Macroeconomic Uncertainty and Risk: Collective Optimism of Small-Business Owners

Abstract

How small-business owners assess uncertainty and risk in the macroeconomic environment to report optimism toward their business activity is an important research question for policymakers and academics alike. Drawing on a monthly optimism index and a variety of uncertainty and volatility (risk) trackers from January 1986 to June 2017, in line with the expectation that business owners dislike uncertainty in economic policy, economic policy uncertainty may lower optimism. Business investment and sentiment volatility, however, contributed positively to optimism. The findings provide preliminary implications for policymakers in targeting indicators salient to the optimism index of small-business owners.

Keywords

risk uncertainty collective optimism

The drivers of the economic sentiment of small-business owners are of keen interest to policymakers (Dunkelberg & Wade, 2009; Phillips, 2002). The longest-running small business optimism index in the United States is the National Federation of Independent Businesses (NFIB) Small Business Optimism Index (Dunkelberg & Wade, 2009; Storey, 2011). Released monthly, the optimism index has a score of 100 as the baseline, with scores above 100 indicating a positive mood and scores lower than 90 signaling a recession. The NFIB optimism index remains among the most reliable proxies of the pulse of small businesses in the United States.

The NFIB optimism index represents the shared optimism or shared positive beliefs and expectations of a group of individuals (Santero & Westerlund, 1996). The collective sentiment is studied widely in assessing consumer, executive, supply chain manager sentiments, among others (Christiansen et al., 2014; Santero & Westerlund, 1996). Though individual optimism is partly driven by psychological (Cooper et al., 1988), genetic (Conrad, 2001), or environmental factors (Yuh et al., 2010), collective optimism of the business owners primes economic activity and informs policymakers on the needs of small-business owners.

For this research note, our contribution is primarily not theoretical, but we take a policy perspective by asking which of the available macroeconomic uncertainty and risk indicators influence the small-business owner optimism index in the United States. We focus on the role of two widely assessed elements in entrepreneurship—uncertainty and risk—in driving the collective optimism of small-business owners (Rakow, 2010). The assumption about the nature of uncertainty for our work requires discussion.

Based on Knight (1921), uncertainty is an “unknown, unknown” and risk is a “known, unknown”; as such, uncertainty may be empirically unmeasurable. Knight proposed three types of uncertainty: known distributions and unknown draws, unknown distributions and unknown draws, and nonexistent distributions with unclassifiable instances or true uncertainty. Our conceptualization of uncertainty is based on the first two and not the third type, as it can only be resolved by human action (Buchanan & Vanberg, 2002). As such, we do not focus on pure uncertainty, but based on Knight’s (1921) the uncertainty in the event spaces—set of all possible configurations of expected events—as a driver of decision-making under uncertainty.

Uncertainty- and risk-related macroeconomic factors that drive collective optimism have significant estimation challenges. Uncertainty- and risk-related factors could be correlated, and multiple indicators can aggregate in an unobservable fashion. Which types of uncertainty and risk indices do small-business owners take into consideration in forming optimism are not straightforward. Small-business owners may perceive the absence of reliable probability distribution that can be estimated from current data sources to predict future outcomes.¹ Aggregate small-business owner optimism index is an individual-level aggregation of a variety of firm-specific and environmental factors that are combined in an unobserved fashion, making it notoriously difficult to assess and attribute weights to understanding inclusion, combinations, and aggregations. Furthermore, due to a multitude of uncertainty and risk indicators, the traditional vector autoregressions (VAR) or time-series econometrics-based approaches may be of limited use. The VAR approach has limitations, as it may not be suitable for collapsing or compressing high-dimensional data (Stock & Watson, 2012), and others use graph-theoretic approaches and VARs in specifying risk and uncertainty (e.g., Balli et al., 2019). An additional limitation is that decision-makers cannot fully explain their evaluation process.

One approach to addressing this challenge is to use the stochastic multi-attribute acceptability analysis (SMAA), where the small-business owner is not able to fully report the parameters necessary for devising the optimism index (Doumpos et al., 2016, 2017). In this research note, based on SMAA, we use a Bayesian least absolute selection and shrinkage operator (LASSO) regression model built on the “ $σ - μ$ efficiency analysis”² (Greco et al., 2019a). We build on prior work by adding the Bayesian aspect to model for priors. Modeling for priors is essential because business owners report optimism based on past information and learn from macroeconomic indicators over time. We further use LASSO regression to overcome the challenge of an unobservable combination of attributes across a multitude of indicators. We caution that the approach, though with possible benefits, should not be construed as the only reliable method given there is no valid benchmark to assess the reliability of this measure against.

We use the monthly optimism index from NFIB from January 1986 to June 2017 and combine these data with monthly economic policy uncertainty and volatility (risk) indicators. We use both the aggregate and individual dimensions of uncertainty and volatility as we ex-ante do not know how the uncertainty and risk indicators could combine in explaining aggregate optimism index. The LASSO component of the proposed Bayesian LASSO regression model in this paper accommodates these multicollinearities between aggregates and individual indicators.

Overall, our goal with this research note is twofold—focus on an important index relevant to small-business owners in the United States and apply a method that helps overcome several estimation challenges in identifying the key uncertainty and risk-related drivers of such an index. We aim to make the following contributions. First, our results show that economic policy uncertainty contributes negatively and business investment and sentiment volatility contribute positively to optimism. The traditional theoretical frameworks in entrepreneurship have focused on positive and negative aspects of risk and uncertainty; however, we jointly consider these two theoretically studied factors to assess how small-business owners consider these indicators in reporting optimism. However, we also caution that several untested assumptions about functional forms and the interactions of risk and uncertainty entering as linearly separable terms in perceptions of small-business owners are required to simultaneously account for risk and uncertainty. Ours is a preliminary step aimed at untangling the risk and uncertainty perceptions in influencing small-business owners’ expectations. Despite the associated negative elements of risk that carry a loss proposition, inconsistent with the assumption of risk aversion among small-business owners (Fairlie & Holleran, 2012), the most consistent effect is the negative effect of economic policy uncertainty. More surprisingly, we find a positive effect of the business investment and sentiment Equity Market Volatility (EMV) tracker.³ Considering the two findings, business owners, as expected, may not positively perceive uncertainty in economic policy; however, they may perceive volatility in investment and sentiment more positively.⁴

Second, equally interesting are the findings on the limited role of widely espoused indicators discussed in the popular press. We find that uncertainty in monetary or fiscal policy, health care policy, or regulatory uncertainty has a negligible impact on optimism.⁵ Though monetary and fiscal policy uncertainty can impact access to capital or health care and are also popular when discussing challenges for small-business owners (Kitching, 2019; Snowe, 2006), these factors do not seem to be too critical in influencing optimism. Though regulation is considered an important consideration for small-business owners in the United States, regulatory uncertainty is not a driver of lower optimism. As such, for policymakers considering a gamut of factors that small-business owners are exposed to, our study is an early study that may help parse the key drivers of small-business owners’ optimism. Our primary contribution is empirical. We identify an important consideration for policymakers—in forming business optimism, business owners react negatively to economic policy uncertainty and positively to the business investment and sentiment EMV tracker (Åstebro et al., 2014; Greco et al., 2008).

