Do More Hours Equal More Learning? Probably, But It Isn't Cheap

In many countries around the world, universal primary school enrollment has been achieved. But quality remains an ongoing challenge. How do you get students to learn more? One solution that comes up often in Latin America and the Caribbean is to increase the length of the school day. From Mais Educação in Brazil to Jornada Escolar Completa in Chile, many governments are considering or are already rolling out additional hours to the school day.

Will more hours help?
There are a range of constraints to the quality of education, and hours of schooling aren’t necessarily the first to jump to mind. But in fact, before the roll-out of extended day programs, students were spending very little time in school. According to Alfaro & Holland (2012), students in Río Negro, Argentina, were in school just four hours a day before the roll-out of Jornada Escolar Extendida beginning in 2006. In São Paulo, Brazil, the length of the school day was four hours at about the same time. Ultimately, this can translate into far fewer hours of instruction over the course of a year. For example, across countries participating in the international student assessment PISA, the annual compulsory instructional time ranged from under 600 hours in Poland to more than 1,000 in Italy, as you can see in the figure below.


Source: Holland, Alfaro, & Evans (2015), using 2009 PISA data

One can imagine a number of ways in which additional hours of schooling could potentially translate into improved student learning: More school time could mean more productive time learning reading, math, and other academic subjects. More school time could make it easier for parents to work, increase income, and invest in complementary educational resources.

At the same time, if the quality of instruction is low or if additional hours are used for recreation rather than learning, then additional hours may not actually translate into improved learning. (In fact, one can imagine a scenario where extra-curricular activities in the longer school day substitute for home study and actually result in a negative impact.) As the figure above also shows, there is no simple correlation between hours of instructional time and reading scores. Now, that doesn’t necessarily mean there isn’t a link: There are many factors determining both decisions about instructional time and reading scores, so a simple cross-country correlation isn’t the best way to figure out whether more hours of schooling are effective.

What do the studies show for educational impacts?
The clearest way to calculate the impact of extending the school day would probably be a randomized trial within a country. In the absence of that, researchers have sought to take advantage of geographical variation in the rollout of these programs. Because this is a priority issue in many Latin American and Caribbean countries, we carried out a review of local evidence, available here. We identified 11 studies that show the impact of extending the school day on learning outcomes, 4 studies (with some overlap!) with other educational outcomes such as grade promotion, and 4 studies with impacts on other outcomes. Of the 11 with learning outcomes, 8 are propensity score matching studies, and the others use some variation on difference in differences. Most of the studies find positive, statistically significant effects, although at least two find negative and significant effects on test scores. At the same time, the size of the effects is often small. An evaluation of a full-time school program in Uruguay yielded gains of 0.07 standard deviations in math and 0.04 in language ( Cerdan-Infantes & Vermeersch). An evaluation in Brazil by De Aquino found no significant impacts on math, and just a few points different in language on a national test ( De Aquino). Another evaluation in Brazil found impacts of 0.05 for language and 0.06 for math (in fourth grade; no effects in eighth grade - Xerxenevsky). And one evaluation in Chile found no impact in math and an impact of just 3 points (out of about 230) on a language test in the public school sample ( Garcia). There are a couple of exceptions, but you get the idea: Even though there is a tendency toward significant impacts, they tend to be small.

What about other impacts?
Only a handful of studies have looked at other impacts. Pires & Urzua find a negative impact on adolescent motherhood and high school dropout in Chile. Likewise, Berthelon & Kruger examine Chile (as well!) and find the same negative impact on adolescent motherhood, as well as a reduction in municipal crime rates. One study examines the impact on women’s labor force participation: Contreras, Sepúlveda, & Cabrera find a positive significant impact of full-day schooling in Chile on whether women participate in the labor force (the external margin), even though the average hours drops (the internal margin).

Is it worth it?
We collected data on the cost of expanding the school day in Uruguay and then combined that with the impact estimates from Cerdan-Infantes & Vermeersch. We then used the measures of cost-effectiveness from a bunch of other educational interventions, from Evans & Popova. You can see the result in the figure below. (Actually, you can only barely see the result: Expanding the school day is so much less cost effective than any of the other interventions that it barely registered on the graph.) That's done using exchange rates, which is what Dhaliwal et al. recommend in their cost effectiveness analysis. If we use purchasing power parity numbers, Uruguay goes from 1/10 as cost-effective as the next intervention to 1/5 as cost-effective. So it doesn't change the story.

Source: Holland, Alfaro, & Evans (2015) calculate cost-effectiveness for Uruguay. The rest of the figure is adapted from Evans & Popova (2014), using data from J-PAL (2014). 

We aren’t the first to notice this. Cerdán-Infantes & Vermeersch, without doing a formal cost analysis, note their small estimated impacts and say, “Even though the program appears to have yielded positive effects on learning outcomes of participating schools, the full-time school program is expensive.” Likewise, Valenzuela – looking at Chile’s experience – notes that “the small magnitude of the effects raises doubts about its cost-effectiveness.”

Another key consideration is that – unlike a remedial tutoring program or an information provision program – expanding the school day is much more likely to be politically irreversible.

This might be where cost-effectiveness breaks down. Cost-effectiveness measures the additional benefit per $100 (of some other amount). It requires a single outcome, and it’s really useful for comparing programs with one principal goal (such as improving student learning). Cost-benefit analysis requires monetization of diverse benefits (student learning, increased maternal employment, reduced crime); ideally, it permits comparison of more diverse programs, but it also requires strong assumptions.

If the argument is that governments should expand the school day, not for educational purposes (because clearly there are more cost-effective ways to do that), but for female labor force participation purposes or reduced juvenile crime reasons, then one has to ask, Do we really believe this is the most effective and most cost-effective way to achieve those outcomes?

Longer schools days have their benefits; but better teaching or other interventions may do much more for kids and cost much less to governments.