1.1 Literature Review

This paper relates to studies exploring the relationship between access to financing and impact on growth and investment (Cooley and Quadrini, 2001; Albuquerque and Hopenhayn, 2004; Clementi and Hopenhayn, 2006; Caggese and Cunat, 2013; Buera and Karmakar, 2017; Vaziri, 2021; Ferreira et al., 2021). Empirically, a strand of the literature has investigated the extent to which the degree of financial development in a country has facilitated growth (Rajan and Zingales, 1996). lore recent studies use firm- level data to measure the effect of financial constraints on economic performance (Chen and Guariglia, 2013; Ferrando and Ruggieri, 2018; lanaresi and Pierri, 2019; Cevik and liryugin, 2022).

A challenge faced by the empirical literature is the measurement of access to financing. A common method in measuring these constraints is using firms’ balance sheet information such as cash holdings or firm’s leverage (Chen and Guariglia, 2013; Gomis and Khatiwada, 2017; Levine and Warusawitharana, 2021). However, taking cash holding as an example, it is an endogenous financial choice it is not immediately clear if this variable would always correlate with better access to liquidity. For example, it is possible that firms decide to increase their cash holdings if they do not have access to alternative financing methods. Alternatively, studies have used survey data or exploited exogenous shocks to banks and impact on firms using data on bank to firm linkages (e.g., Buera and Karmakar 2017)). These methods while providing a more exogenous measure, either require data that is not easily available or only cover a smaller subset of firms. To overcome these challenges, we follow the methodology of Ferrando and Ruggieri (2018) combining information from a set of balance sheet variables to mitigate the endogeneity issues without increasing the data requirements. Ferrando and Ruggieri (2018) use Orbis dataset and do a cross-country analysis of Euro area firms finding that financing constraints negatively affect firms’ productivity especially in countries with lower degrees of financial development. With respect to their work, our data set provides a better coverage of young and SME firms. Therefore, the administrative data allows us to further explore the effect of financing constraints on firms’ observable characteristics.

Our paper also contributes to the literature studying the effect of financial constraints on transition and emerging economies. In this respect Gatti and Love (2008) use survey data to study the Bulgarian economy and find that access to credit is positively associate with an improvement in productivity. Chen and Guariglia (2013) study the Chinese manufacturing sector and find that an increase in access to internal finance, measured by cash flow, is associated with an improvement in total factor productivity. To the best of our knowledge, our paper provides the most comprehensive analysis of the impact of financial constraints in the context of transition economies.

Our paper also adds to the discussions around the role of policy in alleviating the impact of financial frictions on firms. Hadlock and Pierce (2010) provide a detailed discussion on validity of various measures of financial constraints collecting detailed qualitative information from firms’ financial filings and combining it with the balance sheet of publicly listed firms in the US using Compustat. They find that age and size have a high prediction power for the level of financial constraints and the effect of these constraints dampens as firms mature and grow.³ Similar to Hadlock and Pierce (2010) our results highlight the importance of firm age and size, however, we also uncover further nonlinearities in this respect. In particular we find that among young firms, larger firms have a higher probability of being constrained while the relationship reverses for older firms. This could be explained by large young firms having a higher propensity to grow even further as they are most likely the very productive firms of the sample. Therefore, our results suggest that by combining information from firms’ age and size, it is possible to target firms more effectively.

3.1 Measure of Financial Constraints

Our first step is to create a clean measure of financial constraints. The challenge, how- ever, is that these constraints are not empirically observable in the firms’ balance sheets. Additionally, financial decisions of firms themselves are endogenous and are not indicative of whether the firm is constrained or not. Further, methods that exogenously capture the financial conditions of the firm often have high data requirements which are not readily available (for example Campello et al. (2010) use survey data, and Buera and Karmakar (2017) and Chodorow-Reich (2014) use credit registry data to create an exogenous shock based on firm to bank relationships).

