GEM sponsorluse toetuskiri_2026

Lp Sigrid Rajalo 13.märts 2026 Pöördume Tallinna Tehnikaülikooli GEM-i uurimisrühma nimel palvega jätkata Globaalse Ettevõtlusmonitooringu (Global Entrepreneurship Monitor, GEM) Eesti uuringu läbiviimise toetamist 2026. aastal. Majandus- ja Kommunikatsiooniministeerium on toetanud GEM Eesti uuringu teostamist juba kolmel järjestikusel aastal – 2023, 2024 ja 2025 – ning tänu sellele on GEM-ist kujunenud järjepidev ja rahvusvaheliselt võrreldav andmeallikas Eesti ettevõtlusaktiivsuse ja -keskkonna hindamiseks ning poliitikakujundamise toetamiseks. 2025.aasta GEM Eesti uuringu aruanne on lõpp-toimetamisel ning edastame selle ministeeriumile esimesel võimalusel. Veebruari lõpus avaldati ka GEM-i globaalne aruanne, mis on kättesaadav GEM-i kodulehel GEM Global Entrepreneurship Monitor ning sisaldab Eesti näitajaid rahvusvahelises võrdluses. Kuna GEM uuring on oma olemuselt iga-aastane ja tsükliline – nii riiklikul kui globaalsel tasandil kogutakse andmeid igal aastal, et tagada trendide jälgitavus ja riikidevaheline võrreldavus –, oleme valmis sama uurimis- meeskonnaga jätkama GEM Eesti uuringu läbiviimist 2026. aastal ning palume ministeeriumilt rahastust 2026. aasta GEM uuringu teostamiseks summas 89 585 eurot. Kinnitame, et oleme tutvunud ministeeriumi esitatud lisatingimustega ning oleme nendega nõus. Võtame need 2026. aasta tööplaanis ja väljundites selgelt arvesse: täiendame raportite lõpus poliitikasoovitusi nii, et need oleksid senisest detailsemad ja lihtsamini poliitikakujundamises kasutatavad; koostame ja viime ellu selge kommunikatsiooniplaani; korraldame ministeeriumi soovil töötoa või seminari koos MKM-i ja sidusrühmadega, kus GEM Eesti 2026 tulemusi ühiselt arutada ja tõlgendada; ning loome raamistiku andmete jagamiseks teiste ülikoolidega, et avaliku raha eest kogutud andmeid saaks võimalikult laialt kasutada ka teadus- ja õppetöös (sh lõputöödes), järgides GEM-i reegleid ja andmekaitsenõudeid. Kommunikatsiooniplaani osas näeme 2026. aastal ette, et pärast andmekogumise ja analüüsi lõppu valmivad GEM Eesti 2026 uuringuaruanne ja lühike poliitikabrief peamiste TALLINNA TEHNIKAÜLIKOOL Ehitajate tee 5 Tel 620 2002 19086 Tallinn E-post [email protected] Rg-kood 74000323 www.taltech.ee järelduste ning soovitustega; seejärel toimub ministeeriumiga kooskõlastatult tulemuste tutvustamine ja arutelu töötoas/seminaril MKM-i ja sidusrühmadega, mille eesmärk on järelduste ja poliitikasoovituste ühisarutelu ning võimalike rakendussammude täpsustamine. Tulemuste laiemaks levikuks koostame ministeeriumiga kooskõlastatult pressiteate ja veebikommunikatsiooni, valmistame ette kokkuvõtliku esitlusmaterjali ning võimalusel visuaalsed lühivormid (nt infograafikud), et tulemused jõuaksid arusaadavalt nii poliitikakujundajate kui ka praktikute ja laiema avalikkuseni. Olulised sõnumid ja ajastuse kooskõlastame ministeeriumiga enne avalikustamist. Käesolevale pöördumisele lisame eraldi failina artikli, mille on koostanud GEM Eesti projektijuht Sirje Ustav ning mis valideerib NES-i (ekspertuuringu) tulemuste valiidsust. Palume Majandus- ja Kommunikatsiooniministeeriumilt jätkuvat toetust GEM Eesti 2026 uuringu läbiviimiseks. Vajadusel oleme valmis kohtuma ja vastama täpsustavatele küsimustele. Lugupidamisega, Merli Reidolf Tallinna Tehnikaülikooli ärikorralduse instituudi direktor /allkirjastatud digitaalselt/ Lisainformatsioon: Helena Rozeik, kontakt +372 527 2314; Sirje Ustav, kontakt +372 508 7488 TALLINNA TEHNIKAÜLIKOOL Ehitajate tee 5 Tel 620 2002 19086 Tallinn E-post [email protected] Rg-kood 74000323 www.taltech.ee  GEM Timeline for the 2026 research cycle The GEM data collection process follows a similar cycle every year, although the dates and deadlines do vary slightly Data Collected by Teams Data Processed by GEM Teams Check Data CONFIRM SPONSORS Global Report Produced DEADLINE: Annual Schedule signed – FEBRUARY 16th DEADLINE: GEM Global Services Fee paid – MARCH 31st DEADLINE: NES Proposals – APRIL 3rd DEADLINE: APS Proposals – MAY 8th Annual Meeting and Launch of Global Report DEADLINE: APS and NES Data – JULY 10th Jan 26 Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Jan Feb Request For Proposal Package and Administrative Final APS and NES Results – OCTOBER 9th Documentation Sent to Teams – JANUARY 26th 2nd APS Results – SEPTEMBER 25th 1st APS Results – AUGUST 28th APS = Adult Population Survey 1st APS Results – AUGUST 29th NES = National Expert Survey N.B: Late APS and NES Proposals as well as late data submissions are only allowed when agreed directly with GEM Global Data Supervisor Francis Carmona and NES Coordinator Alicia Coduras.  D oes expert consensus hold : T ime dynamics in ecosystem ratings from GEM expert surveys ABSTRACT Expert assessments are widely used to diagnose entrepreneurial ecosystem conditions, yet we know surprisingly little about the time dynamics of such ratings: do annual expert evaluations behave as stable barometers, or do they exhibit sharp, domain-specific shifts that can be mistaken for broad ecosystem change? We address this critique using three consecutive years of Estonia’s Global Entrepreneurship Monitor (GEM) National Expert Survey (NES) (2023–2025; N = 117 experts). Using SPSS General Linear Models , we estimate year effects while accounting for experts’ field of expertise, gender, and age, examine stability and change at item and domain levels and apply false discovery rate controls for multi-item testing. We further assess robustness to endpoint response styles on the 0–10 scale and incorporate a 2025 supplement measuring experts’ perceived participation capacity and enabling activity. Results show that temporal movement is concentrated: most domains are stable and robust to expert composition, while a policy/program-related item displays a pronounced 2023–2024 decline that remains low in 2025 and is not explained by response style. We contribute a time-dynamics lens for interpreting expert-based ecosystem diagnostics, distinguishing ‘consensus and ‘break domains’ that stress the meso level position of experts rather than generalized narratives of ecosystem decline. Keywords: entrepreneurial ecosystems; expert surveys; GEM NES; ecosystem governance; temporal dynamics; measurement robustness INTRODUCTION Entrepreneurial ecosystems (EE) have become one of the dominant lenses for explaining why entrepreneurship varies so sharply across region s (Stam, 2015) . By Stam t he ecosystem view emphasizes interdependent systemic conditions (e.g., finance, policy, talent, support infrastructure, cultural legitimacy) that jointly enable venture creation and growth. Ecosystem scholarship has matured rapidly (Wurth et al. , 2022) , yet measurement remains a persistent challenge: many ecosystem constructs are latent, multi-dimensional, and difficult to observe directly, which creates demand for diagnostics that are both interpretable for policymakers and usable for researchers (Crotti et al., 2025) . Moreover , Rietveld & Patel (2023) claim that experts disagree so greatly in their evaluations of entrepreneurial framework conditions (GEM, EFCs) that meaningful cross-country comparisons and within-country (longitudinal) analyses are precluded. On the other hand, Amorós, Bosma, & Levie (2013) give a warn ing about misuse of data by researchers such as incorrect weightings or misattribution of constructs to variables. Precisely because ecosystems are complex and multi-actor, research and policy communities look for diagnostics that can summarize “how the ecosystem is doing” in ways that are comparable, communicable, and actionable ( Schrijvers et al., 2023) , indicating to a need to integrate more regional data . One of the most widely used tools for ecosystem diagnosis still is the Global Entrepreneurship Monitor (GEM) and National Expert Survey (NES), which captures experts’ assessments of the Entrepreneurial Framework Conditions (EFCs) that shape entrepreneurial activity and policy leverage points (GEM, 2025) . NES-based measures are routinely used for benchmarking, ecosystem “bottleneck” identification, and national reporting, and they increasingly inform comparative ecosystem narratives and indices (Crotti et al . , 2025; World Bank, 2026). Yet, despite of this extensive use, we know less about what these expert ratings represent individually or over time horizons , especially because each year the expert cohort s vary . S o are they stable indicators of ecosystem conditions, or do they respond sharply to year-specific policy and macro-context shifts? As researchers indeed we see divers ity in expert evaluations for the same condition (EFC) and studies picking it up ( Ritveld and Patel 2023; Pfeifer et al., 2021). Pfeifer et al. (2021) suggests that expert perceptions of entrepreneurship conditions can shift over time and can diverge by their specialization . As a result, ecosystem researchers and policymakers face an interpretive ambiguity: when an NES score changes, does it reflect a real ecosystem movement, shifts in expert composition, or measurement biases ? This reflects problematics in understanding EE meso-level agency , we need to learn more about. There is ongoing discussion about macro policy or micro level challenges in EE, recommendations to seek for a bottom-up action in EE (Krueger, 2024; Hruskova, 2024; Candeias & Sarkar, 2023), good overview of ongoing discussions in EE research by Wurth, Stam and Spigel (2023) , to argue that meso level studies are scarce. We see e xperts as experienced in entrepreneurial financing, mentoring, policy program implementation, education , up to infrastructure building , who function “in between”, at the meso-level, representing the institutional layer, influencing and shaping the EE through their specialized knowledge and enabling position s , who act as orchestrators shaping collaboration and coordination in ecosystems ( Hernández-Chea et al., 2021). Or do they act as own limited expert-area based informants only? The gap matters because expert surveys are vulnerable to key-informant and common method concerns ( Ritveld & Patel, 2023). T his paper aims to address that gap to empirically establish experts’ mediating role in EE by developing and applying a time-dynamics lens for GEM ecosystem ratings and experts individual characteristics (field of expertise, age, education) , distinguishing domains where expert consensus holds ( i.e. ratings remain stable and robust) from domains where consensus breaks ( i.e. domain-specific shifts over time), using Estonia’s GEM NES data across three consecutive years. Research questions for investigation are : 1 . To what extent are expert assessments of EFCs stable across years versus sensitive to year-specific shifts? 2 . Which ecosystem (RFC) domains show the largest year-to-year movement in Estonia years 2023 to 2025? 3 . Is apparent disagreement driven by expert cohort composition (field/gender/age), or by time-varying context? 4. How see experts themselves their ( meso )position in EE – are they active enablers or just informants/reporters? This design explicitly addresses key-informant concerns by modelling expert attributes and applying conservative multiple-testing SPSS analysis. The GEM NES framework is particularly well suited to this purpose because it targets policy-relevant ecosystem domains (EFCs) using a consistent methodology, while also permitting examination of whether changes are concentrated in specific areas rather than diffuse across the instrument or depend on experts’ personal priorities. Our focus on time dynamics yields a clear interpretive trust for entrepreneurial ecosystem assessment scholarship. If NES ratings are largely invariant to expert composition and mostly stable over time, this strengthens their use as a credible barometer of ecosystem conditions. Conversely, if ratings exhibit concentrated year-to-year shifts in particular domains, this suggests that some components of the ecosystem are more sensitive and thus researchers and policymakers should interpret changes as potentially meaningful signals rather than as generic fluctuation. In doing so, we speak directly to ongoing debates about how ecosystems should be conceptualized and measured, and how diagnostic instruments can support theory development and policy learning rather than just providing rankings. We intend to make three contributions. First, we advance entrepreneurial ecosystem research by proposing a stability-versus-break framework for interpreting expert-based ecosystem indicators over short time horizons , addressing the recent critique . Second ly , we provide evidence on the robustness of GEM NES ratings with respect to expert composition, addressing a key validity concern raised in the broader key-informant literature and in recent critiques of NES-based indices ( Rietveld & Patel (202 3 ) . Third, we offer practical implications for ecosystem monitoring: a dashboard logic that distinguishes broadly stable “consensus domains” from “break domains” where sharp shifts may require closer institutional and policy attention . In the following sections we test our argument s within entrepreneurial ecosystem theory, describe the NES data and measures based on GEM Estonia (2023-2025) surveys , and discuss implications for ecosystem scholarship and empirical evidence -based entrepreneurship policy. THEORETICAL OVERVIEW To understand the positions experts hold in entrepreneurship ecosystems (EE) and thus, evaluation trustworthiness, we learn from theoretical discussions of entrepreneurship ecosystems configurations, governance and actors, with a focus on meso-level enabling power. Entrepreneurial ecosystem as a systemic configuration of enabling conditions Entrepreneurial ecosystem (EE) research has become a prominent discussion for explaining why some places generate more productive entrepreneurship than others (Stam, 2015). More than decade ago it was stated that entrepreneurial ecosystems, are that work to support not just quantity but quality of entrepreneurial activity and more specifically, local communities are those in the cente r of economic development (Krueger, 2012). Rather than treating entrepreneurship as an outcome of a single input (e.g., finance or education), the ecosystem perspective emphasizes configurations of interdependent actors and factors whose alignment enables entrepreneurial experimentation, resource mobilization, and growth ( Schrijvers et al., 2024; Theodoraki & Messeghem , 2017). A central contribution of this synthesis is to differentiate framework conditions (e.g., formal institutions, culture, demand) from systemic conditions (e.g., networks, leadership, finance, talent, knowledge, and support services). These systemic conditions form the platform of the ecosystem moderating the relationship between a variety of entrepreneurial actions and regional economic development ( Audretsch & Belitski , 2021). More broadly, recent work consolidates a consensus in definition of entrepreneurial ecosystems as “a set of interdependent actors and factors coordinated in such a way that they enable productive entrepreneurship within a particular territory,” and calls for more evidence on the mechanisms through which domains interact and change (Wurth, Stam & Spigel, 2022). This systemic emphasis is theoretically consequential for measurement and inference. If ecosystems operate through complementarities and bottlenecks, then evaluating any single domain in isolation can be misleading (Stam & Spigel, 2016; Wurth et al. (2023), deficiencies in one domain (e.g., policy implementation capacity) may constrain the value of strength in another (e.g., availability