In the following sections, we start by providing a brief theoretical background followed by the empirical specification and results. We close after discussing the implications for policymakers and academics.

Theoretical Background

Since the 1940s, economists have focused on measuring collective sentiment with the underlying premise that the opinions, feelings, and beliefs of the collectives in societal segments could help predict broader changes in the economy. The measure of economic sentiment is not only informative for the policymakers but is also, at times, a leading indicator of the economic cycle. Timely sentiment indices inform policymakers to act toward devising business-friendly policies.

Among the sentiment indices, NFIB’s optimism index represents the longest-running publicly available sentiment index of small-business owners in the United States. Though idiosyncratic measures are valuable for individual business owners, the sentiment index aims at improving population-averaged assessment. Therefore, moving from individual business owner optimism assessments studied widely in entrepreneurship literature, we zoom out and focus on the optimism of small-business owners’ assessment of expected future business conditions. A variety of studies have shown that consumer and investor sentiments affect market returns and despite their limitations as aggregate self-reported data, sentiment indices remain critical to macroeconomic decision-making (Christiansen et al., 2014).

Small-business owners and entrepreneurs are generally optimistic (Crane & Crane, 2007). We rely on Scheier and Carver’s (1985) definition of optimism as the extent to which the positive expected events of the future outweigh the negative expected events. The optimism index based on the outlook and planned hiring and investments, in the face of uncertainties and risks, explains the strategic and tactical actions of small-business owners. The sentiment index focuses on the time-varying assessments of the underlying sentiment based on changing macroeconomic conditions. Just as investors or consumers reporting sentiments are driven by a variety of personal and social factors, on the aggregate the goal is to assess the responsiveness of investors and customers to changing economic conditions. Similarly, the NFIB optimism index is based on the changing optimism of small-business owners with the changing macroeconomic environment. Collective optimism, defined as shared positive expectations that outweigh shared negative expectations, is previously studied at the team, organizational, community, or national level (Bennett, 2011). Our theoretical assumption, consistent with sentiment indices, is that of an aggregative index (Woehr et al., 2015). In classical microeconomics, aggregation, by summing the indicators of optimism, is problematic as the summated index may not inherit the key attributes of small-business owners’ preferences, and it may also not inherit transitivity among individuals whose optimism levels are summed.

Optimism, Risk, and Uncertainty

Risk and uncertainty are central to entrepreneurial endeavors (Knight, 1921). Knight (1921) proposed that risk refers to unknown outcomes from known distribution, whereas uncertainties refer to unknown outcomes from unknown distributions. In other words, risk refers to variations in possible outcomes that are known or previously experienced. Related to uncertainty, even though events are experienced they may be uncertain when no relevant information is available or when few individuals model “black swan” outcomes in their probabilistic reasoning (Taleb, 2010). Based on bounded rationality (Simon, 1997), though entrepreneurial decision-making is not random and irrational, studies have shown that entrepreneurs engage in deliberative decision-making by seeking information and support and by drawing on a variety of tools, including effectuation (Sarasvathy, 2009), search and satisficing (Berg, 2014), navigating environments by building coalitions (Dew et al., 2008), among others. Knight (1921, p. 237) highlighted that it is correct “to treat all instances of economic uncertainty as cases of choice between a smaller reward more confidently and a larger one less confidently anticipated.” The intuitive approach is faster and requires less effort, but it is subject to errors. The logical approach is cognitively effortful, is rule-based, and driven by conscious attention.

Based on the risk and uncertainty signals in the broader macroeconomy and considering the underlying challenges in processing such signals, a calculus of how multiple signals are interpreted, processed, and aggregated is difficult to ascertain. Therefore, we hypothesize small-business owners to be Bayesian learners who update their expectations in the context of their business and evolving macroeconomic signals. A Bayesian approach may be central to developing an understanding of how small-business owners interpret uncertainty and risk-based signals to devise optimism sentiment.

Small-business owners may not be able to accurately report combinations of uncertainty and risk information used to derive weights for reporting the optimism index. For example, small-business owners in the retail industry may consider an aggregate risk index but may also consider a consumer sentiment risk index. The consumer sentiment risk index is a part of the aggregate risk index; however, this empirically less desirable attention to aggregate and its components can be expected in practice and the traditional regression models may not suffice. The optimism index may not be based on additive parameter weights, though not irrational (Berg & Hoffrage, 2008), which are generally a result of intuitive and less linear thought processes that require accommodation of such errors and also variations in attention to aggregate and components of signals. To accommodate such varying combinations, we develop a model based on LASSO with Bayesian priors.

The methodology we offer is not a panacea for addressing the problem of aggregation and reporting of optimism sentiment in the presence of indicators of uncertainty and risk. Due to the lack of theoretical background, we do not propose a formal hypothesis; however, our research question is how do small-business owners interpret (positively or negatively) uncertainty and risk-based signals from the macroeconomy? Though the theoretical framework around this framework could gestate in the future, our primary goal is to empirically test the proposed research question. Based on the above discussion, we present our conceptual model in Figure 1, and below we provide a formal model of our methodology.

Figure 1

A conceptual overview. “Uncertainty must be taken in a sense radically distinct from the familiar notion of Risk, from which it has never been properly separated.... The essential fact is that ‘risk’ means in some cases a quantity susceptible of measurement, while at other times it is something distinctly not of this character; and there are far-reaching and crucial differences in the bearings of the phenomena depending on which of the two is really present and operating.... It will appear that a measurable uncertainty, or ‘risk’ proper, as we shall use the term, is so far different from an unmeasurable one that it is not in effect an uncertainty at all.” “Knightian uncertainty is a lack of any quantifiable knowledge about some possible occurrence, as opposed to the presence of quantifiable risk [uncertainty] acknowledges some fundamental degree of ignorance, a limit to knowledge, and an essential unpredictability of future events.”

Empirical Specification

Suppose we have a vector of explanatory variables $x_{t} \in R^{K}$ and a vector of outcomes $y_{t} \in R^{M} .$

A dynamic factor model is as follows.

\begin{matrix} y_{t} = Λ f_{t} + ξ_{t}, \\ f_{t} = ϱ f_{t - 1} + x_{t}^{'} γ + ζ_{t}, t = 1, \dots, n, \end{matrix}

(1)

where $Λ$ is a vector $M \times 1$ of factor loadings, $ξ_{t}$ is an $M \times 1$ vector of error terms, $ζ_{t}$ is the error term of the factor equation, $f_{t}$ is the dynamic factor, and $ϱ, γ$ are unknown coefficients in $R$ and $R^{K}$ , respectively. The dynamic factor model may seem quite general, but it is highly over-parametrized especially when the number of factors is large. Here, we use the one-factor model in Equation (1) only as an illustration and a benchmark to compare with our new model.

