 apprentices at endline; this drop-off was partially by design, in order to limit the duration of the interview with the firm owners (who were asked detailed questions about individual apprentices).

GENERAL: Explicit research question (Laura + Isa)

  • "The research gap is well motivated but the research question could be more explicitly stated; it’s not clear to me what exactly your focusing on (see also my last bullet on the link between the CB and the reg analysis)" (L)
  • "big picture" is still missing a bit

Clean up comparison to Euro context (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 621 in b1b717e

Finally, how do these numbers compare to evidence from high-income countries? Due to the disparities in firm size and productivity of informal firms in Bénin and training firms in Germany and Switzerland, where the majority of studies have been conducted, an informative comparison is difficult to make. Though cost-benefit studies from these countries do not report earnings data of training firms, they do suggest that whether firms recoup the costs of training depends on the firm, trade, and region. For instance, @hauschildt2018 reports that in Germany, productive contributions of apprentices covered about 70 percent of a company's training costs, while in Switzerland, net benefits per apprentice per year amounted to about \euro 2,500 [@strupler2012]. Without a clearer understanding of future benefits and better instruments to measure apprentice productivity, the true benefit of apprenticeship training in informal firms will remain ambiguous.

  • how do Bolli et al do this, if at all?
  • probably best to move entirely to discussion

changes to intro

% in TVET and universities
dropouts in TVET compared to finished secondary
"what about leaving for own business?" - Isa
"how many countries have national TVET systems?"
"no more dual systems [than the ones listed on page 5]?"

"Due to attrition, this subgroup analysis suffers from small group size"

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 401 in 775d267

In Table \@ref(tab:skills) above, scores are pooled at the trade level in each wave. Alternatively, we can observe the change in scores for individuals before taking means. Individual changes in score averaged across trades are shown in Figure \@ref(fig:improvement2) in the Appendix. Viewed thus, improvements in competence and experience scores are much more pronounced, with apprentices in masonry doubling their experience and nearly doubling their competence (on the other hand, improvements for electrical installation appear lower compared to the case where data is pooled). However, due to attrition between the two waves, this subgroup analysis suffers from small group size, with $N=34$ for masonry and $N=20$ for carpentry.

-> "what does that imply"

Cite papers that model apprentice productivity

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 472 in 2a1f299

In a second approach, we keep all components of Model I and add two additional factors. First, we estimate an additional benefit of training to the firm: apprentices' net productive value to the firm. In the competitive model of labor markets (with heterogeneous wages), workers are paid their marginal productivity. We assume competitive labor markets and use detailed wage information elicited from each firm to estimate the total productive output of apprentices. Namely, we assume apprentice productivity is equal to that of an untrained employee with no more than a primary education for the first two years of training, and increases to that of trained employee (who had trained at the training firm) for the final two years. Under these assumptions, the annual productive value generated by apprentice work amounts to the average of these two wages\footnote{A popular alternative to this admittedly unpolished approach involves eliciting specific tasks performed by apprentices and estimating costs savings based on the wages paid to workers who would otherwise be responsible for said tasks [@hauschildt2018]. Our firm-apprentice data did not cover specific tasks and is thus not equipped to carry out such an analysis.}.

  • how do they do it? (one sentence, focus on low-income)

Add attrition table (Isa)

3_cqp/markdown/cnb_apprenticeship.md

Line 200 in 2306971

Summary statistics for the apprentice sample are shown in Table \@ref(tab:appdesc). Data on 432 apprentices working for 199 unique firms was collected at baseline. Of these firms, 155 were available for an endline survey. Data was collected for only 245 apprentices at endline; this drop-off was partially by design, in order to limit the duration of the interview with the firm owners (who were asked detailed questions about individual apprentices).

Change fee means to in-line code (just round)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 409 in be6645b

Apprentices report significantly higher fees than firms, for this specific fee in particular. Firms may underreport fees to avoid accusations of gauging, but are at the same time likely to have more direct knowledge of all fees than apprentices, whose parents and relatives usually pay the craftsmen directly. Finally, firms report collecting higher initiation and application fees at the time of the endline survey. This may indicate a shift to fee payments to the beginning of the apprenticeship as graduation ceremonies (and the concomitant graduation fees) are phased out as legislation prohibiting graduation ceremonies is put into practice over time.