of talent). The ecosystem lens therefore encourages domain-specific diagnosis while maintaining a system logic: understanding entrepreneurship requires attention to both the quality of multiple enabling domains and their coordination. Stam’s causal scheme makes this explicit by connecting framework and systemic conditions to entrepreneurial activity and value creation (Stam, 2015). Ecosystem dynamics over time: stability vs localized change A persistent critique of EE research is that ecosystems are often described as static “lists” of attributes rather than as evolving processes (Spigel & Harrison, 2018). But in parallel there are studies which argue that Entrepreneurial Ecosystems are not merely pre-existing environmental conditions but coordinated systems whose functioning depends on active agency (Stam, 2016, Brown & Mason, 2017, Roundy, 2018). The dominant definition of ecosystems emphasizes interdependent coordinated actors and factors to enable productive entrepreneurship, making governance and coordination mechanisms central (Wurth et al., 2022). Process view emphasizes how ecosystems are reproduced and transformed through sequences of events, shifting relationships among actors, and reconfigurations of institutions and resource flows. Spigel and Harrison (2018) argue for kind of process-based view that can explain ecosystem evolution and transformation, including the emergence of different ecosystem structures over time. Empirically, for example, intermediary organizations such as incubators operate beyond simple brokerage by shaping collaboration dynamics and even configuring ecosystem structure and service provision (Hernández-Chea et al., 2021). Stam (2015) specifies that a system approach helps identify the “weakest link” that limits entrepreneurial ecosystem performance. Bottleneck in any element can have adverse consequences on others, reducing the overall capacity of the ecosystem (OECD, 2025) where missing elements can constrain outcomes ( Schrijvers , 2024). Mack and Mayer (2016) present an evolutionary framework for EE development, making time and regional history central. Analyzing missing elements, they recommend the stage of EE by division of evolution – birth, growth, sustainment and decline. Similar approach is proposed again by Cantner et al. (2020). However, these dynamics lack empirical ground as considering combinations of social, political, economic, and cultural elements within a region that support the development and growth of innovative start-ups and encourage nascent entrepreneurs (Spigel, 2017) – these are not so much dependent on regional evolution but entrepreneurial mindset, knowledge and professional planning, democratic leadership and action. Many ecosystem elements can be sticky (e.g., deep cultural norms, long-term institutional arrangements), while others (e.g., the functioning of support programs, policy implementation practices, or coordination among intermediaries) can shift more rapidly (Spigel & Harrison, 2018). More recent work from Harima & Harima (2024) foregrounds orchestration and leadership capabilities of anchor organizations as mechanisms through which networks and resources are mobilized and adapted over time. Hence, the collective action among ecosystem actors’ forms pipelines and shared infrastructures that sustain entrepreneurial processes (Harima & Harima, 2024; Hruskova, 2024). Together, these perspectives position ecosystems as partially created, but sometimes also destabilized, through meso-level orchestration, leadership, and intermediary action, rather than being determined by macro policy alone (Santos, 2023, Roundy, 2024). Johnson and Schaltegger’s (2020) propose that the meso level plays a key role in mediating the bi-directional causal mechanisms. The implication here is that ecosystem change is often uneven across domains. Over short horizons (such as a three-year window), it is theoretically plausible to observe broad stability across many ecosystem conditions alongside localized shifts in domains that are more immediately exposed to policy, macroeconomic turbulence, or changing discourse. Recent synthesis work similarly argues that EE research must move beyond general claims how coordination problems and institutional change shape outcomes (Wurth, Stam & Spigel, 2022). This motivates our “time dynamics” lens that looks at temporal movement if it is a diffuse drift across all indicators or potentially concentrated in particular ecosystem . To capture the genesis and evolution of EE, the network of interactions of individual elements should be studied ( Audretsch et al., 2018). Governance and orchestration and the special role of enabling Ecosystems are not “just there” but are actively shaped through orchestration and leadership at the meso - level (Harima & Harima, 2021). Feldman and Zoller (2012) argue that leadership is provided by what they call dealmakers: experienced entrepreneurial actors who link other actors in an ecosystem and define entrepreneurial networks. This enables the characterisation of three main ecosystems: the entrepreneurial ecosystem (macro level), the entrepreneurial support ecosystem ( meso level), and the business incubator ecosystem (micro level). The results highlight the importance of studying the interplay among sub-ecosystems ( Theodoraki & Messeghem , 2017). Experts in entrepreneurship field often act at meso-level, understanding the micro level needs and challenges on one hand, and political objectives, agendas on the other and represent or lead structures in-between. Ecosystems are not simply collections of resources; they are coordinated - sometimes effectively, often imperfectly , through the actions of intermediaries and institutional actors who connect entrepreneurs to markets, capital, knowledge, and policy instruments. Stam’s systemic conditions framing highlights leadership, networks, and support services as central coordinating features of the ecosystem. (Stam, 2015). Though in research we acknowledge a nchor organizations’ strategic functions in managing networks and resources can be understood as orchestration (Harima & Harima, 2024) which as well can come in variety of qualities. Ecosystem functioning depends not only on the presence of formal structures but also on the relational and cognitive foundations that enable coordination; thus, domain-specific shifts in governance-facing items may signal localized disruptions in operational legitimacy and collaborative capacity rather than generalized ecosystem decline ( Theodoraki , Messeghem and Rice, (2018). After the establishment of an extensive network with local ecosystem stakeholders as bridging social capital, anchor organizations orchestrated mechanisms can transform this network into bonding social capital. (Harima & Harima, 2024). So, c oordination among actors and factors is essential for enabling productive entrepreneurship and that stronger theoretical progress requires specifying the mechanisms that connect domains and shape collective action (Wurth, Stam & Spigel, 2022). Spigel (2017) draws attention to the lack of a strong theoretical foundation, governance structures, and how individual factors contribute to the activities of entrepreneurial ecosystems . I ndividual factor connectedness within entrepreneurial systems, individual institutional impact, is indeed vague ( Alvedalen and Boschma , 2017) , and indicates to the role of enablers. External enabler is a collective label for non-trivial changes to the business environment such as new technologies, regulatory changes, demographic and sociocultural trends, macroeconomic swings, and changes to the natural environment, which are expected to trigger, shape, or enhance some entrepreneurial pursuits (Davidsson, 2020). Entrepreneurship enablers are individuals, organizations, and institutions