In many instances, we want to infer the effect of explanatory variables on outcomes via a composite indicator such as

C_{t} = y_{t}^{'} w, t = 1, \dots, n,

(2)

where $w \in S = {w \in R_{+}^{M} : \sum_{m = 1}^{M} w_{j} = 1} .$ Then the problem becomes how to fit the effect of explanatory variables $x_{t}$ on the composite indicator $C_{t}$ . If the weights were known, one could use least squares (LS) in the following regression.

y_{t}^{'} w = x_{t}^{'} b + e_{t}, t = 1, \dots, n,

(3)

where $e_{t}$ is an error term, and $b \in R^{K}$ is a vector of associated coefficients fits by LS. As the coefficients $w$ are unknown, this procedure does not work. Instead, we propose to use

y_{t}^{'} w = y_{t}^{*} + ε_{t},

(4)

where $ε_{t}$ is an error term, and $y_{t}^{*}$ is essentially the value of the composite indicator $C_{t}$ in Equation (2), so we may refer to it as a modified composite indicator. Equation (4) allows for the fact that a composite indicator is not precisely what we have on the left-hand side of Equation (3) and allows a two-step justification of sentiment. First, $y_{t}^{*}$ is modeled as a linear combination of the outcomes, as in Equation (4), and, second, it is expressed as a linear combination of the explanatory variables as in Equation (3).

Greco et al. (2019a) introduce a methodology for constructing composite indicators known as $σ - μ$ efficiency analysis. As discussed earlier, the additive composite indicators may not be representative of the underlying indicators. For the $σ - μ$ analysis, first, mean, μ, and the standard deviation (SD), σ, of their evaluations of components of the NFIB index is computed to maximize the mean and minimize the σ. The σ represents the instability in the evaluations concerning the variability of weights (Greco et al., 2019a). Greco et al. (2019a, p. 942) consider optimism index reported in a year-month to be Pareto–Koopmans efficient “with respect to μ and σ if there is no convex combination of μ and σ of the rest of the units with a value of μ that is not smaller, and a value of σ that is not greater, with at least one strict inequality.” The locus of all Pareto–Koopmans efficient units represent the first Pareto–Koopmans frontier, with each year-month assigned to one of the sequences of Pareto–Koopmans frontiers. Each year-month local efficiency relative to each frontier and its global efficiency relative to all planes in frontier is computed to improve the explanatory power of the resulting index.

The $σ - μ$ efficiency analysis is associated with another major problem, namely, the problem of “representative agent,” which can be mitigated by the use of multi-criteria decision analysis and SMAA. SMAA involves a decision-maker that is unable to provide the parameters required for the evaluation process (see e.g., Doumpos et al., 2016, 2017, as well as Greco et al., 2008, 2010, 2018, 2019b). As Greco et al. (2019a, p. 943) write: “More specifically, by considering a probability distribution on the set of feasible weight vectors, SMAA reveals the probability that a unit attains a given ranking position, as well as the probability that a given unit is better than another.” If desired, the decision-maker can prioritize the weights, for example $w_{1} \leq \dots \leq w_{M} .$ Tsionas (2020) showed how the analysis in Greco et al. (2019a) can be cast into a Bayesian framework.

Model

Our model consists of Equation (5) plus the following.

y_{t}^{*} = ρ y_{t - 1}^{*} + α (σ_{t}) + x_{t}^{'} β + v_{t},

(5)

where $v_{t}$ is an error term, $β \in R^{K}$ is a vector of unknown coefficients, and $a (σ_{t})$ is a function of the sigma component:

α (σ_{t}) = a_{1} + a_{2} σ_{t} + \frac{1}{2} a_{3} σ_{t}^{2},

where $σ_{t}$ is the SD of the composite indicator (see Greco et al., 2019a).

This is defined as

σ_{t}^{2} = M^{- 1} \sum_{m = 1}^{M} w_{m} {(y_{t, m} - y_{t}^{*})}^{2}

that is, the weighted variance of the different components in the composite indicator. Normally, we would expect $a_{2} > 0$ and $a_{3} < 0$ . Equation (5) defines a mu-sigma frontier between “risk” (σ_t) and “return” ( $μ_{t} = y_{t}^{`} w$ ) in the language of portfolio analysis. Part of the appeal of the sigma-mu frontier model is that we do not need a separate model for the variance ( $σ_{t}$ ) as this follows directly from the definition of the composite indicator.

Notice that $y_{t}^{*}$ does not have the same interpretation as the dynamic common factor $f_{t}$ in Equation (1). The common factor is a “representative” variable of the $y_{t} S$ , whereas $y_{t}^{*}$ is a linear combination of the $y_{t} S$ that allows imposition of prior notions of importance and, in any case, it can be interpreted better as a “dynamic canonical correlation” for lack of a better name.

As our explanatory variables are economic policy variables and our outcomes are entrepreneurship-related variables, we modify Equation (5) as follows.

y_{t}^{*} = ρ y_{t - 1}^{*} + x_{t}^{'} β + v_{t} - u_{t},

(6)

where $u_{t} \geq 0$ is a random variable representing inefficiency on how economic policy affects entrepreneurship. Economic policies are not always designed optimally and the very notion of an “optimal” economic policy is ambiguous. Here, we do not take a specific view on what is an optimal economic policy which, even if it were known to us, was not implemented; so, it is impossible to affect sentiment. Instead, we define a policy frontier as in Equation (6), and we examine how the policy variables affect the outcomes and what the slacks, $u_{t}$ are.

Equation (6) provides a way to summarize the effect of policy variables $x_{t}$ on the dynamic composite indicator $y_{t}^{*}$ and, at the same time, figure out where economic policy introduces systematic inefficiencies in entrepreneurial performance. Though earlier, we discussed the anomalies in aggregation by small-business owners based on Berg and Hoffrage’s (2008) heuristics, as an alternative mental process, may help maximize performance. We make the following distributional assumptions⁷:

\begin{matrix} ε_{t} ∣ x_{t} \sim N (0, σ_{ε}^{2}), v_{t} ∣ x_{t} \sim N (0, σ_{v}^{2}), \\ l n u_{t} = ψ_{o} + ψ_{1} l n u_{t - 1} + x_{t}^{'} γ + η_{t}, \\ η_{t} ∣ x_{t} \sim N (0, σ_{u}^{2}) . \end{matrix}

(7)

We denote all observed data by $Y = {y_{t}, x_{t}}_{t = 1}^{n} .$ Our sample includes 378 year-month observations from January 1986 to June 2017.

