PhD thesis · simplified guide

Can Training Founders Change How Judges See Them?

A plain-English digest of the legitimacy mind map — the argument, the evidence, the experiment, and the defense.

Open the interactive mind map → Paper–journal network GitHub repo

The thesis in one paragraph

Judges — investors, clients, institutions — hold a mental yardstick for what a "real business" looks like. Founders send signals; judges deliver a verdict. Scholars have studied each piece separately, but nobody has tested whether training founders to send better signals actually changes the judge's verdict. This thesis runs that test: a field experiment where half of the founders in the ATENA program receive a 12-hour legitimacy-training module and half don't.

Expectations→Conformity→Assent→Legitimacy The judge's yardstick → the founder's signals (what the module trains) → the judge's verdict (what Siraz measured) → legitimacy conferred.

The 5 claims I defend

#	Claim	Main risk	Hedge
1	Judges hold a yardstick	It differs France ≠ Italy	Context as a formal moderator
2	The module teaches the right signals	A signal could backfire	Teach consistency + audience-appropriateness, not tricks
3	Judges respond to these signals	Siraz used fake profiles	Real behavior + milestones on top of her lab evidence
4	Nobody connected the two sides	Someone names a paper that did	Precise claim: first field RCT, not first test ever
5	I can run the experiment	Dropouts, contamination, self-report	ITT + imputation + cohort separation + objective measures

The scholars — who showed what

Scholar	In one line	Role in thesis
Schoon 2022	Legitimacy lives in the relationship: expectations + conformity + assent	The umbrella framework
Suchman 1995	Legitimacy = "appropriate within the system" (pragmatic / moral / cognitive)	The definition everything footnotes
Siraz 2023–26	Looking like a normal, recognizable business beats endorsements; mixed signals stall judges; women, immigrants & minorities punished harder after identical failures	Proved which signals judges care about → what the module teaches
Claes 2025	Industry-favorite defenses actively hurt legitimacy — strategies backfire	Why training must be tested, not assumed
Ghods 2021–26	Legitimacy makes international expansion pay; meaning varies by country; auto-scorable from texts	Cross-border rationale + pitch-deck scoring
Bertrand 2025	Built the self-legitimacy scale; founders who feel illegitimate freeze (impostor syndrome)	The psychological mediator
Aldebert 2021–24	Accelerators lend legitimacy only once trusted themselves	What ATENA as a program can/can't do
Überbacher 2014	The field's map — shows the evaluator-cognition gap	Positions the research gap
Delmar & Shane 2004	Ventures that legitimate first survive longer (real data)	Empirical backbone of "legitimacy first"

The 5 legitimacy states (Siraz et al. 2023)

Legitimacy is not yes/no — it's a spectrum.

1 · LegitimateNobody even questions it — a bakery on your street.

2 · ConditionalAccepted if you meet conditions — GMOs with labels.

3 · Unknown — where ATENA founders startThe judge can't decide. Undetermined = invisible (fix: get noticed) · Undecided = seen but unplaceable (fix: give a familiar frame — "it's like X for Y" ← the module's job) · Polarized = half love it, half hate it (fix: different framing per audience).

4 · Conditionally illegitimateImproper in principle, tolerated out of necessity — medical cannabis.

5 · IllegitimateActively rejected — scams.

Key insight: a venture's state can flip purely because the societal yardstick shifted — no change in its own behavior.

The method in brief

Design: randomized field experiment (Article 1) + interviews explaining the why (Article 2) + repeated surveys (T0 → T1 → T4).

Define variables on Schoon's dyad — train conformity, measure assent.
Adopt Bertrand's self-legitimacy scale + a perceived-legitimacy scale.
Power analysis before recruiting (~100–175 per arm for d≈0.3–0.4).
Recruit → T0 baseline for everyone → randomize → separate cohorts.
Deliver the 12h module — 5 capabilities: Cognitive (fit a known category) · Pragmatic (show value) · Social (get vouched for) · Factual (show milestones) · Regulative (show compliance).
Survey both arms at T0/T1/T4 + collect objective milestones + pitch decks.
Track dropouts (leaver/stayer comparison + imputation + ITT).
Analyze: ITT → TOT → DiD → effect size + confidence interval.
Interview high and low responders; Gioia coding; second coder + kappa.
Write it as the missing link — Siraz proved judges respond to signals; Campos proved training works; the connection was never tested in the field.

Every problem has a fix: small sample → pre-registered power analysis · attrition → leaver/stayer + imputation · contamination → cohort separation + staggered rollout · self-report → objective milestones + text scores · coding subjectivity → dual coders + kappa · CMV → temporal separation + varied formats.

The numbers that matter

+30%profits from psychology-based training — Campos et al. 2017, Science (traditional training: +11%, n.s.)