that facilitate and support the development and growth of entrepreneurial activities ( Kimjeon & Davidsson, 2021). They provide various forms of assistance, including mentorship, coaching, training, access to finance, networking, and other resources that can help entrepreneurs overcome barriers to entry and succeed in their ventures. Entrepreneurship enablers can be government agencies, incubators, accelerators, universities, business associations, investors, and other support organizations that operate in the entrepreneurial ecosystem. Their role is to create an environment that encourages entrepreneurship, fosters innovation, and promotes economic growth and development. I n this paper we argue that experts have a position of EE enabler. Experts can represent any of abovementioned institutions. They gain knowledge, experience and expertise in entrepreneurship, but also move between different organisations, carrying their expertise around different domains of the ecosystem. Domains can respond quickly to changes in government priorities, administrative capacity, funding cycles, or institutional trust. Or domains tied to slower-moving foundations (e.g. cultural attitudes, deep institutional arrangements) which may shift more gradually. C oordination failures, legitimacy shocks, or policy turbulence may produce a measurable difference in how actors evaluate certain domains, even when other domains remain stable. This domain sensitivity becomes especially relevant when scholars rely on expert assessments as real-time diagnostics of ecosystem functioning. Experts as embedded diagnosticians: meso-level actors and key informants Experts as ecosystem actors are not random respondents , they occupy relevant positions in the ecosystem , their entrepreneurial behavio u r is enabled and constrained by their networks (Aldrich & Zimmer, 1986 ). These yields privileged information and positional frames (Pfeifer et al., 2021) . Noting at the same time that t he development, reproduction, and outputs of entrepreneurial stakeholders depend on the social ties between actors (Spigel, 2017). Consistent with a multi-level view of entrepreneurial ecosystems, in which interacting sub-ecosystems include an entrepreneurial support ecosystem (Stam, 2015) , we conceptualize experts as embedded diagnosticians whose positions in policy, support organizations, finance, and education shape informational access and interpretive frames ( Theodoraki & Messeghem , 2017) . T hus , they may both , experience ecosystem fluctuations firsthand and participate directly or indirectly in the coordination processes , which also may be evaluated by themselves . This embeddedness provides informational advantages (experts may observe implementation problems, coordination gaps, or institutional bottlenecks early), but it also implies that their ratings may reflect positional vantage points and interpretive frames rather than a single objective truth. This is consistent with the classic key-informant methodology literature, which argues that informants can validly report on complex interorganizational and institutional phenomena, but that validity hinges on informant selection and knowledge , and can be threatened when reports systematically differ across informants ( Kumar, Stern & Anderson , 1993) . They emphasize that interorganizational research often relies on single informants despite known risks and discuss the need to consider perceptual agreement and informant selection to improve reliability and validity. Meaning in ecosystem diagnostics that expert disagreement is not just an error but may also be a signal of heterogeneous access to information and divergent interpretive frames - precisely the mechanisms expected when evaluators occupy different meso-level positions. These validity concerns have become particularly salient in the context of GEM NES–based indices (Pfeifer et al. 2021). Recent critiques argue that the subjective nature of expert evaluations can undermine strong claims based on cross-country rankings and that limited precision constrains within-country trend assessment when uncertainty is properly acknowledged. Importantly, these critiques (e.g. Ritveld & Patel, 2023) do not imply that expert surveys are unusable; rather, they sharpen the methodological implication that expert-based diagnostics require disciplined inference and careful interpretation , especially when analysts examine many items across many ecosystem domains (Crotti et al., 2025) . To sum up , these arguments lead to the conceptual model for the current study (Figure 1 ). We posit that time-varying context (macro conditions) can alter perceived ecosystem conditions in ways that are not uniform across domains. Those domain conditions inform expert ratings (NES), while experts’ embeddedness shapes what they observe and how they interpret domain performance. ================== Figure 1 about here ================== Finally, response processes act as a measurement overlay that can influence observed dispersion and apparent change, motivating robustness tests rather than ad hoc interpretation. This model motivates our core empirical aim: to distinguish domains if and where expert consensus holds from domains where consensus breaks. METHOD OLOGY Studying the expert consensus about entrepreneurial ecosystem conditions if they remain stable or break, we us e three years, 2023,2024 and 2025 of Estonia’s Global Entrepreneurship Monitor (GEM) National Expert Survey (NES) study results, which are carefully following the rigour of GEM NES methodology (G EM, 2025 ). Our empirical strategy builds on the conceptual model developed in the Theory section: time-varying context (policy shifts, macro conditions, technology and discourse) shapes ecosystem conditions across domains, which in turn inform expert ratings, while experts , positioned as embedded diagnosticians , may also differ systematically in their assessments. Methodologically, this implies that changes in NES scores must be interpreted cautiously unless they are robust to expert composition and multiple testing across many items. Accordingly, the analysis proceeds in four steps. First, we describe the samples and measures for each NES study and harmonize items across years to ensure comparability. Second ly , we test for domain-specific time dynamics by estimating year effects for each item and for aggregated domain ( research construct) scores, comparing the magnitude and concentration of movement across ecosystem pillars. Third ly , we assess composition robustness by incorporating expert characteristics : field of expertise (1–9), gender (0/1), and age—into the same models and evaluating whether these factors explain systematic differences once year effects are controlled. Because the NES contains many items, we emphasize construct-level aggregation to reduce the risk of over-interpreting isolated chance findings. Finally, to strengthen the interpretation of experts as embedded diagnosticians rather than purely external observers, we incorporate a supplementary 2025 module (Theme X) that measures experts’ perceived participation capacity and enabling activity (e.g., involvement in policy shaping, collaboration with policymakers and entrepreneurs, practical contributions, and expert networking). This module is used descriptively and, where appropriate, linking perceived embeddedness to ecosystem ratings. Our methodological choices follow directly how entrepreneurial ecosystem conditions are typically operationalized in GEM, using SPSS databases and calculation . For this study three separate (year based) databases are consolidated for statistical comparison. Sample Our analysis combines three years of Estonia’s GEM National Expert Survey (NES): 2023 (N=45), 2024 (N=36), and 2025 (N=36), yielding a combine d sample