The major assumption in Equation (7) is that policy-related inefficiency is dynamic or persistent and also depends on the policy variables $x_{t} .$ . Moreover, $ψ_{o}, ψ_{1}$ and $γ \in R^{K}$ are unknown coefficients. This model allows to figure out the effect of policy variables $x_{t}$ on both the modified dynamic composite indicator $y_{t}^{*}$ as well as on policy-induced entrepreneurial inefficiencies. The objective of the study is to examine systematically how policy variables affect economic sentiment. Inefficiency is defined via the frontier in Equation (6) and answers the question given the policy instruments, how effective was the particular configuration of the instruments? For example, if we find, say, u = 10% on the average, then the particular configuration of the instruments yielded 10% less of what it could, realistically, produce. We do not wish to define optimal economic policy here, which may depend on variables that we do not observe or particular configurations of instruments that are known to policymakers but not to us. Instead, we take a more pragmatic view: Given the instruments and the outcomes and a “technology” that maps the former to the latter, is it the case that instruments have been fully exploited to yield the best possible outcome?

When the number $K$ of explanatory variables $x_{t}$ is large relative to sample size (in our application, we have $K = 56$ , which produces 112 parameters in the indicator and inefficiency equations plus two dynamic parameters and the scale parameters. So, $K$ is roughly one-fourth of our sample size, which is not excessive but we would like, nevertheless, to restrict the number of explanatory variables whose effect is near 0. This is accomplished by using a LASSO prior (Bae & Mallick, 2004; Figueiredo, 2003; Park & Casella, 2008; Tibshirani, 1996; Yuan & Lin, 2005).

To our knowledge, this is the first instance in which a LASSO prior is placed on a dynamic latent variable model including both the latent dynamic modified composite indicator ( $y_{t}^{*}$ ) and latent dynamic inefficiency ( $u_{t}$ ).

Denote collectively all regression-like parameters by $α = [β', γ'] \in R^{d}$ , where $d = 2 K .$ Then, the LASSO prior has the following form:

α ∣ τ = [τ_{1} \dots, τ_{d}]' \sim N_{d} (0_{d}, D_{τ}),

(8)

where $D_{τ} = d i a g [τ_{1}^{2}, \dots, τ_{d}^{2}]$ denotes a diagonal matrix with elements $τ_{1}, \dots, τ_{d}$ along its main diagonal, and

p (τ) \propto \prod_{j = 1}^{d} \frac{λ}{2} e x p {- λ τ_{j}^{2}} .

(9)

Here, $λ$ is a positive unknown smoothing parameter. Posterior conditional parameters of $α, τ, λ$ are in standard families so Gibbs sampling can be performed efficiently. Our prior for $λ$ is $p (λ) \propto λ^{- 1}$ yielding the conditional posterior $p (λ ∣ τ, Y) \propto λ^{d - 1} e x p {- λ \sum_{j = 1}^{d} τ_{j}^{2}}$ , which is a gamma distribution, $G (d, τ^{'} τ)$ (notice that as we remarked before, $Y$ denotes all available data). The smoothing parameter is itself fitted from the data as, in Markov Chain Monte Carlo (MCMC), we can draw from its conditional posterior distribution. Moreover, we should notice that for a linear regression of the form $y = X b + v$ , the LASSO would minimize the objective function

min \underset{b}{(} y - X b)' (y - X b) + \sqrt{λ} \cdot \sum_{j = 1}^{d} ∣ b_{j} ∣ = (y - X b)' (y - X b) + \sqrt{λ} \cdot ∣ ∣ b ∣ ∣_{1},

(10)

for $λ > 0$ and $∣ ∣ b ∣ ∣_{1}$ is the $L_{1}$ -norm of the vector of regression parameters. This problem is not related to Equations (5)—(9) but is given here to define the LASSO. For full details on the model and our MCMC techniques (e.g. Tierney (1994)) we refer to Online Appendix A. Moreover, we should notice that, unlike other priors, the LASSO prior is nondifferentiable at the origin (making it possible to have some coefficients exactly equal to 0 rather than “close to 0” as it would be the outcome from a normal prior).

Data and Methods

We use three datasets: (a) the Small Business Optimism Index from NFIB; (b) the categorical policy uncertainty data (a proxy for uncertainty); and (c) categorical volatility index data (a proxy for risk). All data are monthly data from January 1986 to June 2017.

The NFIB index is comprised of responses from small-business owners on prevailing economic conditions. The index comprises 10 components and the seasonally adjusted optimism is derived from subtracting negative responses from positive responses on seasonally adjusted business expectations. Additional details on the NFIB index are provided in Online Appendix B.

The categorical policy uncertainty data are from Baker et al. (2016) and based on a range of subindices based on results from the Access World News database of over 2000 U.S. newspapers. Each subindex is based on terms related to economy, uncertainty, and policy. The time-series are normalized to have a mean of zero and an SD of one. We provide a list of all the categories in Online Appendix C. The index is used in a variety of recent studies (e.g., Acemoglu et al., 2016; Arellano et al., 2019; Bloom et al., 2018).

The components for volatility (risk) index are the newspaper-based EMV tracker that moves with the Chicago Board Options Exchange Volatility Index (VIX) and with the realized volatility of returns on the S&P 500. The policy-related economic EMV tracker is created for 30 categories every month. The details on the 30 categories are listed in Online Appendix D.I, and additional details on the calculation of the measure are available at https://www.policyuncertainty.com/EMV_monthly.html.

Results

Posterior means and SDs of select parameters are reported in Table 1 and Figure 2. In Figure 2(a)–(c), we report marginal posterior densities of weights that are used in the construction of the new composite indicator. Each of the parts in Figure 2(a)–(c) shows marginal posterior densities for a different set of weights. For the most part, these densities are nonnormal showing the dangers of asymptotic theory applied in our instance. The most important subcomponent is the OPTINDEX (optimism index) in Figure 2(a). An interesting question is whether the weights for the subcomponents can be rankedAltmann et al. (2014). This is taken up in Figure 2(d) where we report marginal posterior cumulative densities. From this evidence, it is easy to see that OPTINDEX stochastically dominates all others, followed by expected business conditions, expected real sales, past earnings, good time expansion, among others. The lion’s share is due to OPTINDEX whose weight ranges from 0.3 to over 0.5 (Figure 2(a)) followed by expected business conditions and expected real sales (Figure 2(b)) whose posterior mean weights are 0.16 and 0.14, respectively.

Table 1

Posterior Moments of $w$ .

	Posterior mean	Posterior SD
Optimism index	0.419	0.0383
Plans to increase employment	0.064	0.0039
Plans to make capital outlays	0.067	0.0083
Plans to increase inventories	0.050	0.0086
Expect economy to improve	0.165	0.0089
Expect real sales higher	0.144	0.0089
Current inventory	0.019	0.0016
Current job openings	0.058	0.0043
Expected credit conditions	0.036	0.0028
Now a good time to expand	0.079	0.0054
Earnings trends	0.096	0.0052

Figure 2

Marginal posterior densities of $w$ .

