This section clarifies the analytical role of each variable used in the evaluation of Google and Microsoft corporate accelerator programs using the database compiled from Crunchbase. The purpose here is to explain how each variable contributes to the structure of the statistical testing, what their limitations are, and why they were chosen. The section also outlines the logic behind their inclusion from the ANOVA (analysis of variance) testing and binary comparisons that form the empirical core of the dissertation.
This post represents a series of articles related to a research and dissertation called “Are corporate accelerators springboards for startups: a performance analysis of the Microsoft’s and Google’s accelerated.
Why Variables Were Chosen
The selected variables represent key startup performance signals based on both academic literature and practical outcomes measurable within the scope of a corporate accelerator’s influence. These include:
- Funding, amounts in five periods of time to assess financial growth.
- IPO events, as indicators of public market readiness.
- Acquisitions and acquisitions made, to capture exit strategies and strategic capabilities.
- Closure status, as a measure of business failure or unfeasibility.
- Patents, to reflect pre-acceleration innovation capacity and possible selection bias.
These variables align directly with the research objectives of understanding whether corporate accelerators contribute to growth, exit-to-market readiness, or early termination.
Variable Function Mapping
The table below categorizes the analytical function of the most relevant variables:
| Variable | Analytical Role | Used In |
|---|---|---|
| Funds raised before acceleration | Funding baseline | ANOVA |
| Funding at acceleration date | Initial corporate investment impact | ANOVA |
| Funding in Years 1, 2, and 3 | Growth trajectory measurement | ANOVA |
| IPO | Binary performance indicator (1/0) | Descriptive tables |
| Acquired | Binary performance indicator (1/0) | Descriptive tables |
| Acquisition Made | Strategic behavior indicator (1/0) | Descriptive tables |
| Closed | Risk/survival indicator (1/0) | Descriptive tables |
| Patents before acceleration | Contextual innovation signal, (1/0) and quantity | Chapter 4 analysis |
| Accelerator name | Grouping variable (Google/Microsoft) | All tests |
Clarifying Statistical Use (Role in ANOVA)
The one-way ANOVA model is the primary statistical tool used in this dissertation. It was chosen to compare average or mean funding levels values and if there is significant difference. The groups of startup performance are between the two accelerator groups (Google vs. Microsoft) across defined periods of time.
The binary outcomes (e.g., IPO, closure) are not included in the ANOVA but are examined through descriptive and comparative tables (see Section 4.3). These variables help complete the picture of accelerator effectiveness but are not statistically tested for significance.
ANOVA was chosen because it offers a clear and replicable method to compare group means without assuming causality. It fits the cross-sectional structure of the database and supports group-level comparison over time.
Justification of Exclusions
Some variables were deliberately excluded from statistical testing, including:
- SIC Code: Too fragmented across 72 codes to provide group-level comparisons.
- City and Accelerator program: Not relevant for our study.
- Previous acceleration experience: Not uniformly reported and may disorient results.
Limitations in Variable Interpretation
While the variables were carefully selected and validated, a few caveats remain:
- Closure may include strategic exits or operational pivots, not necessarily failure.
- IPO within 3 years is rare and may be influenced by market timing and context, not accelerator impact.
- Acquisitions made could be minor or strategic, not necessarily indicators of dominance.
These limitations do not undermine the value of the data but must be considered in the broader interpretation of results.
Conclusion
This section has outlined the roles, analytical purpose, and reasoning behind the inclusion or exclusion of key variables used in this dissertation. By framing their use within the ANOVA model and supporting descriptive statistics, the study ensures a focused, consistent, and transparent approach to assessing corporate accelerator performance.
References
- Assenova, V., & Amit, R. (2024). Startups and innovation ecosystems: Performance signals and patent activity. Working Paper. University of Pennsylvania.
- Crunchbase. (n.d.). Crunchbase: Discover innovative companies and the people behind them. Retrieved from https://www.crunchbase.com/
- Fehder, D. C. (2023). Where do startup accelerators matter? Ecosystem complementarities and program effectiveness. Small Business Economics, 60(1), 149–173. https://doi.org/10.1007/s11187-023-00791-1
- Seitz, N., Krieger, B., Mauer, R., & Brettel, M. (2023). Corporate accelerators: Design and startup performance. Small Business Economics. https://doi.org/10.1007/s11187-023-00732-y