of N=117 experts (Table 1) . Experts are distributed evenly across the nine expertise fields (factors 1 - 9), with four experts per field (total 2x36) in years 2024-2025 and five experts per field (total 45) in 2023. Gender composition is overall balanced, and the age distribution centers in midlife (mean age ~ late 40s to early 50s, depending on year ) which reflects well that gaining expertise requires years and experience, and that expertise was looked for. Items cover the standard NES domains ( 1-9 ) . ================== Table 1 about here ================== Measures Th is study uses GEM NES constructs, assessed by experts (GEM, 2025) in three years concequtive national studies We report consolidated results , scores given by 117 experts (Figure 2 ) at two levels: a) item-level models that test for domain-specific movement and composition effects b) construct-level summaries (EFCs) averaging across items within each domain. IBM Statistical Package for Social Sciences (SPSS) is used for statistical analysis. Theme X: Expert activity and embeddedness (2025 extra) To capture experts’ embeddedness and perceived influence in the entrepreneurial ecosystem beyond their evaluations of ecosystem conditions , we fielded a short supplementary module in 2025 (“Theme X: Expert activity. In my country…”). The module consists of six statements assessing experts’ perceived opportunities and capacity to (a) participate in entrepreneurship policy shaping, (b) collaborate across policymakers - expert organizations - entrepreneurs, (c) be heard by policymakers, (d) contribute practically to entrepreneurship development (e.g., training, mentoring, investing, support measures), (e) rely on an active and effective expert network supporting entrepreneurship, and (f) push sustainability and responsible economic practices in firms. Responses were recorded on a 0 - 10 scale (higher values indicate stronger agreement). All items were completed by the full 2025 expert sample (N = 36). We operationalize Expert Embeddedness & Activity as the mean of the six items (X1, X2, X4, X6, X8, X10). Internal consistency of the six-item scale is satisfactory (Cronbach’s α = 0.79), supporting aggregation into a single composite measure for descriptive reporting and exploratory analyses. Table 2 reports item-level means and standard deviations. =================== Table 2 about here =================== Results F irst f or each NES item (and for each domain/construct average), we estimated year-to-year differences for each harmonized item across 2023 vs 2024 vs 2025, using the same logic as the main models (year effects, with conservative inference for many items). All analyses were conducted in IBM SPSS Statistics 29. In order to explore differences between years, one-way ANOVA and post hoc tests ( Bonferroni ) were applied. Across the full set of analyzed items (98 base items harmonized across years), we find limited evidence that expert ratings systematically differ by field of expertise, gender, or age once year is controlled (Table 2) . Although a small number of items show nominal significance (p<0.05) for these variables, none of these effects remains statistically valid. The main movement , which is in focus of RQ2, is concentrated in Domain B (political conditions). A one-way ANOVA confirms a significant year effect for Domain B (F(2,114) = 7.42, p = 0, 001; η² = 0, 115), indicating a moderate-to-large time-related shift concentrated in this pillar. Pairwise comparisons (Bonferroni-adjusted) show that Domain B is significantly lower in both 2024 vs. 2023 (p = 0, 003) and 2025 vs. 2023 (p = 0, 013), while 2025 vs. 2024 is not statistically different. ==================== Table 2 about here ==================== This pattern is consistent with the “consensus holds” argument of our framework: expert assessments of ecosystem conditions appear broadly shared and not strongly dependent on expert profile. Item-level ANOVAs indicate selective time dynamics (depicted on Figure 2) in a small set of items Domain B items show notable declines from 2023 to 2024 (and limited recovery in 2025), including B07, B06, and B05 (all p < .05). For B04, mean ratings fall sharply from 2023 (M = 7.12, SD = 2.42, n = 42) to 2024 (M = 5.24, SD = 2.34, n = 34) and remain similarly low in 2025 (M = 5.17, SD = 2.06, n = 36). The year effect (Figure 3) is statistically strong (F(2,109) = 9.26, p < .001; η² = .145). Bonferroni-adjusted pairwise comparisons show that B04 is significantly lower in 2024 vs. 2023 (p = .003) and 2025 vs. 2023 (p < .001), while 2025 vs. 2024 is not different (p = 1.00 ). ================= Figure 3 about here ================= To sum up , the claim that time dynamics are not common across the NES instrument , i nstead, they are concentrated in Domain B, with B04 emerging as the most robust item-level indicator of a 2023→2024 “break” that persists through 2025 . Other items (notably additional B items) show suggestive movement reinforcing the inference that the temporal signal is localized rather than general. This concentration is theoretically consistent with the argument that certain governance-/policy-adjacent ecosystem components can be more shock-sensitive over short horizons, while much of the broader ecosystem diagnosis remains comparatively stable . Thirdly we asked if ra tings are systematically associated with expert profile like field of expertise , gender or age . To answere the RQ2 the General Linear Model (GLM/UNIANOVA) framework was used : Year (2023/2024/2025) as a fixed factor, field of expertise ( Expertfactor 1–9) and gender (0/1) as fixed factors, and age as a covariate. We first tested composition effects at the domain (construct) level, where each domain score is computed as the respondent-level mean of items within that domain. As shown in Table 4, once Year is controlled, no domain exhibits a composition effect that remains robust after FDR correction (q > .05 for all field, gender, and age tests across domains). This suggests that, at the level of ecosystem “pillars,” expert assessments are broadly composition-robust . There are a small number of nominal (uncorrected) signals , e.g., gender effects for Domains A and B (p ≈ .020–.023) and an age effect for Domain B (p ≈ .037) , but appearing statistically not important . In other words, any profile-linked differences at the domain level appear small and not reliable . The item-level results indicate that disagreements like those observed on specific items (e.g., C01) are not systematically explained by stable expert-profile segmentation. Instead, item-level variability appears more consistent with vantage-point heterogeneity and interpretation differences that do not align neatly with field/gender/age categories , precisely the reason we emphasize construct-level aggregation and conservative inference. To sum up, t he domain-level results support the “consensus holds” logic: experts’ aggregate ecosystem evaluations are not systematically segmented by field/gender/age once year effects are accounted for. Additionally, we focused on final check if observed shifts remain robust under conservative inference and response-style , whether the time dynamics identified in RQ1 should be interpreted as meaningful “breaks” rather than chance findings arising from many simultaneous item tests or response-style artifacts on a 0 - 10 scale. T his is a question whether the conclusions hold when we account for multiple testing across many items and test robustness to endpoint response styles (frequent use of 0 and 10) as it can be witnessed in database . Because the NES uses a 0 - 10 response scale, some respondents may exhibit an “endpoint response style”, a tendency to use 0 and 10 more frequently than others. Such response styles can inflate dispersion at the item level and, in principle, could