3.1. Data
This study utilises the most recent firm-level data from the World Bank Enterprise Survey for Vietnam (Available at
http://www.enterprisesurveys.org) (accessed on 10 February 2026) [
37], a nationally representative survey designed to capture firm characteristics, innovation activities, performance outcomes, and business environment conditions in developing and emerging economies. The WBES is widely used in empirical research on innovation and firm performance due to its standardised methodology, cross-country comparability, and detailed coverage of both formal and informal innovation activities [
38,
39,
40].
The WBES applies a stratified random sampling design to ensure representativeness of the formal private sector. In Vietnam, firms are stratified along three dimensions: industry, firm size, and geographic location. Industry stratification follows the International Standard Industrial Classification (ISIC), covering manufacturing and selected service sectors. Firm size is defined based on the number of permanent full-time employees and categorised as small (5–19 employees), medium (20–99 employees), and large (100 or more employees). Geographic stratification ensures coverage across major regions and economic centres within Vietnam.
Within each stratum, firms are selected randomly from an official business registry, which reduces sampling bias and allows inference at the national level. Sampling weights provided by the World Bank are available to adjust for unequal probabilities of selection.
The Vietnam WBES covers a broad cross-section of firms operating in manufacturing and services, including both domestic and foreign-invested enterprises. This diversity is particularly valuable for analysing innovation behaviour, as firms differ substantially in access to finance, technology, and international markets. Export-oriented and foreign-owned firms tend to exhibit higher innovation intensity, while domestic small and medium-sized enterprises (SMEs) face greater resource constraints.
The initial dataset consists of 1028 surveyed firms. Observations with missing information on key variables—namely innovation indicators, sales growth, employment growth, or core control variables—are excluded from the analysis. Additionally, to mitigate the influence of outliers, we follow [
41] to winsorise the sales and employment growth variables at the 2.5th and 97.5th percentiles. After excluding observations with missing values for key variables and constructing growth measures, the final estimation sample comprises 667 firms. The exclusion of observations due to missing data may introduce potential selection bias if the dropped firms differ systematically from those retained in the sample. However, the final sample continues to exhibit substantial variation across firm size, sector, ownership structure, and innovation activities, suggesting that the main results are unlikely to be driven solely by sample selection.
Survey weights are available in the WBES dataset but are not applied in the baseline analysis, as the focus is on estimating relationships rather than population-level aggregates.
The WBES is particularly well suited to the Vietnamese context, where innovation often takes incremental and non-patent-based forms. Many firms engage in product or process innovation without conducting formal R&D, making traditional innovation indicators such as patents or R&D intensity inadequate. The survey’s direct questions on product innovation, process innovation, and innovation-related expenditures allow for a more accurate representation of firm-level innovation behaviour.
In addition, the WBES includes firm-reported measures of business environment constraints, such as access to finance and regulatory obstacles, which are central to Vietnam’s STI environment. Incorporating these variables enables the analysis to control for institutional factors that jointly influence innovation decisions and firm growth outcomes. Detailed definitions and descriptions of all variables used in the analysis are provided in
Table A1 in the
Appendix A.
Table 1 presents the summary statistics for the key variables used in the empirical analysis, including firm performance indicators, innovation activities, institutional constraints, and firm characteristics in the sample. The average employment growth rate is −0.804 per cent, with a median of zero, indicating that many firms experienced little or no employment expansion during the period considered. The large standard deviation (14.604) and the wide range from −77.119 to 41.421 suggest substantial heterogeneity in employment outcomes across firms. In contrast, sales growth is on average positive at 1.304 per cent, with a median of 4.447. However, the high standard deviation (23.329) and wide range from −96.838 to 42.551 indicate that firms differ significantly in their revenue performance.
The mean of 0.081 indicates that only 8.1 per cent of firms in the sample report introducing a new product, while 6.3 per cent implemented a new process innovation. In absolute terms, this corresponds to 54 firms introducing product innovations and 42 firms implementing process innovations. The relatively small number of innovative firms may limit statistical precision, particularly for estimates related to innovation variables. However, this distribution is consistent with existing evidence that innovation activity is concentrated among a subset of firms in emerging economies. Engagement in R&D-related activities is also low, with only 3 per cent of firms reported spending on R&D. Although the mean value of R&D expenditure is 15.868 million VND, the median is zero, indicating that most firms do not invest in formal R&D activities. This pattern suggests that innovation among Vietnamese firms is concentrated among a small subset of enterprises and is likely to be incremental rather than research intensive.