Before proceeding with our main analysis, we present the distribution based on violin plots of all the explanatory variables, descriptive statistics of all the explanatory variables, and time-series distribution of all the indicators (Online Appendix D). An additional concern is that structural breaks may bias the results. The original authors of the uncertainty and the risk measure adjusted for this possibility; however, we also provide tests for structural breaks in Online Appendix D. We used the difference between each indicator and used the estat sbcusum routine in Stata 16. The test assesses whether the coefficients of a time-series regression are stable over time by constructing a cumulative sum of recursive residuals. In Online Appendix D, we present structural breaks plots for the main uncertainty and risk variables, and then we present the plots for all the remaining explanatory variables. Note, if the line is within the confidence interval, the null hypothesis of “no structural break” is not rejected. In the figures, all the explanatory variables are within the confidence interval.

In Table E.1 of Online Appendix E, we report the results of estimating the dynamic factor model in Equation (1) using maximum likelihood state-space models. In turn, we use the new model to provide marginal posterior densities of the modified composite indicator shown in Figure 3. Figure 4 presents the sample distribution of posterior mean fits of inefficiency.

Figure 3

Posterior means. (a) Posterior mean fits of the modified composite indicator. (b) Risk ( $σ_{t}$ ) of the composite indicator.

Figure 4

Posterior mean policy inefficiency.

Our inferences are based on Table 2. We note that all the variables are standardized for the analysis; the optimism index is not standardized as the model forces it to be a sum of 1. Before discussing the results, we reiterate that inclusion of combined and individual dimensions is not subject to multicollinearity concerns in LASSO fits. For mean estimation, $ρ$ is controlled for and also to control past errors in combinations to derive optimism index, $ψ$ is controlled for. In interpreting the coefficients for $y_{t}^{}$ the equation is similar to a typical regression fit, whereas the fit of inefficiency equation represents the variation in deriving fits (i.e., inefficiency in aggregating $y_{t}^{}$ ). In interpreting the effect sizes, we note that our explanatory variable and outcome variables are standardized, and as such, the interpretation is a unit increase in the SD of explanatory variable results in the corresponding increase or decrease in the SD of optimism index. Our discussion of effect sizes is based on a one-tenth increase in the SD of the explanatory variable.

Table 2

Posterior Moments of $β$ and $γ$ .

	$y_{t}^{*}$ equation	Inefficiency equation
$ρ$	0.714 (0.016)	–
$ψ_{1}$	–	0.818 (0.019)
Economic policy uncertainty	–0.082 (0.011)	0.121 (0.015)
Monetary policy	0.0032 (0.0004)	0.004 (0.001)
Fiscal policy (taxes or spending)	0.0040 (0.0010)	0.017 (0.0021)
Taxes	–0.093 (0.006)	0.034 (0.005)
Government spending	0.0015 (0.0003)	0.017 (0.006)
Health care	–0.0032 (0.0011)	–0.013 (0.002)
National security	0.0000	–0.0032 (0.001)
Entitlement programs	0.0000	0.0044 (0.001)
Regulation	–0.014 (0.0005)	0.0052 (0.0013)
Financial regulation	0.017 (0.003)	–0.0033 (0.0012)
Trade policy	0.0000	–0.0052 (0.0013)
Sovereign debt, currency crises	–0.0144 (0.003)	0.0255 (0.007)
Overall EMV tracker	0.344 (0.021)	0.0000
Policy-related EMV tracker	0.174 (0.021)	0.0000
Macroeconomic news and outlook EMV tracker	0.0000	0.0014 (0.0002)
Macro—broad quantity indicators EMV tracker	0.0000	0.0035 (0.001)
Macro—inflation EMV indicator	–0.022 (0.006)	0.035 (0.006)
Macro—interest rates EMV tracker	–0.014 (0.002)	0.012 (0.003)
Macro—other financial indicators EMV tracker	0.0000	0.044 (0.016)
Macro—labor markets EMV tracker	–0.0032 (0.0013)	0.0017 (0.0005)
Macro—real estate markets EMV tracker	0.0000	–0.0013 (0.0002)
Macro—trade EMV tracker	0.0044 (0.0010)	0.0000
Macro—business investment and sentiment EMV tracker	0.313 (0.015)	–0.022 (0.006)
Macro—consumer spending and sentiment EMV tracker	0.355 (0.018)	–0.162 (0.023)
Commodity markets EMV tracker	0.0000	0.0000
Financial crises EMV tracker	–0.014 (0.002)	0.022 (0.004)
Exchange rates EMV tracker	0.0000	0.0044 (0.0012)
Health care matters EMV tracker	0.0000	0.0013 (0.0004)
Litigation matters EMV tracker	0.0000	0.0000
Competition matters EMV tracker	0.0000	0.0022 (0.0005)
Labor disputes EMV tracker	0.0000	0.0000
Intellectual property matters EMV tracker	0.0013 (0.0002)	0.0012 (0.0005)
Fiscal policy EMV tracker	0.0000	0.0022 (0.0019)
Taxes EMV tracker	–0.044 (0.005)	0.032 (0.007)
Government spending, deficits, and debt EMV tracker	0.0000	0.041 (0.013)
Entitlement and welfare programs EMV tracker	0.0000	0.0014 (0.001)
Monetary policy EMV tracker	0.0000	0.0000
Regulation EMV tracker	0.0000	0.0144 (0.003)
Financial regulation EMV tracker	0.0044 (0.0012)	–0.0032 (0.0011)
Competition policy EMV tracker	0.0000	0.0022 (0.0004)
Intellectual property policy EMV tracker	0.0032 (0.0004)	0.0000
Labor regulations EMV tracker	0.0000	0.0000
Immigration EMV tracker	0.0000	0.0000
Energy and environmental regulation EMV tracker	–0.014 (0.004)	0.017 (0.005)
Lawsuit and tort reform, supreme court decisions EMV tracker	0.0000	0.0000
Housing and land management EMV tracker	0.0000	0.0000
Other regulation EMV tracker	0.0000	0.0000
National security policy EMV tracker	0.0000	0.0000
Government-sponsored enterprises EMV tracker	0.0000	0.0000
Trade policy EMV tracker	0.0014 (0.0002)	0.0000
Health care policy EMV tracker	–0.0013 (0.0003)	0.0000
Food and drug policy EMV tracker	0.0000	0.0000
Transportation, infrastructure, and public utilities EMV tracker	0.0000	0.0000
Elections and political governance EMV tracker	0.0041 (0.0013)	0.0000
Agricultural policy EMV tracker	0.0000	0.0000
General economic policy uncertainty current	0.0000	0.0000
σ_t	0.035 (0.0045)	0.051 (0.006)
$σ_{t}^{2}$	−0.014 (0.003)	−0.017 (0.004)

Note. Zero entries mean that the respective coefficients are zero as we use a LASSO prior. EMV = Equity Market Volatility.

Related to uncertainty signals, a one-tenth increase in SD in economic policy uncertainty has a negative association with optimism by −0.0082 SDs; however, there is also greater variability in how economic uncertainty is perceived (inefficiency = .0121). The effect size is meaningful in the sense that the optimism index has a posterior SD of 0.0383 (Table 1), indicating that a one-tenth SD change is feasible, and the corresponding decline of −0.0082 represents a 21.4% change in SD relative to the posterior SD.

