Belo Broadband

24
Broadband in Schools: Good or Bad for Students’ Performance? Rodrigo Belo (CMU, IST-UTL, FCEE-UCP) Pedro Ferreira (IST-UTL) Rahul Telang (CMU) June 28, 2010 (CMU) Broadband in Schools June 28, 2010 1 / 15

Transcript of Belo Broadband

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 1/24

Broadband in Schools: Good or Bad for Students’

Performance?

Rodrigo Belo (CMU, IST-UTL, FCEE-UCP)Pedro Ferreira (IST-UTL)

Rahul Telang (CMU)

June 28, 2010

(CMU) Broadband in Schools June 28, 2010 1 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 2/24

Background and Motivation

Broadband in Schools

Information and Communication Technologies (ICTs) perceived as powerful

tools to improve the quality of education:Widening access to information

Fostering new learning methods (more interaction and feedback)

Increasing students’ interest and performance

(CMU) Broadband in Schools June 28, 2010 2 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 3/24

Background and Motivation

Broadband in Schools

Information and Communication Technologies (ICTs) perceived as powerful

tools to improve the quality of education:Widening access to information

Fostering new learning methods (more interaction and feedback)

Increasing students’ interest and performance

Governments have been investing considerable amounts of money toprovide schools with ICTs, in particular for Broadband connectivity.

Portugal (FCCN)

2006 - Deployment of Broadband Internet connection (ADSL) to all

schools

(CMU) Broadband in Schools June 28, 2010 2 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 4/24

Background and Motivation

Broadband in Schools

Information and Communication Technologies (ICTs) perceived as powerful

tools to improve the quality of education:Widening access to information

Fostering new learning methods (more interaction and feedback)

Increasing students’ interest and performance

Governments have been investing considerable amounts of money toprovide schools with ICTs, in particular for Broadband connectivity.

Portugal (FCCN)

2006 - Deployment of Broadband Internet connection (ADSL) to all

schools

Little empirical evidence that broadband connectivity generates significantbenefits.

(CMU) Broadband in Schools June 28, 2010 2 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 5/24

Background and Motivation Research Question

Research Question

“How does broadband use  aff ect students’ performance?”

Is broadband use beneficial for student performance?

What’s the magnitude of this eff ect?

How does Broadband use aff ect diff erent schools?

(CMU) Broadband in Schools June 28, 2010 3 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 6/24

Research Context

MethodologyUnit of analysis

Unit of analysis: School

(CMU) Broadband in Schools June 28, 2010 4 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 7/24

Research Context

MethodologyUnit of analysis

Unit of analysis: SchoolPerformance - average national exam scores for students in a school.The standardized exams are administered for 9th and 12th graders. Wehave data at the student level. However, we aggregate the scores atschool level.

(CMU) Broadband in Schools June 28, 2010 4 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 8/24

Research Context

MethodologyUnit of analysis

Unit of analysis: SchoolPerformance - average national exam scores for students in a school.The standardized exams are administered for 9th and 12th graders. Wehave data at the student level. However, we aggregate the scores atschool level.Internet use

- We have data on broadband traffi

c for the school. Weuse monthly average per capita broadband consumption as the keymeasure for Internet use.

(CMU) Broadband in Schools June 28, 2010 4 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 9/24

Research Context

MethodologyUnit of analysis

Unit of analysis: SchoolPerformance - average national exam scores for students in a school.The standardized exams are administered for 9th and 12th graders. Wehave data at the student level. However, we aggregate the scores atschool level.Internet use - We have data on broadband traffic for the school. Weuse monthly average per capita broadband consumption as the keymeasure for Internet use.

Period: Academic years 2005-2009

Broadband use started only from 2006. So 2005 provides us a datapoint before the broadband use. Similarly, 2009 provides us data pointafter the broadband became more prevalent.Use of panel data allows us to control for school specific unobservedeff ects.

(CMU) Broadband in Schools June 28, 2010 4 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 10/24

Research Context

MethodologyEconometrics

Performance depends on Internet use, on socio-economical factors, as wellas on unobserved eff ects

∆ pi = ω∆I i +Xiβ + ui (1)

In which ∆ represents diff erences between 2005 and 2008 (or 2009).

(CMU) Broadband in Schools June 28, 2010 5 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 11/24

Data

Internet usage

Total

Download

Upload

   0

   5   0

   1   0   0

   1   5   0

   A  v  e  r  a  g  e   M  o  n   t   h   l

  y   T  r  a   f   f   i  c   /   S   t  u   d  e  n   t   (   M   B   )

   H   i  g   h   −  s  c   h  o  o   l  s

  a  n   d   M   i   d   d   l  e   −  s  c   h  o  o   l  s

2002 2004 2006 2008 2010Year

Figure: Internet usage evolution 2003-2009: download, upload and total trafficper student

(CMU) Broadband in Schools June 28, 2010 6 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 12/24

Data

Internet usage across schools

       0

       5       0

       1       0       0

       1       5       0

       2       0       0

       2       5       0

       F     r     e     q     u     e     n     c     y

0 200 400 600 800 1000INet Usage / Student (MB)

High−schools and Middle−schools

In 2009 the average student used 117 MB per month, which corresponds

to:Watching one hour of YouTube video per month (at 260 Kbps);

Exchanging 390 emails (at 300 KB per email); or

Loading 900 webpages (at 130 KB per page).

