Closing the Innovator-Inventor Gap: Evidence from Proactive (Opt-Out) Outreach

We will be studying the innovation engagement of technical and engineering staff at a Silicon Valley high-tech firm, who will receive targeted emails encouraging them to participate in innovation. We will work with the firm to randomly select a set of people from the technical and engineering staff who have never engaged or only engaged a limited amount in innovation practices such as invention disclosure, inventor training, and patent filing. In shorthand, we refer to these people as “never submitters.” The firm will then send one or more targeted emails to this randomly selected subset of "never submitters," inviting them to participate in the patenting process. This email outreach from the firm's patent office will include a link to a short video to help the "never submitter" get started on their innovation journey. Subsequently, the firm will follow business as usual, noting who participates in innovation according to their normal processes.

2024-05-07

2025-04-30

Our primary endpoints are divided into four categories: initial reaction (e.g., email opened), participation in early-stage innovative activities (e.g., learning about the patent submission process through website, attending an IP training, attending a brainstorming session, requesting a meeting with a patent team member, reaching out on Slack, meeting with a patent ambassador (limited to offices with patent ambassadors), sought advice on an inventive idea (only survey), worked on projects likely to yield patentable inventions (only survey)), involvement in later-stage invention activities (e.g., submission of an inventive idea, that idea being filed as a patent application, that idea being abandoned, that idea receiving other IP protection, and the size of the team submitting the idea), and perceptions of self and workplace (e.g., identifying as an inventor, desire to work on tasks leading to being a named inventor, the extent to which teamwork and collaboration are part of the culture and norms are developed around it).

See PAP.

The population will be divided into control and 4 treatment arms. A description of the messages the groups will receive is included in our pre-analysis plan. The treatment arms include:

Control: A group of never-submitters who will not get additional emails through this experiment. They will, however, periodically receive neutral messages without any specific framing or motivational factors from the patent team that are part of the routine course of business.
Arm 1 (personalization + intrinsic motivation)
Arm 2 (personalization + extrinsic motivation)
Arm 3 (generic + intrinsic motivation)
Arm 4 (generic + extrinsic motivation)

Not available

The computer randomized the assignment into groups using the seed number "04232024" and a random uniform distribution. Before implementation, the control and treatment arms were assessed for their covariate balance. The treatment and control groups were statistically indistinguishable for gender, race, citizenship, geographic location, job title, and years working at the firm.

Randomized at the individual level.

No

No clusters

988 control never-submitters, and 3933 treated never-submitters.

The treated never-submitters are grouped into four treatment arms, each with 1002, 945, 1000, and 986 never-submitters, respectively.

For the early-stage involvement in inventive activities, we have data for the precise population under study from our baseline survey. To collect future outcomes, we will have a research assistant record requests for contact based on the patent professionals' emails and correspondences on Slack. The same research assistant will also keep track of participation in brainstorming events and IP training. Since the probability of being assigned to each treatment is equal across observations, our power calculation follows a common format across all treatment-control comparisons. In particular, there is a control group with 988 engineers and a combined treatment group with 3933 engineers. That treatment group is further split into four arms. Arm 1 has 1002, arm 2 has 945, arm 3 has 1000, and arm 4 has 986 observations. Our baseline survey shows that 33\% of never-submitters have done at least one step in the invention process, and the standard deviation is 47\%. Further, we know that the total number of early steps completed is 0.6, and the standard deviation is 1.08. We assume a statistical significance level of 0.05 and a desired power of 0.80. The MDES, based on a two-sided test using our assumed $\alpha = 0.05$ and $\beta = 0.80$, is 6.58 percentage points for the proportion of never-submitters who will complete at least one step in the invention process. By comparison, the MDES for the proportion of never-submitters who will complete at least one step in the invention process for only the extrinsic treatment group (i.e., Arm 2 + Arm 4 = 945 + 986 = 1931) relative to the control group is 8.58 percentage points. The higher MDES is attributable to fewer never-submitters receiving the extrinsic treatment rather than the proactive outreach treatment. When we perform similar calculations for the total number of early steps completed, we see that the MDES for the total number of early steps completed is 0.151 steps for the proactive outreach messaging and 0.196 for the extrinsic arms for the proactive outreach. This means that we can detect a statistically significant average treatment effect for the extrinsic treatment group for the total number of early steps completed when it differs from the control group by 0.196 steps or more. For additional details, please see our PAP.

Pre-analysis Plan for ``Closing the Innovator-Inventor Gap: Evidence from Proactive (Opt-Out) Outreach''