Similarly, tax-related uncertainty has a negative association with optimism (−0.0093, or 24.89% of the change in SD relative to the posterior SD) and a much smaller variability (= .034). Surprisingly, monetary or fiscal policy is not associated with optimism and neither is health care policy uncertainty, especially given the lack of insurance for self-employed and insurance as a key consideration for small firm owners and employees. Regulation is negatively associated with optimism by a small amount; however, financial regulation is positively associated with optimism by a small amount.

Related to risk signals, policy-related EMV tracker has a positive association with optimism (= .0344 for a one-tenth SD of the predictor, or 89.81% of the change in SD relative to the posterior SD). Followed by the macro–business investment and sentiment EMV tracker also with a positive association (= .0313, with small inefficiency; 81.72% of the change in SD relative to the posterior SD). The macro–consumer spending and sentiment EMV tracker has a positive association with optimism (= .0355, and with lower inefficiency = −0.162; or 92.68% of the change in SD relative to the posterior SD). Several other trackers of volatility have a small to negligible association with optimism.

As an additional test, in Online Appendix F, we specify OLS, partialling out LASSO linear regression, cross-fit partialling out LASSO linear regression, and the post-double selection and post-regularization LASSO. The economic policy uncertainty is negative and consistent with the main inferences. Tax-related uncertainty is negative and nonsignificant in the OLS; however, for the remaining LASSO specifications (models 1–4), it is positive but nonsignificant. The policy-related EMV tracker has a nonsignificant effect, the macro–business investment and sentiment tracker has a positive effect, and the macro–consumer spending tracker has a nonsignificant effect.

Overall, based on the proposed analytical model and the general OLS and standard LASSO regression (Online Appendix F) related to uncertainty, business owners respond negatively to economic policy uncertainty and tax uncertainty. Related to risk, stock market volatility increases optimism, specifically, with macro–business investment and sentiment EMV tracker and macro– consumer spending and sentiment EMV tracker. Considering both the main specification and the additional analysis in Online Appendix F, the most consistent effect is the negative effect of economic policy uncertainty and the positive effect of the business investment and sentiment EMV tracker.

Conclusions and Future Research

For this research note, our goal was twofold—delving into drivers of optimism among small-business owners and assessing how business owner optimism is driven by risk and uncertainty. The proposed estimation approach lowers concerns for prior learning by modeling for Bayesian fits and accounts for unobserved considerations of aggregate and component-based dimensions (LASSO). The findings have the following theoretical implications.

First, the assessment of indices in entrepreneurship could consider the value of the proposed estimation approach. The decision contexts of entrepreneurship in the words of Knight (1921, p. 210–211) are an important consideration:

The ordinary decisions of life are made on the basis of ‘fits’ of a crude and superficial character. In general the future situation in relation to which we act depends upon the behavior of an indefinitely large number of objects, and is influenced by so many factors that no real effort is made to take account of them all, much less to fit and summate their separate significances. It is only in very special and crucial cases that anything like a mathematical (exhaustive and quantitative) study can be made. (as quoted in Rakow, 2010)

The quote from Knight is crucial in considering how entrepreneurs consider, report, and aggregate risk- and uncertainty-related signals. Our approach is a preliminary attempt to model some of the limitations in such aggregates to provide a nuanced view of entrepreneurial response to risk and uncertainty in developing an optimistic outlook. The aggregate mental calculus of developing optimism index could be based on collective intuitions, contagion, or herding, or any other group phenomenon. However, such mechanisms are difficult to measure and are intractable. Our estimation has its limitations; however, we focus on the collective aggregations to understand the associations among the components and we do not attempt to address the process of aggregation. Our approach complements that of Anglin et al. (2018) at the individual–industry level factors and focuses on risk and uncertainty indicators driving small business optimism index that is widely disseminated and representative of respondents from a variety of industries and regions.

Second, risk and uncertainty are studied, mostly, separately in the entrepreneurship literature (cf. Koudstaal et al., 2016). Devised as “unknown, unknown” and “known, unknown,” literature streams on uncertainty and risk have developed mostly on separate trajectories. For example, effectuation can be considered a mode of uncertainty resolution, whereas alertness could be a mode for opportunity search based on the available distribution of information in the Kirznerian framework. In our model, we relax the assumption by considering both risk- and uncertainty-related signals as they co-occur in an economy in a variety of indicators (Williamson, 1998). Our measures of uncertainty and risk are based on recently developed widely used measures. Broadly, the results demonstrate that small-business owners have lower optimism under uncertainty, but higher optimism for risky signals from the economy.

In addition to the indicators we found support for, additional discussion on the nonsignificant effects is warranted. We found that monetary and fiscal policies, government spending, health care, national security, entitlement programs, regulation, and financial regulation had negligible to no effect. Similarly, trade policy uncertainty had no effect and the sovereign debt crisis had a small effect. These nonsignificant or significant findings with negligible effects could be interpreted based on Williamson’s (1998) notions of institutions of governance.⁸ Governance is a “means by which order is accomplished in a relationship in which potential conflict threatens to undo or upset opportunities to realize mutual gains” (Williamson, 1998, p. 37). Based on the classical Northian notion of institutions helping set the rules of the game, it is possible that relaxed monetary and fiscal policies over the last two decades may not be salient to small-business owners (North, 1990). Cheap debt and low inflation in the United States may have lowered the importance of these factors. Similarly, health care programs have been the focal point depending on the administrations and with limited, if any, changes in health care costs for self-employed even after Obamacare, uncertainty in health care may not influence optimism. Health care, fiscal, and monetary policy along with regulation and financial regulations are more policy-oriented and are now increasingly embedded institutionally in their reach and influence and may, therefore, be less of a factor based on uncertainty. Finally, entitlement programs and national security are perhaps far removed from small-business owner decisions. Conversely, economic policy uncertainty and the positive effect of the business investment and sentiment EMV tracker may be more transient than institutions, leading to greater sensitivity toward optimism. The notional substitution of institutionalized fiscal, monetary, and health care policies may reduce the role of these uncertainties in influencing optimism. We caution that our findings may have a limited external validity to non-U.S. contexts or additional uncertainty and volatility indicators not included in the model. Future studies can test whether our findings hold in the non-U.S. context and in alternate economic institutions (e.g., developing country or welfare economy), and may also consider additional indicators and study business optimism under extreme events (e.g., COVID-19).

Third, the findings have implications for policymakers focused on the role of multiple indicators that may calibrate small business sentiment. In addition to the significance of the effects, also of importance is the nonsignificance of the multitude of indicators in the results. Several of the indicators, including health care, labor laws, and environmental regulations, are not associated with optimism. Based on methodologies such as ours perhaps calibration of small-business owner optimism with other observable macroeconomic indicators can be useful.

