Skip to main content
Have a personal or library account? Click to login
Directing young dropouts via SMS: evidence from a field experiment Cover

Directing young dropouts via SMS: evidence from a field experiment

Open Access
|Feb 2022

Figures & Tables

Table 1

Descriptive statistics of all youths and dropouts during army days

Characteristics% of all youths% of all dropouts
(1)(2)
Gender (male)51.1161.15
Age
  16–17 years old95.5575.24
  18–21 years old3.8521.14
  22–25 years old0.603.62
School
  Lower secondary83.9599.77
  Vocational upper secondary10.580.21
  General upper secondary5.050.02
  Postsecondary0.420.00
Literacy
  Level A88.4464.72
  Level B2.9913.23
  Level C1.875.62
  Level D2.668.56
  Level E3.277.04
Directed
  Toward any partner public institution11.7963.49
  Toward ML agencies2.1932.46
Total number of observations5,514,495237,110

[i] Notes: This table reports descriptive statistics about some characteristics of youths and dropouts during army days. “Age” is age at the army day. The category “School” for dropouts corresponds to the level at which youths drop out of the school system. “Level A” for “Literacy” corresponds to “normal literacy”, while “Level E” corresponds to “illiteracy” and Levels B–D range for decreasing levels of medium literacy. Partner public institutions of army days include Établissement pour l’insertion dans l’emploi (EPIDE), Service militaire adapté (SMA), Centres d’informations et d’orientation (CIO), Savoirs pour réussir (SPR), and the MLs.

[ii] Source: SAGA 2013–2019 database, author calculations.

[iii] MLs, missions locales; SAGA, Système d'aide à la gestion des administrés.

Table 2

Control and treatment groups

GroupName
ControlNo SMS
Treatment 1SMS – formal style
HELLO {YOUTH FIRSTNAME}, THE {AGENCY NAME} ADVISES YOUTHS ON THEIR PROJECTS. MORE INFORMATION AT {AGENCY ADDRESS}. THE 1ST MEETING DOES NOT REQUIRE AN APPOINTMENT.
Treatment 2SMS – informal style
HEY {YOUTH FIRSTNAME}, THE {AGENCY NAME} ADVISES YOUTHS ON THEIR PROJECTS. + <SPECIFIC INFO> +. MORE INFORMATION AT {AGENCY ADDRESS}. THE 1ST MEETING DOES NOT REQUIRE AN APPOINTMENT! ☺

[i] Notes: This table reports the different groups in which young dropouts were allocated during the experiment and the content of the text they received. Elements in curly brackets are variables that changed according to individual name and location.

Table 3

ITT effects

OLS estimatesEntry in agency (0/1)
(1)(2)(3)(4)(5)(6)
Panel A: impact of SMS
SMS – all−0.0097 (0.0143)−0.0100 (0.0144)−0.0092 (0.0152)−0.0090 (0.0154)−0.0091 (0.0160)−0.0093 (0.0162)
Constant (reference: no SMS)0.1864*** (0.0146)0.1866*** (0.0146)0.1860*** (0.0159)0.1858*** (0.0160)0.1859*** (0.0166)0.1860*** (0.0130)
Panel B: impact of SMS according to language style
SMS – formal−0.0003 (0.0165)−0.0000 (0.0168)0.0016 (0.0170)0.0020 (0.0174)0.0026 (0.0179)0.0019 (0.0181)
SMS – informal−0.0129 (0.0143)−0.0134 (0.0143)−0.0129 (0.0153)−0.0128 (0.0155)−0.0131 (0.0161)−0.0131 (0.0163)
Constant (reference: no SMS)0.1864*** (0.0146)0.1866*** (0.0146)0.1860*** (0.0159)0.1859*** (0.0160)0.1859*** (0.0166)0.1861*** (0.0130)
βSMS Formal = βSMS Informal0.15990.13140.10990.11070.09570.1020
N4,1034,1034,1034,1034,1034,103
Displayed informationNoYesYesYesYesYes
Individual characteristicsNoNoYesYesYesYes
Agency characteristicsNoNoNoYesYesYes
Location characteristicsNoNoNoNoYesYes
Month fixed effectsNoNoNoNoNoYes

Notes: This table reports OLS estimates, where the dependent variable is a dummy variable equal to one if the individual went to an ML agency after its army day, zero otherwise. “SMS – X” are dummy variables equal to one if the individual received a specific treatment SMS, zero otherwise. Displayed information corresponds to variables that might have been displayed in the different treatment texts as the distance (in kilometers) to the agency and the number of youths enrolled in the agency on the month before the army day. Individual characteristics include dummies for gender, birthplace, age at the army day, literacy level, and region of residency. Agency characteristics include dummies for the number of agencies, number of committee rooms, number of points of contacts, number of firms in portfolio, number of caseworkers, mean age of caseworkers, share of male caseworkers, and average number of caseloads per caseworker. Location characteristics include dummies for disadvantaged area, type of city, local unemployment rate, the number of services, number of stores, number of schools, number of transport modes, and number of leisure facilities. All the control variables are introduced as dummy variables that have been demeaned so as to leave the constant mostly unchanged, as suggested in Athey and Imbens (2017). The estimations include 4,103 observations instead of the full sample of 4,457 because some agency and location characteristics are missing for about 400 observations. Estimates are equivalent whether these individuals are included or not. Robust standard errors are clustered at the month of the army day level and are reported below the coefficients in parentheses. The line βFormal = βInformal reports the p-values associated with a Student test of equality of the SMS estimates.

*** Significant at 1%.

ITT, intention-to-treat; ML, mission locale; OLS, ordinary least squares.

Language: English
Accepted on: Jan 11, 2022
Published on: Feb 21, 2022
Published by: Sciendo
In partnership with: Paradigm Publishing Services
JEL:

© 2022 Jérémy Hervelin, published by Sciendo
This work is licensed under the Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License.