bias year-to-year comparisons if endpoint-heavy respondents are unevenly distributed across waves. To assess whether our two focal patterns, high dispersion on C01 (2025) and the year-to-year “break” on B04 are caused by the endpoint use, we conducted a test at respondent-level on each expert’s endpoints to exclude respondents with increasingly strict endpoint thresholds. Endpoints = ( N of 0s + N of 10s ) / N answered items We then re-estimated results under increasingly strict filters that exclude respondents whose endpoint rate exceeds 0.30, 0.20, and 0.10. B04 year-break remains robust. The 2023 - 2024 decline in factor B04 but persists with similar magnitude under all endpoint filters. In the full sample, B04 averages decline from 7.12 (2023) to 5.24 (2024) and remain low in 2025 (5.17). Excluding endpoint-heavy respondents (pct_endpoints > 0.30; removing 6 observations) yields statistically the same pattern (6.84 for 2023, 5.24 for 2024 and 5.12 for 2025 ) and the year contrasts remain statistically significant (Δ2024 - 2023 ≈ - 1.61, p≈0.004; Δ2025 - 2023 ≈ - 1.72, p≈0.001). Even under a stringent filter (pct_endpoints > 0.10; removing 32 observations), the year differences remain sizeable and significant (Δ2024 - 2023 ≈ - 1.53, p≈0.016; Δ2025 - 2023 ≈ - 1.67, p≈0.006). Overall, these results indicate that the B04 “break” is not driven by a small set of endpoint-style respondents. C01 dispersion remains high after filtering. The C01 item (2025) shows substantial disagreement (Mean ≈ 4.86, SD ≈ 2.44, IQR = 4, range 0 - 10). Removing endpoint-heavy respondents does not reduce this dispersion meaningfully: under pct_endpoints ≤ 0.30, C01 it remains widely spread (Mean ≈ 5.03, SD ≈ 2.36, IQR = 4), and even under pct_endpoints ≤ 0.10 the IQR remains 4 and the SD remains ≈ 2.48. Thus, the observed heterogeneity on C01 appears to reflect genuine differences in assessment and/or interpretation rather than being a modification of endpoint response style. S um ming up, we note that moderate endpoint filters do not remove disproportionately respondents from any single study and this robustness checks strengthen confidence in our argumentation: Estonia’s NES ratings are broadly stable across expert compositions. Finally, research question four was looking into e xpert embeddedness in participation in the ecosystem (Table 3 ). Th is th eme ( X ) captures how strongly NES experts feel they are able to participate in ecosystem shaping, both through governance/policy influence and through practical enabling actions. All items are rated on a 0–10 scale (higher = stronger agreement) . ==================== Table 3 about here ==================== The highest rated item: “I have been able to practically contribute…” ( 8.2 ) indicates that experts perceive themselves as actively contributing through real actions (training, mentoring, investment, support measures). It supports the theoretical framing of experts as meso-level enablers rather than passive observers. Second best: “A professional, active and effective network of experts has developed” ( 7.1 ) suggests solid meso-level infrastructure (expert collaboration and networks) exists, which is very important for ecosystem orchestration. It proves expert networks are perceived as relatively well developed . Further it is impressive to note that s ustainability nudgi ng has a strong normative support . “Experts should take a stronger role in pushing sustainability principles…” ( 7.3 ) indicates that e xperts broadly agree that they should play a stronger role in steering firms toward sustainable practices . T his is a n important takeout of available sense of mission for policymakers to use. Governance interface appears to be weake st : collaboration is moderate, political receptiveness is lowest. “Excellent cooperation across policymakers … entrepreneurs” ( 5.9 ) loudly declare that triangle macro-meso-micro of EE needs bigger connectivity. It correlates well with meso-macro communication deficiency: “Politicians listen to entrepreneurs and experts” ( 5.0 ), the lowest mean of the research construct. This is the key tension: experts report strong ability to contribute and strong networks, but only moderate cross-level collaboration and comparatively weak perceived political listening. T hese six items work together well as a single scale: Cronbach’s alpha ≈ 0.79 (good internal consistency) and supports the meso-level logic: experts are not just observers - they clearly report their ability to participate in governance, reflect the quality of coordination, and their enabling actions (training/mentoring/investing/programs). The result is highly policy relevant . Within 2025, the Theme X scale correlates positively with several NES construct averages , especially with AI construct , which demonstrates ( Spearman ρ = 0.73 ) a very strong connection. Also , the constructs D, B, F, C show moderate , but positive correlations. To sum up, experts who feel more able to engage/enable tend to rate ecosystem (and especially AI readiness) more positively - consistent with embeddedness shaping perceptions and/or optimism. Not claiming causality , it is a very interesting finding for further exploration. Discussion Our findings address directly to recent debates on the validity of NES-based ecosystem indices. Prior work shows that expert evaluations may diverge systematically by specialization ( e.g., entrepreneurs and policymakers can rate the same national context differently ) and that this heterogeneity can undermine strong inferences from aggregated indices if uncertainty is ignored (Pfeifer et al., 2021; Ritvield & Patel, 2023) . Rietveld and colleagues argue that although NES-based composites can be internally consistent, poor interrater reliability and resulting imprecision complicate cross-country rankings and make year-to-year interpretations within countries precarious when measurement uncertainty is properly acknowledged. Our results align with this caution to the extent that many nominal item-level “changes” do not survive conservative inference across a multi-item instrument, and some items indeed display substantial dispersion. At the same time, by focusing on a single-country setting and explicitly separating temporal movement from expert composition while applying response-style robustness checks, we identify a localized and substantively interpretable break (B04) that persists across specifications. The item B04: “ In my country … new businesses can obtain most required permits and licenses in about a week ”, reflects that the situation is worse in 2024-2025 then 2023. This truthfully reflects the outspoken dissatisfaction that has emerged in Estonian society over the past couple of years with the rapidly growing bureaucracy. The government has promised to reduce bureaucracy in the business environment in 240 proposed acts (Government, 2025) . Especially are the permits and licences criticized in planning and construction. In this sense, our study reflects the reality under which expert-based ecosystem diagnostics remain informative: not as precise ranking device, but as disciplined and trustworthy screening tool that distinguish stable “consensus domains” from robust “break signals” that merit deeper institutional investigation and attention. Variations in C01 : “ New and growing businesses can access a variety of government assistance through a single agency ” remain high, but none of specific items can not be systematically explained by expert-profile segmentation. Instead, item-level variability appears more consistent with vantage-point heterogeneity and interpretation differences . To explain the C01, in Estonian entrepreneurship ecosystem the reality can be open to many interpretations . There is one main organisation implementing government-initiated support