The average firm age is approximately 14.7 years, indicating a mix of relatively young and established firms. About 28 per cent of firms hold an internationally recognised certification, while foreign ownership accounts for 12 per cent of the sample. Around 21.9 per cent of firms have export activities, and 17.7 per cent operate internationally. Small firms dominate the sample (65.1 per cent), followed by medium-sized firms (30 per cent). Overall, these statistics highlight the diverse structure of the sample and provide important context for the subsequent empirical analysis.
Table A2 in the
Appendix A presents the pairwise correlation matrix for the main variables used in the analysis. Overall, the correlations between the explanatory variables are relatively low, suggesting that multicollinearity is unlikely to pose a serious concern for the regression analysis. Both employment growth and sales growth show positive but weak correlations with the innovation variables, including product innovation, process innovation, and R&D expenditure and R&D participation. R&D expenditure captures the amount invested in research and development activities, while R&D participation is a binary variable indicating whether the firm spent on R&D activities. The correlation coefficients are generally below 0.15, indicating that although innovative firms tend to experience better performance outcomes, the relationships are not strong.
Among the innovation variables, moderate correlations are observed between process innovation and innovation-related expenditures, particularly R&D expenditure and R&D participation, suggesting that firms engaging in process improvements are also more likely to invest in innovation-related activities. Institutional constraints, such as financing and legal obstacles, exhibit weak correlations with both innovation measures and firm growth indicators. This pattern indicates that while these factors may influence firm behaviour, their effects are unlikely to dominate the relationships. Overall, the low magnitude of the correlations supports the use of multivariate regression to further examine the relationship between innovation and firm performance. To further assess potential multicollinearity, we compute Variance Inflation Factors (VIF) for the combined OLS specification used as a robustness check. The VIF values are generally low, with most variables below 2. Although the small and medium firm indicators exhibit slightly higher VIF values, this reflects their construction as mutually exclusive size categories rather than problematic multicollinearity. All VIF values remain well below the conventional threshold of 10, indicating that multicollinearity is not a serious concern (see
Table A3 in the
Appendix A).
3.2. Empirical Specification
To address potential endogeneity in the relationship between innovation and firm performance, this study employs an instrumental variable (IV) approach. Endogeneity may arise due to reverse causality, measurement error, or omitted variables. For instance, firms with higher growth may be more likely to invest in innovation, leading to reverse causality. In addition, unobserved managerial ability or firm-specific characteristics may influence both innovation decisions and performance outcomes, resulting in biased ordinary least squares (OLS) estimates.
To mitigate these concerns, we instrument innovation variables using access to external financing, measured by the availability of financial sources. This instrument is chosen based on its relevance to firms’ ability to undertake innovation activities, as access to external finance can ease liquidity constraints and enable investment in new products, processes, and development activities. Conditional on observable firm characteristics, access to financing is assumed not to affect firm performance directly, except through its influence on innovation. While financing conditions may shape firms’ overall investment capacity, their direct effect on short-term sales or employment growth is expected to be limited once firm characteristics and market conditions are controlled for. This assumption is consistent with prior studies emphasising the role of financial constraints in shaping innovation rather than directly determining firm performance, thereby supporting the validity of the exclusion restriction.
We apply two-stage least squares (2SLS) to estimate the impact of innovation on firm performance. In the first stage, the innovation variable is regressed on the instrumental variable and other control variables:
where
represents one of the innovation measures (product innovation, process innovation, R&D expenditure, or R&D participation).
denotes the instrument variable,
is a vector of control variables including firm characteristics, ownership structure, and market conditions,
is the error term capturing unobserved factors that influence firms’ innovation activities but are not included in the model.
In the second stage, the predicted values from the first stage are used to estimate their effect on firm performance:
where
denotes either sales growth or employment growth for firm i;
is the predicted values of innovation;
denotes the vector of control variables; and
is the error term. The coefficient
captures the estimated effect of innovation on firm performance. The innovation variables are included separately in the baseline specifications to avoid multicollinearity and to isolate the effects of different types of innovation, which may operate through distinct channels. This approach is also necessary due to identification constraints in the IV framework, as a single instrument does not allow for the simultaneous instrumentation of multiple endogenous innovation variables.
To assess the strength and validity of the instrument, we report first-stage F-statistics for weak instrument diagnostics and the Wu–Hausman test for endogeneity. Following conventional thresholds, an F-statistic above 10 indicates that the instrument is sufficiently strong. These diagnostics are presented alongside the regression results to support the reliability of the IV estimates.