+10.1%average profit effect across all business-training RCTs — McKenzie 2021 meta-analysis

~20,000conjoint ratings across Siraz's 4 experiments

n ≈ 16/d²per group for 80% power — the rule of thumb

Pollack et al. 2012 showed founder preparedness (trainable!) → legitimacy → funding amount — the thesis chain on real money. Tornikoski & Newbert 2007 showed legitimation actions predicted venture emergence more than founder traits — what you DO beats what you ARE. McKenzie's key line: "not finding a significant effect is not the same as finding no effect."

Fresh ammunition — 49 papers + the 3 validated scales

P01–P30 = most-cited canon-citers (OpenAlex) · P31–P46 = Scopus export, sorted by cites. Stable IDs, one line each, reorder freely. Formula first, names second. Grouped by level: the judge → the connection → the object → the field. Most new work sits in the connection — where this thesis lives.

ID	Formula — what constructs into what	Paper	Cit.
L1 · The judge — evaluator side
P16	verdicts → collective validity (masks low consensus) · solves: validity/agreement blur→multilevel split	Haack et al. 2025, JMS	14
P17	societal acceptance ≠ institutional legitimacy · solves: separate literatures→integration	Alsheimer et al. 2024, RSER	14
P18	norms → investor yardstick → IPO diversity↑ · solves: norm-shift mechanism→30-yr panel	Kenney et al. 2024, SBE	14
P20	judgment = variance + prediction + wisdom · solves: loose term→3 perspectives	Tsoukas et al. 2024, AMA	12
P21	CEO humor × schema-fit → social evaluation · solves: schema filtering→humor×fit model	König et al. 2024, AMR	12
P40	validity cues × categorical fit → propriety (fit dominates) · solves: evaluator propriety→factorial survey n=1,866	Thaler, Sievert, Siraz & Pinz 2026, JMS	2
P42	text scoring → legitimacy per dyad at scale · solves: uni-dim measures→multi-dim dyads (our instrument)	Ghods, Ricard, Aldebert & Schoon 2026, BJM	0
P43	multi-source evals → validity↑ → action · solves: single-source→multi-source	Du et al. 2026, BJM	0
P38	content endorsed + agent rejected → cooptive rejection · solves: object-agent tie→dissociation	Bourgoin et al. 2026, JMS	2
P37	misconduct × stakeholder → defence varies · solves: one-size repair→differentiated defence	Oberhauser et al. 2025, CSR	4
L2 · The connection — founder↔evaluator boundary
P09	hype → expectations↑ → resources↑ · solves: hype≠always good→contingent value	Rady et al. 2025, JBV	20
P10	founder updates ↔ backer sentiment · solves: hype decay→longitudinal pairs	Wood et al. 2024, JBV	19
P11	internal challenges → 8 inward strategies · solves: external-only focus→internal repertoire	Loos & Spraul 2024, JMS	19
P13	hype → legitimacy↑ then ↓↓ · solves: hyped crashes→3-phase model	Spivack et al. 2025, JBV	17
P15	local legitimacy × brokerage → diffusion · solves: local-only→diffusion beyond ecosystem	Kansheba et al. 2024, IJEBR	15
P19	crisis comms → crowdfunding outcome · solves: comms black box→dims + vignette	McKenny et al. 2024, ETP	14
P24	rebranding → host-market legitimacy · solves: untested cross-country→6-country case	Ramos da Silva et al. 2024, IMR	9
P26	make-or-break intermediary → 3 pathways · solves: unstudied→accelerator ethnography	Neuberger et al. 2024, AMJ	7
P29	secrecy → 7 discursive strategies · solves: secrecy legitimation→repertoire	Clarke et al. 2024, JBE	7
P28	eWOM → de-stigmatization → legitimation · solves: stigma mechanism→eWOM model	Villers et al. 2024, JSR	7
P32	non-Western ↔ Western investors → tensions · solves: universalism→tension lens	Jones & Murray 2026, JMS	7
P33	challenge×strategy configs → deals · solves: single-strategy→fsQCA 328	Yang, Zhang & Wu 2025, MD	7
P36	wedging+alignment → revised assessments · solves: hybridity liability→reframing	Dahan & Leca 2025, JMS	4
P41	political affinity → endorsements → IPO legitimacy · solves: transfer assumed→pathways	Song, Hu & Xia 2025, JMS	1
P44	contradiction → myths → internal legitimacy · solves: contradiction→myth model	Howard & Brown 2026, JMS	0
P39	crisis narratives → legitimacy managed · solves: disclosure→CDA+ethics	de Castilho Rossoni & Lima 2026, CSR	2
L3 · The object — venture side
P02	green innovation → legitimacy → export↑ · solves: EMF barriers→innovation vehicle	Shu et al. 2024, MOR	35
P04	rival alliances → ecosystem defence · solves: venture vulnerability→rival-alliance defence	Chen et al. 2024, JMS	27
P05	resourcefulness → venture process · solves: fragmented lit→integrated model	Williams et al. 2024, JMS	23
P07	category density → legitimacy spillover · solves: new forms legitimize→spillover	Li, Cai & Bosma 2024, SBE	21
P27	category reconfiguration → new-category legitimacy · solves: challenger legitimation→3 strategies	Rhee, Lo & Durand 2025, OrgSci	7
P31	awareness+motive+weak capability → decoupling · solves: signal w/o substance→config drivers	Xu et al. 2025, JMS	20
P34	responsible ent. → social legitimacy · solves: boundary conditions→time-lag survey	Adomako & Tran 2025, CSR	7
P35	EO → substitutive legitimacy → underconformity · solves: underconformity→substitutive mech	Askarzadeh et al. 2025, JMS	6
L4 · The field — macro environment
P01	strategies×positions×foundations×time×arenas → legitimation · solves: fragmented→5-element framework	Vaara et al. 2024, JoM	68
P03	legitimacy = create→maintain→lose · solves: snapshots→process view	Meyer & Tse 2025, IBR	30
P06	illegitimate actors → institutional repair · solves: threat framing→repair role	Rintamäki et al. 2024, AMR	22
P08	intensive mentoring → accelerator effect · solves: active ingredient→mentoring	Rechter & Avnimelech 2024, SBE	21
P12	university incubator → social intrapreneurship · solves: econ-only→social value	Nafari et al. 2024, Technovation	18
P14	racial identity → lived experience · solves: race as control→identity-centered	Santos et al. 2024, SBE	17
P22	underexploited opportunity → inequality · solves: equalizer myth→inequality	Lewis et al. 2024, AMR	12
P23	outcome = institution + cohort · solves: confound→Bayesian decomposition n=1,350	Avnimelech et al. 2024, SMJ	12
P25	3R strategies → marginalized legitimation · solves: marginalization→3R strategies	Spanuth & Urbano 2024, AMP	8
P30	accelerator × ecosystem = complement/substitute · solves: uniform effect→ecosystem moderator	Qin 2025, JIBS	3
P45	invoked past → evaluative authorization · solves: legacy divergence→authorization theory	Markman et al. 2026, JMS	0
P46	digital forces → evaluation assumptions break · solves: stable audiences→rethink	Clemente et al. 2026, JMS	0
S · Scale anchors — validated instruments (any year)
P47	self-legitimacy scale (4 dims) → impostor syndrome → intention↓ · solves: self-legitimacy never measured → validated scale n=122	Bertrand, Benamara & Aldebert 2025, HAL wp	—
P48	individual perceptions → pragmatic/moral/cognitive legitimacy scale · solves: collective-only measures → individual-perception scale (7 studies)	Alexiou & Wiggins 2019, Strategic Organization	138
P49	entrepreneurial self-efficacy scale — benchmark to discriminate against · solves: no ESE measure → validated scale (Q6's foil)	Chen, Greene & Crick 1998, JBV	~4k