Clarify if baseline endline or average (Laura)

Why self-evaluate at endline only? (Laura)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 209 in 6eeb154

Second, **competency questions** refer to sets of trade-specific tasks which were collected directly from practicing craftsmen in field interviews. Each competency is evaluated on a binary scale for each apprentice: an apprentice is either considered capable of carrying out a given task, or not. Apprentice competency was evaluated by master trainers at both baseline and endline, while apprentices were asked to self-evaluate at endline only. Between 10 and 15 competency questions were used per trade.

make all tables chunks

-rename chunks
-perhaps organize external files as follows:
main body tables
appendix tables
appendix figures

Add table detailing difference between app, not selected, did not apply (Laura)

3_cqp/markdown/cnb_apprenticeship.md

Line 200 in 2306971

Summary statistics for the apprentice sample are shown in Table \@ref(tab:appdesc). Data on 432 apprentices working for 199 unique firms was collected at baseline. Of these firms, 155 were available for an endline survey. Data was collected for only 245 apprentices at endline; this drop-off was partially by design, in order to limit the duration of the interview with the firm owners (who were asked detailed questions about individual apprentices).

Add permutation table for foregone productivity (appendix)

3_cqp/markdown/cnb_apprenticeship.Rmd

Line 474 in b1b717e

Second, we add to Model I an additional estimated cost of training: foregone trainer productivity. Firms estimated the hours trained on the last day the firm stopped all productive activities to train apprentices, as well as the number of days per week that such training occurred. We use this information as the basis for our estimation. We extrapolate weekly hours of training (hours trained on previous day of training x days trained per week) to annual hours by assuming four work weeks per month and multiplying by the number of months the firm owner reported being open in the previous year. As when estimating apprentice productivity, we set reported monthly wages for skilled workers (wage employees who had trained with the current firm) equal to trainer productivity and divide by the approximate hours worked in the past month to arrive at approximate hourly wage per trainer (assuming four work weeks per month and using firm-reported days open last week and hours worked on the last day). Finally, we multiply by the number of trainers and divide by the number apprentices per firm to arrive at an estimated cost, per apprentice, in terms of total foregone employee productivity resulting from training activities.

Add permutation table for firm costs

"how can firms distinguish what goes into training and what not?"
"but training equipment might bought once a year (save for books!)"

  • the owners were asked to estimate MONTHLY expenditures on each item (i.e. last month). i.e. maybe not accurate for individual owners, but should be decently reliable averaged over the sample (no?). we can try different assumptions about frequency of expenditures on equipment; books.

Recommend Projects

  • React photo React

    A declarative, efficient, and flexible JavaScript library for building user interfaces.

  • Vue.js photo Vue.js

    🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.

  • Typescript photo Typescript

    TypeScript is a superset of JavaScript that compiles to clean JavaScript output.

  • TensorFlow photo TensorFlow

    An Open Source Machine Learning Framework for Everyone

  • Django photo Django

    The Web framework for perfectionists with deadlines.

  • D3 photo D3

    Bring data to life with SVG, Canvas and HTML. 📊📈🎉

Recommend Topics

  • javascript

    JavaScript (JS) is a lightweight interpreted programming language with first-class functions.

  • web

    Some thing interesting about web. New door for the world.

  • server

    A server is a program made to process requests and deliver data to clients.

  • Machine learning

    Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.

  • Game

    Some thing interesting about game, make everyone happy.

Recommend Org

  • Facebook photo Facebook

    We are working to build community through open source technology. NB: members must have two-factor auth.

  • Microsoft photo Microsoft

    Open source projects and samples from Microsoft.

  • Google photo Google

    Google ❤️ Open Source for everyone.

  • D3 photo D3

    Data-Driven Documents codes.