Ping Entrepreneurial Ecosystems

Last registered on July 21, 2021


Trial Information

General Information

Ping Entrepreneurial Ecosystems
Initial registration date
July 19, 2021

Initial registration date is when the trial was registered.

It corresponds to when the registration was submitted to the Registry to be reviewed for publication.

First published
July 21, 2021, 5:39 PM EDT

First published corresponds to when the trial was first made public on the Registry after being reviewed.



Primary Investigator


Other Primary Investigator(s)

PI Affiliation
Max Planck Institute for Innovation and Competition
PI Affiliation
Max Planck Institute for Innovation and Competition

Additional Trial Information

In development
Start date
End date
Secondary IDs
A successful Entrepreneurial Ecosystem (EE) depends on productive regional social capital. Social capital is productive if its structural, relational and cognitive dimensions work together (Nahapiet & Ghoshal 1998). In EEs, these dimensions of social capital are ex- pressed through culture, leadership and social networks (Theodoraki et al. 2018, Stam & van de Ven 2021). In order to understand how social capital can help to foster entrepreneur- ship in specific regions, we propose one empirical study design. We employ a Randomized Messaging Experiment (RME) with approximately 58,000 subjects to assess entrepreneurs’ willingness and sources for motivation to support start-up ecosystems. Our hypotheses are that entrepreneurs are more willing to support their own EE; that this willingness varies with EE; and that being part of a common industry is not a defining factor for the entrepreneurs’ support. The studies aim to contribute to the literature by providing an additional quantitative measure of the quality of an EE and by identifying additional levers for policy to increase start-up activity and economic growth.
External Link(s)

Registration Citation

Defort, Aaron, Dietmar Harhoff and Michael Rose. 2021. "Ping Entrepreneurial Ecosystems." AEA RCT Registry. July 21.
Experimental Details


We send out slightly varied e-mails asking to participate in a survey. The e-mails we aim to send out will have different texts, triggering different feelings of belonging in the participants. Otherwise the mails will be the same mentioning the purpose of the survey, a GDPR claimer and the executing institutions.
The control and treatment interventions are corresponding to an ever growing higher level of relatedness to the subjects in order to isolate the effect. The changes in text should have an impact on the response rate which we interpret as stronger feeling of belonging.
Intervention Start Date
Intervention End Date

Primary Outcomes

Primary Outcomes (end points)
Click rate on the survey link and response rate
Primary Outcomes (explanation)
The click rate will be driven by the e-mail content, which is our primary intervention. Response rate is indication for higher commitment

Secondary Outcomes

Secondary Outcomes (end points)
Responses (length, mail: yes/no)
Secondary Outcomes (explanation)
The subjects can write more or less advice into an open text field. If they write a lot, we assume that they are more committed to the cause of helping the mentioned affiliation.

Experimental Design

Experimental Design
The above-mentioned interventions are allocated in two ways. Since we are interested in geographical entities we are clustering the participants according to their location (cluster trial). As discussed above our experiment has four arms in order to isolate the mechanisms (factorial trial). The interventions are allocated randomly inside of the clusters.
Experimental Design Details
Our hypotheses are that all treatments will result in significant changes in the response rate, which we interpret as propensity to support the mentioned group. We hypothesize the following order: C1<T1. For T2, there are competing assumptions: Inkpen and Tsang (2005) would suggest that in industrial districts there is no common goal, or trust that is not process based, which would result in T2<T1, in that case it will be interesting to also see where in the order of treatment and control T2 will find its place. Homophily and a closer identification with similar actors would suggest that T2>T1.

The clustering of the subjects is done via a DBSCAN algorithm. This procedure is necessary in order to allocate subjects that are residing in suburbs to the greater metropolitan area, e.g. Boston and Cambridge were connected, just as well as Silicon Valley Cities like Palo Alto and Mountain View.

Ex-ante analyses will try to explain differences in treatment effects between ecosystems. Some initial hypotheses are, that the differences in effect size can be explained by: (1) Level of diversity in the ecosystems (they support any member of the ecosystem and not only based on homophily); (2) Amount of flagship ventures in the ecosystems (members are inspired by common narrative that creates feeling of belongingness); (3) The size of ecosystems (unclear boundaries result in less attachment); (4) Availability of talent in the ecosystems (a scarce regional resource leads to industry-unspecific competition)

With the experiment we send a 3-question survey asking of the respondents are founders, for advice for entrepreneurs and if they want to leave an e-mail address for a follow-up.
In addition to the response rate we will also interpret a higher number of words in advice and the leaving of an e-mail address as propensity to support.
Randomization Method
Randomization was done in an office by a computer. For this we used the random module in Python with a seed to ensure replicability.
Randomization Unit
Individual randomization inside of clusters
Was the treatment clustered?

Experiment Characteristics

Sample size: planned number of clusters
ca. 109 ecosystems (statistical clusters of regionally close entrepreneurs); 12,038 entrepreneurs are not clustered but still take part in the experiment.
Sample size: planned number of observations
ca. 59,564 entrepreneurs
Sample size (or number of clusters) by treatment arms
ca. 1/3 of entrepreneurs in each cluster
Minimum detectable effect size for main outcomes (accounting for sample design and clustering)
Power test in Stata below. Assumptions: 1/3 of sample in each treatment group = 19854; Base response rate 3% like in earlier pilot of this study (Setup can be found here: RCT ID: AEARCTR-0007473; target audience was changed to only 1 cluster with 684 entrepreneurs.) Change in response rate of 0.5 p.p. can be detected at 0.8 power. +-----------------------------------------------------------------+ | alpha power N N1 N2 delta p1 p2 | |-----------------------------------------------------------------| | .05 .8 39709 19854 19854 -.00461 .03 .02539 | | .05 .9 39709 19854 19854 -.00531 .03 .02469 | +-----------------------------------------------------------------+ +-----------------------------------------------------------------+ | alpha power N N1 N2 delta p1 p2 | |-----------------------------------------------------------------| | .05 .8 39709 19854 19854 .00499 .03 .03499 | | .05 .9 39709 19854 19854 .0058 .03 .0358 | +-----------------------------------------------------------------+

Institutional Review Boards (IRBs)

IRB Name
Ethikrat der Max-Planck-Gesellschaft
IRB Approval Date
IRB Approval Number
Antrag 2021_23


Post Trial Information

Study Withdrawal

There is information in this trial unavailable to the public. Use the button below to request access.

Request Information


Is the intervention completed?
Data Collection Complete
Data Publication

Data Publication

Is public data available?

Program Files

Program Files
Reports, Papers & Other Materials

Relevant Paper(s)

Reports & Other Materials