Measuring the relative productivity of multitasking to sole-tasking in household production: experimental evidence

Abstract

The standard household production model does not incorporate multitasking, although time-diary data reveal that individuals regularly multitask. We incorporate multitasking into a household production model in which time spent in childcare can be sole-tasked or multitasked with another household production activity and we present the results of an experiment designed to measure the productivity parameters of this model. Because utility and productivity are intertwined and difficult to disentangle in any household production model, we vary the utility pay-offs our experimental participants receive in order to determine how our estimated productivity parameters are affected by a change in the utility parameters. Our estimates of the relative multitasking productivities indicate that, while a minute of sole-tasked time produces more of a single commodity than a minute of multitasked time, total household output increases when two outputs are produced simultaneously, hence confirming the economic motivation for multitasking.

Keywords:

• multitasking
• time use
• experiments

JEL Classification:

• D10
• C91

Notes

¹ This range, while not reported explicitly by the authors, is the result of our calculations based on Table 3 of their paper.

² One could allow total time available, T, to depend on a market wage, w: . However, the sign of is ambiguous.

³ Our stylized model is intended to capture the main dynamics at play and naturally may not be ‘correct’. However, this is true of any model one attempts to parameterize. We have experimented with different functional forms and do not have any strong reason apart from tractability and simplicity to choose one over the other.

⁴ Data from the second pilot are included in the analysis in this article because neither the design nor the pay-off schemes were changed between that pilot and the main rounds.

⁵ Unfortunately, due to a programming error, participants were informed that their compensation in the baby-care activity would be calculated based on 2-second intervals of the baby’s mood rather than 1-second intervals, the latter of which were used in generating actual pay-offs from the task. This misalignment was the case throughout the experiment. After careful consideration of this problem, and in light of the specific and accurate information about their pay-offs which was fed to our participants as the experiment progressed (see ensuing text), we do not believe that this error had any material consequence for the results of the experiment.

⁶ Real-dollar pay-offs are calculated by using a fixed exchange rate relating experimental dollars to real dollars, where that rate was chosen to result in the average real-dollar payout being roughly equivalent to the ASBLab’s standard payment of $15–$20 per hour of participant time. Pay-offs calculated in this way from the task-performing stages are paid in addition to a $5 show-up fee, which is paid to all participants, regardless of their performance.

⁷ The experiment used software custom-built by Markus Schaffner using the popular software package zTree (Fischbacher, 2007) as a template.

⁸ The sole-tasked baby-care stage is extremely easy to master, as all that is involved is clicking on a button repeatedly when it appears. By contrast, the sole-tasked clothes-sorting task does require that participants learn how to drag and drop the icons, and that they accurately match clothes to the correct piles. Therefore, we only see a sole-tasking learning effect for clothes sorting.

⁹ In a parallel working paper, we find few gender differences in these parameter estimates.

¹⁰ This response of participants to a higher-powered incentive for the household task echoes the finding in much of the industrial organization literature on multitasking, reviewed in the ‘Introduction’ section, that compensating a person for performing one task can crowd-out his performance of another task (Holmstrom and Milgrom, 1991).