programs to support entrepreneurship and competitiveness - Estonian Business and Innovation Agency (EIS, 2025). However, with own objectives and regulations, there are many enterprises who should look elsewhere when help is needed. Like it is common in Europe, many alternative location-based communities, hubs and incubators or meso level entrepreneurial support ecosystem ( Theodoraki ja Messeghem , 2017). So rather, it is the question of orchestration of anchor organisations (Harima & Harima, 2024) where experts have varying views and so it is only fair to see the wide variety. This does not support that experts may be rating their experienced slice of the ecosystem, when the question intends a national-level judgement . On the contrary, it is in line with Aldrich & Zimmer (1986) claim that they occupy relevant positions in the ecosystem, their entrepreneurial behaviour is enabled and constrained by their networks . This answers the last research question where we can claim that carefully chosen cohorts of experts are (representing meso-level institutions) acting as active enablers (Davidsson, 2020). Estonian experts report high practical contribution (8.2) and a strong expert network (7.1) , but moderate collaboration across levels (5.9) and notably lower perceived listening by politicians (5.0) . This is exactly the kind of ecosystem governance story where capacity exists in the meso layer, but translation into policy responsiveness looks weaker. It is important to focus on the superconnectors , the liaison-animateurs who are proactive and great and connecting the connectors (Krueger, 2024). Hence, according to expectations, additional theme (X) justif ies the objective diagnostician framing and is strengthe ning our conceptual model: e xperts are positioned to evaluate conditions, and they experience the governance interface directly (involvement, collaboration, receptiveness). It is therefore crucial to understand the meso level enabling positions – who is an actual “expert”. Entrepreneurs as well as policy makers per se cannot be considered (obviously they sit on the other side of the table). Having adequate expert sample supports the claim that consensus can hold (shared diagnosis among embedded actors), while breaks may occur in domains where governance responsiveness shifts. The se results are highly policy relevant . Conclusion We propose few contributions based on results of this study. First, the c onceptual contribution: “consensus holds vs breaks” as a time-dynamics lens for expert ecosystem diagnostics , where we see justified agreement and consensus even in breaks. W e add to entrepreneurial ecosystem research by interpreting expert-based EE indicators over short time horizons and with distinctive sample . Second ly , we provide evidence on the robustness of GEM NES ratings with respect to expert cohort composition, addressing a key validity concern raised in the earlier literature and in recent critiques of NES-based indices. Our results indicate that incomparability between EEs might be caused by inability to define clearly “the expert” position, and therefore, when mixing stakeholders from macro, meso and micro levels, the outcome will reflect contradictory objectives. Focusing strictly to meso level actors - who understand the needs coming from micro level and odds from marco , we can have reliable diagnostics. Third, we offer practical implications for ecosystem monitoring: a dashboard logic that distinguishes broadly stable “consensus domains” from “break domains” where sharp shifts may warrant closer institutional and policy investigation , and for the reason. Finally, we add understanding how meso level experts judge their own position in the local EE . Considering their active participation and motivation in EE development, we cautiously propose to consider them as meso level enablers when moving between institutions and organisations orchestrating the cooperation, while also making a mutual claim that the Government is only moderately listening this ( meso )level. Therefore, we find the results of current research relevant to both researchers and policymakers - present ing results on temporal movement and composition robustness across domains , contributing to evidence-informed entrepreneurship policy. And most importantly, that NES data and measures (in controlled conditions) can be trusted. Limitations and future research T he most relevant limitation lies in studying a s ingle , small country case , where the sample per year is rather small for statistical analysis. Also, there is a limitation in generalizability, which are indicated both in literature and our results – different bases on defining “the expert” (e.g. entrepreneur vs policy maker) will be reflected in scores and making them incomparable between EEs. But it gives us the clear indication, where further studies should head . Expert ( NES ) cohorts should be built systematically to have reliable outcome s . For future research we expect to see similar studies in other EE s, as the replication is relatively easy. Noting that t his paper brings in front an idea of an expert who acts as a meso level enabler. There is a n eed for more focused study looking in linking t he objective indicators , understanding the meso level criteria of “experts” who , in our case, serve as trusted diagnostians and credible informants, in consensus. REFERENCES Amorós , J. E., Bosma, N., & Levie, J. 2013. Ten years of Global Entrepreneurship Monitor: Accomplishments and prospects. International Journal of Entrepreneurial Venturing , 5(2) : 120-152. Aldrich H. E., Zimmer C. 1986 . Entrepreneurship through social networks. In Sexton D., Smilor R. (Eds.), The art and science of entrepreneurship. Ballinger. Audretsch , D. B., & Belitski , M. 2021. Towards an entrepreneurial ecosystem typology for regional ecosystem typology for regional economic development: The role of creative class and entrepreneurship. Regional Studies , Vol. 55: 735–756 . Audretsch , D., Mason, C., Miles, M.P., O’Connor, A. 2018. The dynamics of entrepreneurial ecosystems. Entrepreneurship & Regional Development , Call for papers, Vol. 30(3-4) : 471 - 474. Bonomi Santos, J., Fernandes, A.R., deOliveira , P.T., Maia, L.M., Partyka, R.B . 2023. Increasing entrepreneurial ecosystem-level outcomes: Decoding orchestration in entrepreneurial ecosystems. Technovation . Elsevier, Vol. 128(C). Bosma, N., Hill, S., Ionescu-Somers, A., Kelley, D., Guerrero, M., & Schott, T. 2021. GEM Brown, R., Mason, C., (2017). Looking inside the spiky bits: a critical review and conceptualisation of entrepreneurial ecosystems. Small Business Economics , Springer, Vol. 49(1) : 11-30. Candeias, J.C., Sarkar, S. 2023. Entrepreneurial Ecosystems Policy Formulation: A Conceptual F ramework. Academy of Management Perspectives , Vol. 38(1). Cantner, U., Cunningham, J.A., Lehmann, E.E. 2020. Entrepreneurial ecosystems: a dynamic lifecycle model. Small Business Economics , Vol. 57 : 407–423 Crotti, R., Potter, J., Shah, P., Stam, E. 2025. Data and Methods for Entrepreneurial Ecosystem Diagnostics. OECD SME and Entrepreneurship Papers . https://www.oecd.org/content/dam/oecd/en/publications/reports/2025/06/data-and-methods-for-entrepreneurial-ecosystem-diagnostics_0851c7b7/5173f4a2-en.pdf Last accessed: 12.01.2026 Davidsson, P., Recker, J., von Briel, F. 2020. External enablement of n ew venture creation: A Framework. Academy of Management Perspectives , Vol.34, No 3. EIS, 2025. Estonian Business and Innovation Agency. About . https://eis.ee/en/about-the-organisation/?context=company-in-estonia Last accessed: 12.01.2026 Estonian Government . 2025. The government has approved over 240 proposals to reduce bureaucracy. News. https://valitsus.ee/uudised/valitsus-heaks-kiitnud-ule-240- burokraatiat-vahendava-ettepaneku . Last accessed: 12.01.2026. Gemconsortium . 