Export Facilitation through Improving Management

A treatment group of 100 firms will receive a diagnostic, 190 hours of technical assistance, and then participate in a trade fair, while a control group of 100 firms will receive the diagnostic and trade fair only

2018-08-01

2019-06-28

Primary Hypothesis: The program will lead more firms to export, diversity the range of products exported and destinations exported to, and improve the export performance of participating firms.
This primary hypothesis will be measured through the following primary outcomes, all obtained from administrative data on export performance, where past year will denote the year from the start of the intervention phase onwards (anticipated to be the period August 2018-July 2019):
1. Extensive margin: Export at all in the past year: This is a binary variable, defined as one if the firm exports directly at all in the one year period since the intervention begins, and zero otherwise.
2. Number of Distinct Products Exported in the past year: The number of different product categories exported in the past year, using the 6-digit product classification in the harmonized system for the Andean Community. This will be coded as zero for firms that do not export, and will be winsorized at the 99th percentile.
3. Number of Different Countries Exported to in the past year. The number of different countries the firm exported to in the past year, coded as zero for firms that do not export, and winsorized at the 99th percentile.
4. Number of Distinct Product-Country Combinations Exported in the past year: This counts the number of product-country combinations a firm exported to in the past year, coded as zero for firms that do not export, and winsorized at the 99th percentile.
5. Export innovation (new product-country combination): This is a binary variable coded as one if the firm exported to a product-country pair that they had not exported to at all in the past three years, and zero otherwise. Coded as zero for firms that do not export.
6. Inverse Hyperbolic Sine of Total Export Value in the past year. This takes the inverse hyperbolic sine transformation (log(y+(y2+1)1/2) of total exports (measured in US dollars), and is coded as zero for firms that do not export.
7. Inverse Hyperbolic Since of Export Labor Productivity: This is the inverse hyperbolic sine of the ratio of total export value in the past year (measured in US dollars) to the average number of workers used in the past year (obtained from the PILA database, which has monthly data on formal workers). This is coded as zero for firms that do not export, and winsorized at the 99th percentile.
8. A standardized export outcomes index: This index will be calculated as the average of the normalized z-scores of outcomes 1 through 7, where each z-score is defined by subtracting the mean and dividing by the standard deviation of the respective outcome.

1. Proportion of Export-Specific Management Practices Being Used
2. Proportion of General Management Practices Being Used
1. Energy efficiency
2. Production costs index
3. Labor productivity
4. Total factor productivity
1. Employment:
2. Inverse hyperbolic sine of annual revenue
3. Inverse hyperbolic sine of annual profits

See registered report/pre-analysis plan

The sample consists of the 200 firms that applied for the program, which are then randomly assigned to two groups of 100 firms each.

These 200 firms were randomly assigned to two groups of 100 each on April 11, 2018. The application form data were used to stratify firms by size (small, medium, or large) , and whether or not the firms had exported at all in the last 3 years. An additional two strata were added: one stratum of 19 export outlier firms (defined in terms of having export values, the number of destinations exported to, or the number of different products exported above the 95th percentile in the self-reported export data on the application form), and one stratum of 1 firm that was missing firm size information. We then formed an index of the proportion of 11 exporting management practices (defined in Appendix A) that firms were using. Within each of the eight strata, we then ranked firms by this export practices index, and formed quadruplets, with two firms from each quadruplet assigned to control (benefits 1) and two firms to treatment (benefits 2). In total this gives us 54 strata defined by these export practice quadruplets inside the eight original strata.
The program website, and the description given to firms made clear that the program had two benefit schemes (benefits 1 and benefits 2), and that firms would be randomly allocated by the World Bank to one or another using a process that guaranteed transparency and equality of opportunities for selection. Random assignment was carried out by two of the authors using Stata, and was livestreamed to both members of the Programa de Transformación Productiva, as well as to applicant firms.

Public computer randomization (see above)

Firm

No

200 firms

200 firms

100 control firms, 100 treatment firms
Time series data: monthly data on outcomes for first year after intervention.

See pre-analysis plan/registered report

Registered Report Draft/Pre-analysis Plan

Registered Report Draft/Pre-analysis Plan

Policymakers often test expensive new programs on relatively small samples.
Formally incorporating informative Bayesian priors into impact evaluation offers the
promise to learn more from these experiments. We evaluate a Colombian program
for 200 firms which aimed to increase exporting. Priors were elicited from academics,
policymakers, and firms. Contrary to these priors, frequentist estimation can not
reject null effects in 2019, and finds some negative impacts in 2020. For binary
outcomes like whether firms export, frequentist estimates are relatively precise,
and Bayesian credible posterior intervals update to overlap almost completely with
standard confidence intervals. For outcomes like increasing export variety, where
the priors align with the data, the value of these priors is seen in posterior intervals
that are considerably narrower than frequentist confidence intervals. Finally, for
noisy outcomes like export value, posterior intervals show almost no updating from
the priors, highlighting how uninformative the data are about such outcomes.

Iacovone,Leonardo; Mckenzie,David J.; Meager,Rachael. Bayesian Impact Evaluation with Informative Priors : An Application to a Colombian Management and Export Improvement Program (English). Policy Research working paper ; no. WPS 10274; Impact Evaluation series Washington, D.C. : World Bank Group.

http://documents.worldbank.org/curated/en/099807301092338354/IDU08080a50008e3c047f40a8de0e4e6b95c2d2c