The committee's hardest questions

"Isn't it just a gap nobody filled?"

No — a theoretical tension: legitimacy lives in the judge's head, training operates in the founder's. Theory can't predict whether signals cross that boundary.

"What if you get a null result?"

Still a contribution: it proves founder-side training can't cross the boundary → policy should target judges/platforms instead.

"Isn't it all self-report?"

No — objective milestones + Ghods text-scoring of pitch decks + a planned conjoint study with real bank officers.

"Why Bertrand's scale, not self-efficacy?"

Self-legitimacy measures entitlement to be a founder, not skill — it's what triggers impostor syndrome.

"Won't bias override training?"

Possibly — that's why subgroup analysis (gender, origin) is built in. A ceiling effect is itself a finding.

"France ≠ Italy?"

Ghods shows legitimacy representations differ by context → modeled as a formal moderator, not pooled.

"Why only 12 hours?"

Clean causal identification — long incubators confound training with peer effects and mentoring.

Where each paper goes in the thesis

Chapter	Core argument	Key papers
Introduction	The legitimacy deficit is structural, not merit-based	Siraz JBE 2026, AOM 2020; Bertrand 2025
Theory	Legitimacy = multi-state relational judgment	Siraz JMS 2023; Schoon 2022; Suchman 1995
Hypotheses	Categorical fit dominates; training improves cue coherence	Thaler/Siraz JMS; Siraz EMR 2026; Aldebert 2021/23
Boundary conditions	Strategies can backfire; bias may cap effects	Claes EMR 2025; Siraz JBE 2026
Methods	Founder-side node of the dyad + objective validation	Schoon 2022; Ghods BJM 2026; Bertrand 2025
Contribution	First field RCT crossing the founder→evaluator boundary	Full network synthesis + Ghods 2021 template

"Everyone studied one link of the chain. I test whether pulling on the middle link — training the founder — actually moves the last link — the judge's verdict."