(CMU) Broadband in Schools June 28, 2010 7 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 13/24

Data

Exam scores

9th grade

12th grade

   5   0

   5   2

   5

   4

   5   6

   5   8

   6   0

   A  v  e

  r  a  g  e  g  r  a   d  e

2002 2004 2006 2008 2010Year

Figure: Exam score evolution 2002-2008: average exam score over time for 12thand 9th grades (0-100 scale)

(CMU) Broadband in Schools June 28, 2010 8 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 14/24

Endogeneity and Instrument

Endogeneity

While school fixed eff ects should control for across school variations, wemay still worry about unobserved time-varying  eff ects that influence bothexam scores and Internet use and might lead to inconsistent estimates.

Examples: changes in internal organization, technical savviness, or

resources available to a schoolWe may get a positive estimate on broadband use even if there is no causaleff ect.

(CMU) Broadband in Schools June 28, 2010 9 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 15/24

Endogeneity and Instrument

Endogeneity

While school fixed eff ects should control for across school variations, wemay still worry about unobserved time-varying  eff ects that influence bothexam scores and Internet use and might lead to inconsistent estimates.

Examples: changes in internal organization, technical savviness, or

resources available to a schoolWe may get a positive estimate on broadband use even if there is no causaleff ect.Solution: Instrumental Variable (IV)

1 Correlated with change in Internet usage

2 Not correlated with the unobserved eff ect

(CMU) Broadband in Schools June 28, 2010 9 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 16/24

Endogeneity and Instrument

MethodologyInstrumental Variable

(CMU) Broadband in Schools June 28, 2010 10 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 17/24

Endogeneity and Instrument

MethodologyInstrumental Variable

(CMU) Broadband in Schools June 28, 2010 10 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 18/24

Endogeneity and Instrument

MethodologyInstrumental Variable

Distance to the ISP’s Central Office:

1 Correlated with change in Internet usage

2 Not correlated with “change in grade”.

(CMU) Broadband in Schools June 28, 2010 10 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 19/24

Results

ResultsCorrecting for Endogeneity

The IV specification yields a negative coefficient on broadbandusage for the 2005-2008 period.

For 2008, the average broadband use in schools of  51 GB/month in2008 leads to a decrease of  8.9% in the average exam scores, i.e., adecrease of about one standard deviation in 2008 scores.

This eff ect is still negative for the 2005-2009 period, though itbecomes smaller in magnitude and statistically insignificant.

(CMU) Broadband in Schools June 28, 2010 11 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 20/24

Results

Boys and Girls Behave Diff erently

To explore the distraction eff ect in more detail, we use a survey conducted

by ANACOM on the activities performed by students on the Internet (659students; 10-12th):

Table: Participation in activities by gender (%).

Activity Male Female Diff .

Search for Scientific Info 67.5 74.1 -6.6**Email 93.1 89.5 3.7**Chat 89.4 88.2 1.1Radio 48.4 42.5 6.0*TV 27.8 13.9 14.0***Music 75.6 52.7 22.9***

Games 71.9 34.9 36.9****** p<0.01, ** p<0.05, * p<0.1 (t-tests eq. var.)

Boys report to engage in distracting activities more than girls

(CMU) Broadband in Schools June 28, 2010 12 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 21/24

Results

Boys and Girls Behave Diff erently

To explore the distraction eff ect in more detail, we use a survey conducted

by ANACOM on the activities performed by students on the Internet (659students; 10-12th):

Table: Participation in activities by gender (%).

Activity Male Female Diff .

Search for Scientific Info 67.5 74.1 -6.6**Email 93.1 89.5 3.7**Chat 89.4 88.2 1.1Radio 48.4 42.5 6.0*TV 27.8 13.9 14.0***Music 75.6 52.7 22.9***

Games 71.9 34.9 36.9****** p<0.01, ** p<0.05, * p<0.1 (t-tests eq. var.)

Boys report to engage in distracting activities more than girlsIf our story is correct, boys are more likely to be adversely aff ected

(CMU) Broadband in Schools June 28, 2010 12 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 22/24

Results

ResultsBoys and Girls Behave Diff erently

For 2005-2008, the average broadband use in schools of  51GB/month in 2008 leads to a decrease of  12.1% in the average

exam scores of boys, i.e., an decrease of about 1.5 standarddeviations in 2008 scores. The adverse eff ect of Internet use is notstatistically significant for girls.

This eff ect reduces in magnitude and loses significance for both boysand girls in the 2005-2009 diff erence.

(CMU) Broadband in Schools June 28, 2010 13 / 15

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 23/24

Results

ResultsLow Performance vs. High Performance Schools

We split our sample of schools in quartiles based on their 9th grade averageexam score in 2005 and apply our IV setup separately for each group of schools.

Broadband Internet use does not aff ect all schools in the same way:Schools in the lowest performance quartile are more negatively

aff ected by broadband in 2008

Schools might need a certainly level of maturity to be able to eff ectivelycounter the disruptive eff ect that the introduction of broadband Internet

into schools might entail.

(CMU) Broadband in Schools June 28, 2010 14 / 15

C l i

8/8/2019 Belo Broadband

http://slidepdf.com/reader/full/belo-broadband 24/24

Conclusion

Discussion & Policy Implications

Mere provisions of broadband pipes does not seems to generate tangible

benefits

(CMU) Broadband in Schools June 28, 2010 15 / 15