Our study is not without limitations. We draw on the U.S.-based optimism index; however, due to variations in cultural and institutional dimensions, the findings may not be generalizable. We draw upon and build over a recent method on the mean-efficiency index to assess how the aggregation of composite indicators can be better modeled. Nevertheless, future studies can develop models. For example, $σ - μ$ efficiency analysis can be applied in addition to the latent variable models we use here if individual responses are available. This might be important as it would allow us to classify firms in the $σ - μ$ space and determine policy effectiveness more precisely. Additionally, our models can be related (directly, more or less) to the “multiple causes blessing” model of Wang and Blei (2019) and establish causal relations with more precision. According to Wang and Blei (2019), a strong assumption in causal inference from observational data is that variables influencing the causal variables and the outcome variables are observed, a standard but an untestable assumption. By including a variety of uncertainty and volatility indicators, we lower concerns related to the strength of such assumption.

Finally, we used newspaper-based uncertainty and risk measures that are publicly available and used in several studies. However, future studies could also use survey-based measures or richer measures based on text analysis from alternate sources including social media. Though we rely on the measures of economic uncertainty and volatility from http://www.policyuncertainty.com/, additional measures of uncertainty based on commodity, gold, oil, and energy pricing uncertainties can be used in future studies (e.g., Balli et al., 2019; Naeem et al., 2020).

In closing, based on the monthly optimism index and uncertainty and volatility trackers over 32 years, we found that economic and tax policy uncertainty and tax uncertainty increased optimism, and surprisingly, monetary or fiscal policy, health care policy, or regulatory uncertainty did not impact optimism. Aggregate volatility, representing risk in the markets, increases optimism. For small-business owners, the findings highlight a positive response to risk but a negative response to uncertainty.

Supplemental Material

Supplementary Material 1 - Supplemental material for Macroeconomic Uncertainty and Risk: Collective Optimism of Small-Business Owners

Supplemental material, Supplementary Material 1, for Macroeconomic Uncertainty and Risk: Collective Optimism of Small-Business Owners by Pankaj C. Patel and Mike Tsionas in Entrepreneurship Theory and Practice

Footnotes

Acknowledgements

The authors thank Karl Wennberg, senior editor, and the three anonymous reviewers for their excellent comments and suggestions.

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

ORCID iDs

Pankaj C. Patel

Mike Tsionas

Supplemental Material

Supplemental material for this article is available online.

Notes

Author Biographies

Pankaj C. Patel is a professor of management at Villanova University. His research interests are at the intersection of technology and governance. He received his PhD from the University of Louisville.

Mike Tsionas is a Professor of Economics at Montpellier Business School and Lancaster University. His research interests are in the area of econometrics. He received his PhD from the University of Minnesota.

References

Acemoglu

Johnson

Kermani

Kwak

Mitton

. (2016). The value of connections in turbulent times: Evidence from the United States. Journal of Financial Economics, 121(2), 368–391.https://doi.org/10.1016/j.jfineco.2015.10.001

Altmann

McLaughlin

Dobigeon

. (2014). Sampling from a multivariate Gaussian distribution truncated on a simplex: a review. In 2014 IEEE Workshop on Statistical Signal Processing (SSP) (pp. 113–116). IEEE.

Anglin

A. H.

McKenny

A. F.

Short

J. C

. (2018). The impact of collective optimism on new venture creation and growth: A social contagion perspective. Entrepreneurship Theory and Practice, 42(3), 390–425.https://doi.org/10.1111/etap.12256

Arellano

Bai

Kehoe

P. J

. (2019). Financial frictions and fluctuations in volatility. Journal of Political Economy, 127(5), 2049–2103.https://doi.org/10.1086/701792

Åstebro

Herz

Nanda

Weber

R. A

. (2014). Seeking the roots of entrepreneurship: Insights from behavioral economics. Journal of Economic Perspectives, 28(3), 49–70.https://doi.org/10.1257/jep.28.3.49

Bae

Mallick

B. K

. (2004). Gene selection using a two-level hierarchical Bayesian model. Bioinformatics, 20(18), 3423–3430.https://doi.org/10.1093/bioinformatics/bth419

Baker

S. R.

Bloom

Davis

S. J

. (2016). Measuring economic policy uncertainty. The Quarterly Journal of Economics, 131(4), 1593–1636.https://doi.org/10.1093/qje/qjw024

Balli

Naeem

M. A.

Shahzad

S. J. H.

de Bruin

. (2019). Spillover network of commodity uncertainties. Energy Economics, 81, 914–927.https://doi.org/10.1016/j.eneco.2019.06.001

Bennett

. (2011). Cultures of optimism. Cultural Sociology, 5(2), 301–320.https://doi.org/10.1177/1749975511401270

10.

Berg

. (2014). Success from satisficing and imitation: Entrepreneurs’ location choice and implications of heuristics for local economic development. Journal of Business Research, 67(8), 1700–1709.https://doi.org/10.1016/j.jbusres.2014.02.016

11.

Berg

Hoffrage

. (2008). Rational ignoring with unbounded cognitive capacity. Journal of Economic Psychology, 29(6), 792–809.https://doi.org/10.1016/j.joep.2008.03.003

12.

Bloom

Floetotto

Jaimovich

Saporta-Eksten

Terry

S. J

. (2018). Really uncertain business cycles. Econometrica, 86(3), 1031–1065.https://doi.org/10.3982/ECTA10927

13.

Buchanan

J. M.

Vanberg

V. J

. (2002). Constitutional implications of radical subjectivism. The Review of Austrian Economics, 15(2–3), 121–129.https://doi.org/10.1023/A:1015754302238

14.

Christiansen

Eriksen

J. N.

Møller

. (2014). Forecasting us recessions: The role of sentiment. Journal of Banking & Finance, 49, 459–468.https://doi.org/10.1016/j.jbankfin.2014.06.017

15.

Conrad

. (2001). Genetic optimism: Framing genes and mental illness in the news. Culture, Medicine and Psychiatry, 25(2), 225–247.https://doi.org/10.1023/A:1010690427114

16.

Cooper

A. C.

Woo

C. Y.

Dunkelberg

W. C

. (1988). Entrepreneurs’ perceived chances for success. Journal of Business Venturing, 3(2), 97–108.https://doi.org/10.1016/0883-9026(88)90020-1

17.

Crane

F. G.

Crane

E. C

. (2007). Dispositional optimism and entrepreneurial success. The Psychologist-Manager Journal, 10(1), 13–25.https://doi.org/10.1080/10887150709336610

18.

Dew

Read

Sarasvathy

S. D.

Wiltbank

. (2008). Outlines of a behavioral theory of the entrepreneurial firm. Journal of Economic Behavior & Organization, 66(1), 37–59.https://doi.org/10.1016/j.jebo.2006.10.008

19.

Doumpos

Gaganis

Pasiouras

. (2016). Bank diversification and overall financial strength: International evidence. Financial Markets, Institutions & Instruments, 25(3), 169–213.https://doi.org/10.1111/fmii.12069

20.

