Do apps create jobs? Discussing the evidence

Just when world economies struggle to create jobs, it came as great news this paper (pdf) from the University of Maryland claiming that Facebook apps created 180K jobs in the US.

This is surprising to me. In a study we carried out for DG INFSO earlier this year, it came out clearly that developers dont make much money out of apps. They say that it’s good to be visible towards venture capital, but in terms of money they’re better off selling their services to companies who want their apps, rather than selling them directly (see for example this article and this).  A recent book by the most respected scholar on the economic impact of ICT (Brynjolfsson) paint a very careful and nuanced, if not negative, picture.

I am not an econometrician but the Maryland study looks weak to me. It is 7 pages long, so that it does not seem a study but some back of the envelop analysis.

The main doubt I have is that it seems to calculate as impact of Facebook app any job created by companies that at any point in time have created an app in Facebook. It states “We excluded 13 firms, such as Blizzard Entertainment, Electronic Arts, and Yahoo from our sample because we could not identify the proportion of Facebook induced full‐time employment.” How could they establish the Facebook-induced full time employment for any other company?

The second doubtful statement says: “The number of employees of each firm would be a function of the number of developers and the number of active users” . This appears to link employees to the number of user, as if developing an app for 15 users requires less manpower than an app for 1M users. Why? The very basic competitive advantage of the web is that is allows to scale at very low costs.

Based on this statement, it formulates that the facebook app economy has generated more than 50K direct jobs.

Another puzzling issue is that it mention passingly that only few apps are for pay, and that you pay with “credits”. But then, these credits data are not used in the estimation.

Another important statement is about the multiplier effect:” For the industry “Internet and other information services,” RIMS states that for the most populous state, California, the creation of one job leads to 3.41 additional jobs in California alone.” This is far more important evidence  that the overall app economy study. There is robust evidence that Internet multiplies jobs more than other industries such as communication.

In summary, my impression is that the paper:

  1. calculates the total number of developers in companies that have ever created a facebook app, and attribute this number exclusively to the app
  2. multiplies this number by some factor related to the number of users in order to deduce employees, to arrive at 50K.
  3. It then multiplies this total by the (somewhat robust) multiplier of 3.41 to obtain 180K.

Steps 1 and 2 are strongly flawed in my opinion.

But I am no econometrician, I don’t understand the formulas. So far I have only seen news article about this study, and no serious academic discussion. It seems to me one of the classical cases that with complex formulas and general equilibrium theory we can justify everything.

So here I call all economists:

– can you point me to serious discussion of this study?

– do the calculation make sense?

Do apps create jobs? Discussing the evidence

4 thoughts on “Do apps create jobs? Discussing the evidence

  1. Francesco Mureddu says:

    Econometric Technicalities
    The scope of the study is very narrow, as the authors do not take into account an extensive amount of variables which might influence both the dependent and the explanatory variables. There might also be multicollinearity in the form of a relationship between the MAU (monthly active users) and the number of developers, thereby there might be a bias. And endogeneity as number of employees, monthly active users and developers are correlated.
    In the analysis there are no diagnostic tests whatsoever for endogeneity nor for multicollinearity or for missing variables.

    General Comments

    1) It is difficult to determine how many employees in each firm are really engaged in applications development
    2) The data from the study come from Facebook, so that their reliability is doubtful
    3) The study claims that those considered are newly created job, but it might be that investments in applications simply drained jobs and money from other areas of economy (crowding out), expecially because it might take few hours to develop and application
    4) There well known issues about using multipliers, computed from input-output tables which are of static nature by definition
    5) It is likely that Facebook decreases the productivity of people employed in other industries

  2. David, I don’t think your step 1 is true. The phrase “We excluded 13 firms, such as 
    Blizzard Entertainment, Electronic Arts, and Yahoo from our sample because we could not identify the 
    proportion of Facebook induced full‐time employment” (p. 3) seems to imply that they asked the rest of the hand-collected sample how many developers were doing FB apps; then they built the model to project FB-related employment onto the universe of 8,308 firms.” The problem there is that they don’t specify if the hand-collected sample is a well-behaved random sample from the 8,308 firms: I guess NOT, because the FB dataset was anonymized. If Maryland collected data from companies more deeply involved in the FB apps economy than the average in the FB dataset, the model would overestimate employment. This would almost surely have happened if the researchers had called up companies that they knew were working a lot on FB apps.

    Agree with Francesco on potential issues with multicollinearity and endogeneity.

    Also, I am quite puzzled as why they specified those models. They seem quite ad-hoc. Interestingly, the model does not make use of the data on credits, which would seem to be a better proxy of economic value than visits (after all, that cash pays the developers’s salaries).

  3. Thanks Francesco for the insight.
    Alberto I agree with you. My impression is that the phrase you quote is far too generic to counter my doubts. And I also noticed the credits were mentioned but never used.
    But point 2 for me is what hinders the robustness most.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s