Consistent with prior research, firm performance is measured using sales growth and employment growth, which capture distinct but complementary dimensions of firm expansion. Sales growth reflects market performance and demand-side effects of innovation, while employment growth captures labour demand responses and broader developmental implications [
42,
43]. The literature emphasises that innovation may affect these outcomes through different channels and with varying intensity, underscoring the importance of analysing them separately [
13,
44].
Innovation activities are operationalised through four variables reflecting different aspects of innovation investment and engagement. Our approach is consistent with the view that innovation is a multidimensional process and that different types of innovation may generate heterogeneous performance effects. (1) Product innovation, measured by the introduction of new or significantly improved products, captures firms’ ability to differentiate offerings and respond to market demand. The literature finds that product innovation is positively associated with sales growth and, in some cases, employment growth, as firms expand output and market reach [
19,
28]. Product innovation is therefore expected to exert a positive effect on both performance measures. (2) Process innovation reflects improvements in production or delivery methods aimed at enhancing efficiency and reducing costs. Prior studies show that process innovation improves productivity and competitiveness, which can indirectly support firm growth [
34]. However, its effect on employment is theoretically ambiguous, as efficiency gains may reduce labour intensity even as output expands. (3) The amount of R&D expenditure is included as an indicator of firms’ commitment to knowledge creation and capability development. From a resource-based and dynamic capabilities perspective, R&D investment strengthens absorptive capacity and supports long-term innovation outcomes. Empirical evidence suggests that R&D expenditure is positively related to firm performance, particularly sales growth, though effects may be delayed or context-dependent in emerging economies [
35]. (4) R&D participation captures whether firms engage in innovation activities, measured as a binary variable indicating whether the firm spent on R&D in the last fiscal year. This measure complements R&D expenditure by capturing whether the firm undertakes R&D at all, rather than the amount spent. Firms participating in R&D activities achieve higher sales growth. R&D participation may facilitate the commercialisation of innovation and generate more immediate performance effects. Prior studies find that such investments are positively associated with firm performance, especially when firms face technological and institutional constraints [
11].
We employed a set of control variables to account for firm characteristics and institutional conditions that influence performance outcomes, thereby reducing omitted variable bias. Financing obstacles are included to capture constraints on firms’ ability to invest in innovation and scale operations. The literature identifies access to finance as a critical determinant of firm growth in emerging economies, with financing constraints shown to negatively affect sales and employment growth [
37]. Legal and regulatory obstacles reflect the institutional environment in which firms operate. Regulatory burdens and weak legal frameworks can increase compliance costs and hinder firm expansion, particularly in terms of employment. Prior studies suggest that such obstacles are associated with weaker growth outcomes [
37]. Firm age is included to account for lifecycle effects.
Older firms may benefit from accumulated experience and networks but may also face organisational rigidity and slower growth. Empirical evidence indicates that the relationship between age and growth is context-dependent, with effects that may be negative or neutral in emerging economies [
34]. Quality certification is used as a proxy for managerial capability and operational efficiency. Certified firms are often better positioned to compete in domestic and international markets [
45]. Export status captures access to international markets, which can enhance learning, scale, and innovation incentives. Exporting firms tend to exhibit stronger performance outcomes due to exposure to global competition and demand [
7]. International linkages, such as licencing agreements or participation in global networks, reflect knowledge spillovers and access to external technology. These linkages are expected to be positively associated with innovation outcomes and firm performance [
46]. Foreign ownership is included to control for differences in capital access, technology, and managerial expertise. Foreign-owned firms often outperform domestic firms in sales, though employment effects may vary depending on production structure and technology intensity [
36]. Industry controls, including a manufacturing sector indicator, are included to account for structural differences across sectors. No directional effect is imposed a priori, as industry effects primarily serve to control for heterogeneity. Firm size captures scale effects. Smaller firms may exhibit higher relative growth rates but also face greater resource constraints. The literature suggests that the relationship between size and growth is mixed and context-specific [
40].
To further assess potential multicollinearity, we compute Variance Inflation Factors (VIF) for the combined OLS specification used as a robustness check. The VIF values are generally low, with most variables below 2. Although the small and medium firm indicators exhibit slightly higher VIF values, this reflects their construction as mutually exclusive size categories rather than problematic multicollinearity. All VIF values remain well below the conventional threshold of 10, indicating that multicollinearity is not a serious concern (see
Table A3 in the
Appendix A).
All statistical analyses are conducted using R version 4.5.3.