Doumpos

Hasan

Pasiouras

. (2017). Bank overall financial strength: Islamic versus conventional banks. Economic Modelling, 64, 513–523.https://doi.org/10.1016/j.econmod.2017.03.026

21.

Dunkelberg

W. C.

Wade

. (2009). NFIB small business economic trends. National Federation of Independent Business, 5, 19.

22.

Fairlie

R. W.

Holleran

. (2012). Entrepreneurship training, risk aversion and other personality traits: Evidence from a random experiment. Journal of Economic Psychology, 33(2), 366–378.https://doi.org/10.1016/j.joep.2011.02.001

23.

Figueiredo

M. A. T

. (2003). Adaptive sparseness for supervised learning. IEEE Transactions on Pattern Analysis and Machine Intelligence, 25(9), 1150–1159.https://doi.org/10.1109/TPAMI.2003.1227989

24.

Greco

Mousseau

Słowiński

. (2008). Ordinal regression revisited: Multiple criteria ranking using a set of additive value functions. European Journal of Operational Research, 191(2), 416–436.https://doi.org/10.1016/j.ejor.2007.08.013

25.

Greco

Słowinski

Figueira

Mousseau

. (2010). Robust ordinal regression. In Greco

Matthias

José Rui

(Eds.), Trends in multiple criteria decision analysis (pp. 241–283). Springer.

26.

Greco

Ishizaka

Matarazzo

Torrisi

. (2018). Stochastic multi-attribute acceptability analysis (SMAA): An application to the ranking of Italian regions. Regional Studies, 52(4), 585–600.https://doi.org/10.1080/00343404.2017.1347612

27.

Greco

Ishizaka

Tasiou

Torrisi

. (2019a). Sigma-mu efficiency analysis: A methodology for evaluating units through composite indicators. European Journal of Operational Research, 278(3), 942–960.https://doi.org/10.1016/j.ejor.2019.04.012

28.

Greco

Ishizaka

Tasiou

Torrisi

. (2019b). On the methodological framework of composite indices: A review of the issues of weighting, aggregation, and robustness. Social Indicators Research, 141(1), 61–94.https://doi.org/10.1007/s11205-017-1832-9

29.

Keynes

J. M

. (1937). The general theory of employment. The Quarterly Journal of Economics, 51(2), 209–223.https://doi.org/10.2307/1882087

30.

Kitching

. (2019). Regulatory reform as risk management: Why governments redesign micro company legal obligations. International Small Business Journal: Researching Entrepreneurship, 37(4), 395–416.https://doi.org/10.1177/0266242618823409

31.

Knight

F. H

. (1921). Risk, uncertainty and profit. Houghton-Mifflin.

32.

Koudstaal

Sloof

van Praag

. (2016). Risk, uncertainty, and entrepreneurship: Evidence from a lab-in-the-field experiment. Management Science, 62(10), 2897–2915.https://doi.org/10.1287/mnsc.2015.2249

33.

Naeem

M. A.

Balli

Shahzad

S. J. H.

de Bruin

. (2020). Energy commodity uncertainties and the systematic risk of US industries. Energy Economics, 85, 104589.https://doi.org/10.1016/j.eneco.2019.104589

34.

North

D. C

. (1990). A transaction cost theory of politics. Journal of Theoretical Politics, 2(4), 355–367.https://doi.org/10.1177/0951692890002004001

35.

Park

Casella

. (2008). The Bayesian LASSO. Journal of the American Statistical Association, 103(482), 681–686.https://doi.org/10.1198/016214508000000337

36.

Phillips

G. M

. (2002). Four leading economic confidence measures: Who’s optimism matters? The Journal of Business Forecasting, 21(4), 22.

37.

Rakow

. (2010). Risk, uncertainty and prophet: The psychological insights of Frank H. Knight. Judgment and Decision Making, 5(6), 458.

38.

Santero

Westerlund

. (1996). Confidence indicators and their relationship to changes in economic activity. OECD Publishing. Document Number.

39.

Sarasvathy

S. D

. (2009). Effectuation: Elements of entrepreneurial expertise. Edward Elgar Publishing.

40.

Scheier

M. F.

Carver

C. S

. (1985). Optimism, coping, and health: Assessment and implications of generalized outcome expectancies. Health Psychology, 4(3), 219–247.https://doi.org/10.1037/0278-6133.4.3.219

41.

Simon

H. A

. (1997). Models of bounded rationality: Empirically grounded economic reason (Vol. 3). MIT press.

42.

Snowe

O. J

. (2006). Small business health plans: A critical step in solving the small business health care crisis. Harvard Journal on Legis, 43, 231.

43.

Stock

J. H.

Watson

M. W

. (2012). Generalized shrinkage methods for forecasting using many predictors. Journal of Business & Economic Statistics, 30(4), 481–493.https://doi.org/10.1080/07350015.2012.715956

44.

Storey

D. J

. (2011). Optimism and chance: The elephants in the entrepreneurship room. International Small Business Journal: Researching Entrepreneurship, 29(4), 303–321.https://doi.org/10.1177/0266242611403871

45.

Taleb

N. N

. (2010). Why did the Crisis of 2008 Happen? Available at SSRN 1666042.

46.

Tibshirani

. (1996). Regression shrinkage and selection via the LASSO. Journal of the Royal Statistical Society: Series B, 58(1), 267–288.https://doi.org/10.1111/j.2517-6161.1996.tb02080.x

47.

Tierney

. (1994). Markov chains for exploring posterior distributions. The Annals of Statistics, 22(4), 1701–1728.https://doi.org/10.1214/aos/1176325750

48.

Tsionas

M. G

. (2020). A note on sigma–nu efficiency analysis as a methodology for evaluating units through composite indicators. European Journal of Operational Research, 286(3), 1187–1196.https://doi.org/10.1016/j.ejor.2020.03.076

49.

Wang

Blei

D. M

. (2019). The blessings of multiple causes. Journal of the American Statistical Association, 114(528), 1574–1596.https://doi.org/10.1080/01621459.2019.1686987

50.

Williamson

O. E

. (1998). The institutions of governance. The American Economic Review, 88(2), 75–79.

51.

Woehr

D. J.

Loignon

A. C.

Schmidt

P. B.

Loughry

M. L.

Ohland

M. W

. (2015). Justifying aggregation with consensus-based constructs: A review and examination of cutoff values for common aggregation indices. Organizational Research Methods, 18(4), 704–737.

52.

Yuan

Lin

. (2005). Efficient empirical Bayes variable selection and estimation in linear models. Journal of the American Statistical Association, 100(472), 1215–1225.https://doi.org/10.1198/016214505000000367

53.

Yuh

J.-I.

Neiderhiser

J. M.

Reiss

. (2010). Genetic and environmental influences on dispositional optimism and depressive symptoms in adolescence. International Journal of Human Ecology, 11(2), 15–23.

Supplementary Material

Please find the following supplemental material available below.

For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.

For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.

0.00 MB

3.03 MB

2.33 MB