2025. Global Entrepreneurship Monitor. https://www.gemconsortium.org/ . Last accessed 20.12.2025. Harima, A., Harima, J. 2024. Ecosystem orchestration: Unpacking the leadership capabilities of anchor organizations in nascent entrepreneurial ecosystems. Entrepreneurship Theory and Practice , Vol. 48(6). Harima, J., Harima, A. 2025. Ecosystem leadership in catalyzing the emergence of entrepreneurial ecosystems: social capital perspectives. Journal of Entrepreneurial Behavior & Research , Vol. 31(5) : 1133-1420. Hernández-Chea, R., Mahdad, M., Minh, T.T., Hjortsø, C.N. 2021. Moving beyond intermediation: How intermediary organizations shape collaboration dynamics in entrepreneurial ecosystems . Technovation , Vol. 108. Hruskova, M. 2024. Ecosystem pipelines: Collective action in entrepreneurial ecosystems. International Small Business Journal : Researching Entrepreneurship , Vol 42(1). Kim, P.H., Wenneberg , K., Croidieu , G. 2016. Untapped riches of meso-level applications in multilevel entrepreneurship mechanisms. Academy of management Perspectives , Vol. 30, pp. 273-291. Kimejong , J., Davidsson, P. 2022. External Enablers of Entrepreneurship: A Review and Agenda for Accumulation of Strategically Actionable Knowledge. Entrepreneurship: Theory and Practice 46(3). Krueger, N. 2012. Bridging town and gown: Best practice? An essay on growing the local entrepreneurial ecosystem. International Journal of Business and Globalisation , 9(4): 347–358 . Krueger, N. 2024. Top Ten Things that Research Tells Us About Ecosystems (+ Implications for Us!). ICSB International Council for Small Business . Kumar, N., Stern, L.W., Anderson, J.C. 2017. Conducting Interorganizational Research Using Key Informants. Academy of Management Journal , Vol. 36(6). Leendertse , J., Schrijvers , M., & Stam, E. 2022. Measure twice , cut once : Entrepreneurial ecosystem metrics . Research Policy , 51(9). Mack , E., Mayer , H. 2016. The evolutionary dynamics of entrepreneurial ecosystems . Urban Studies , Vol . 53(10) : 2118-2133. Mason,C . , Brown, R. 2014. Entrepreneurial ecosystems and growth-oriented entrepreneurship. Final report to OECD , Paris , 30:77-102. Miles, M. P., & Morrison, M. 2020. An effectual leadership perspective for developing rural entrepreneurial ecosystems. Small Business Economics, 54(4): 933–949. OECD. 2025. Data and methods for entrepreneurial ecosystem diagnostics . Paris: OECD Publishing. Pfeifer, S., Singer, S., Š arlija , N., Oberman Peterka, S. 2021. Perception of the National Entrepreneurship Conditions – Differences across time and expert specialization . South East European Journal of Economics and Business , Volume 16 (1 ): 1-17 . Rietveld, C.A., Patel, P.C. 2023. A critical assessment of the National Entrepreneurship Context. Entrepreneurship Theory & Practice , Vol. 47(6). Schrijvers, M., Stam, E., Bosma, N. 2024. Figuring it out: configurations of high-performing entrepreneurial ecosystems in Europe. Regional Studies , Vol. 58(5). Spigel, B. 2017. The Relational Organization of Entrepreneurial Ecosystems. Entrepreneurship: Theory and Practice 41(1): 49–72. Spigel, B, Harrison, R. 2018. Toward a process theory of entrepreneurial ecosystems. Strategic Entrepreneurship Journal , Vol. 12(1) : 151-168. Stam, E. 2015. Entrepreneurial Ecosystems and Regional Policy: A Sympathetic Critique. European Planning Studies , 23 ( 9 ): 1759-1769 . Spigel, B., Harrison, R. 2018. Toward a process theory of entrepreneurial ecosystems. Strategic Entrepreneurship Journal , 12(1): 151–168. Stam, E., Spigel, B. 2016. Entrepreneurial ecosystems. Utrecht School of Economics Discussion Paper Series : 16–13 . Stam, E. 2021. Entrepreneurial ecosystem elements . Small Business Economics , Vol. 57: 809– 832. Sternberg , R. Bloh , J. and Coduras, A. 2019. A new framework to measure entrepreneurial ecosystems at the regional level . Zeitschrift für Wirtschaftsgeographie . Vol . 63(2-4): 103-117. Rietveld, C. A. 2023. A critical assessment of the National Expert Survey data of the Global Entrepreneurship Monitor . Entrepreneurship Theory and Practice , Vol 47(6). Roundy, P. T. 2024. Entrepreneurial ecosystems as multiteam systems . Journal of Business Venturing Insights , Vol. 21 . Theodoraki , C. 2024. Building Entrepreneurial Ecosystems Sustainably. Foundations and Trends in Entrepreneurship , Vol 20(4). Now Publishers. Theodoraki , C., & Messeghem , K. 2017. Exploring the entrepreneurial ecosystem in the field of entrepreneurial support: A multi-level approach. International Journal of Entrepreneurship and Small Business , 31(1): 47–66. World Bank (2026). Data360, National Expert Survey (NES). https://data360.worldbank.org/en/dataset/GEM_NES . Last accessed 05.01.2026 Wurth, B., Stam, E., & Spigel, B. 2022. Toward an entrepreneurial ecosystem research program. Entrepreneurship Theory and Practice , 46 (3) : 729-778. TABLES Table 1. The sample. Summary of expert groups for 2023, 2024 and 2025 years, descriptive statistics NES field of expertise N M F av Age Edu cation 1 Financing 13 9 4 46,5 4,8 2 Governmental policies 13 8 5 48,8 4,4 3 Governmental programs 13 7 6 43,5 4,7 4 Education and training 13 5 8 56,5 5,0 5 R&D transfer 13 8 5 47,9 4,2 6 Commercial infrastructure 13 4 9 47,5 4,2 7 Internal market ope n ness 13 11 2 50,4 4,7 8 Physical infrastructure 13 9 4 46,7 4,9 9 Cultural and social norms 13 4 9 50,8 4,7 117 65 52 48,7 4,6 Source: Authors’ calculations based on GEM NES statistics 2023-2025 Table 2 . NES EFC domain wise mean scores and standard deviations 2023-2024 NES EFC Domain Mean_2023 N=45 Mean_2024 N=36 Mean_2025 N=36 SD_2023 SD_2024 SD_2025 A 5,41 5,24 5,15 1,75 1,49 1,43 B 6,03 4,82 5,06 1,68 1,49 1,29 C 5,84 5,27 5,39 1,92 1,87 1,43 D 5,75 4,90 5,66 1,83 1,66 1,97 E 4,79 4,50 4,50 1,90 1,95 1,34 F 5,68 5,46 6,02 1,92 1,65 1,34 G 5,86 5,72 5,62 1,39 1,41 1,28 H 6,98 6,81 6,89 1,90 1,56 1,72 I 7,77 7,30 7,30 1,47 1,59 1,85 P 4,62 4,33 4,92 1,80 1,2 0,82 SDGC 6,79 6,33 6,35 1,93 2,11 1,47 SDGN 6,45 6,43 5,97 1,79 1,79 1,81 SDGG 4,58 4,36 4,71 2,09 2,25 1,98 SDGS 5,76 6,05 5,49 1,88 1,89 1,80 AI 5,83 5,94 1,66 1,71 Source: Authors’ calculations based on GEM NES survey 2023,2024,2025 Table 3 . Expert activity and embeddedness in the ecosystem (2025, N = 36) In my country … M ean 0-10 As an expert in the field, I can be involved and have a say in shaping policies concerning the business environment 6,9 there is excellent cooperation between different levels – policymakers – expert organizations and entrepreneurs 5,9 politicians listen to and take into account the opinions and suggestions of entrepreneurs and experts 5,0 I have been able to practically contribute to the development of entrepreneurship (through training, mentoring, investment, support measures, etc.) 8,2 A professional, active and effective network of experts supporting entrepreneurship has developed 7,1 Experts and business landscape designers should take a more active role in pushing the principles of sustainability and sustainable management into companies' operations 7,3 Mean (N=36) 6,7 Source: Autohors’ calculations based on GEM NES survey (extra) data 2025 FIGURES Figure 1 . Conceptual model of the research (Time-varying context → ecosystem conditions → expert ratings), with expert embeddedness , activity and perceptions as a filter that can amplify sensitivity to certain domains Source: Authors’ deduction from theoretical background Figure 2. Mean scores for 9 EFCs given by three different groups of experts years 2023-2025 Source: Authors’ calculations based on GEM NES data 2023-2025 Figure 3. B04 – GEM NES factor (In my country … new businesses can obtain most required permits and licenses in about a week) ratings differentiation 2023-2025 Source: Autohors’ calculations based on GEM NES survey data 2023-2025