To overcome these problems, we create a measure of financial constraint from firms’ balance sheet data using information on the asset-side and the liability-side of the balance sheet. Motivated by Ferrando and Ruggieri (2018) and Pal and Ferrando (2010) we classify firms into groups with various degrees of financial constraint. The advantage of this method is that by creating a number of scenarios we can capture both the desire and the ability of firms to invest and thus overcome the common endogeneity problems associated with choosing a single indicator for deciding the extent of financial constraint. A further advantage of this method is its relatively low data requirements along with flexibly adjusting to any data set with balance sheet information. A disadvantage, however, is that it may not be possible to categorize all firms, therefore, leading to a smaller sample size.¹⁰

^{Table 2:}

Classification Scheme

^{Table 2:}

Classification Scheme

Case	Categor y	Financial Gap	∆ Debt	investment	other
Case I	U	cash ≥ investment	> 0	>0	$\frac{investment}{L.asset}>-{\overline{c}}_1$
Case II	U	cash ≥ investment	> 0	<0
Case III	C	cash < investment	> 0	>0
Case IV	C	cash < investment	< 0	>0	$Debtquartile>{\overline{c}}_2$
Case V	C		<0	<0	$\frac{investment}{L.asset}>-{\overline{\overline{c}}}_3$

View Table

^{Table 2:}

Classification Scheme

Case	Categor y	Financial Gap	∆ Debt	investment	other
Case I	U	cash ≥ investment	> 0	>0	$\frac{investment}{L.asset}>-{\overline{c}}_1$
Case II	U	cash ≥ investment	> 0	<0
Case III	C	cash < investment	> 0	>0
Case IV	C	cash < investment	< 0	>0	$Debtquartile>{\overline{c}}_2$
Case V	C		<0	<0	$\frac{investment}{L.asset}>-{\overline{\overline{c}}}_3$

Table 2 contains the classification scheme considering five distinct scenarios ( I – V) and categorizing firms into two different groups of constrained and unconstrained indicated by C and U respectively. Financial gap is defined as fixed investment minus cash flow. For example, case I refers to firms that are investing (desire to invest) and have cash holdings above their investment choice while increasing their debt from the previous years (ability to invest). Case II includes firms who are disinvesting, but they have positive cash holdings and are able to increase their debt. Thus, these firms while having the ability to invest do not have the desire of doing so. The final condition indicated under “other” limits the extent of liquidation of assets by a threshold ${\overline{c}}_1$ and thus ensures that financial decisions of firms are driven by holding above optimal assets rather than financial problems.

Cases III to V represent firms that are financially constrained relative to cases I and II. In case III firms’ internal sources financing is not sufficient therefore they depend on external sources thus making these firms relatively more constrained compared to previous cases. In cases IV and V firms face stronger constraints as they are not even able to increase their debt. In case IV there is an additional condition requiring firms to be on the ${\overline{c}}_2$ debt quartile thus focusing on the subset of firms that depend on debt as their main source of financing instead of equity, where ${\overline{c}}_2$ is a given threshold for debt quartile. Next, we assign score 0 to the unconstrained group “U” and score 1 to the constrained group “C”. To test the validity of the classification scheme Figure 6 in the appendix depicts the share of firms categorized as constrained. As expected in years leading to the Global Financial Crisis the share of constrained firms was at its lowest value due to lax lending conditions. Starting from 2009 the tightening of financial conditions is reflected in this share increasing to its highest value. Since the crisis and in the following years the share of constrained firms has declined and stabilized at roughly 55 percent.

Further, financial constraint shows persistence over time. Table 7 presents the transition matrix for the indicator capturing share of firms moving between different categories from one period to the next. Roughly 53 percent of firms that are unconstrained remain un- constrained the next period, while about 47 percent face financial constraint in the following year. As for the constrained firms, there seems to be more persistence with around 71 percent of firms remaining in the same category and roughly 29 percent becoming unconstrained in the subsequent period.

3.2 Empirical Strategy

The previous subsection discussed how firms are classified as financially constrained using information from the asset side and the liability side of the balance sheet. While this classification measure overcomes certain weaknesses of using a single variable to proxy for such constraints e.g., using collateral captures the ability of firms to borrow but not their desire) at its current form it has limited variability on the degree of access to finance. The objective of this subsection is first to create an index of financial constraint using the classification method described in Table 2. Next, this index can be used to estimate the effect of financial constraint on productivity and investment.

To create a continuous index motivated by the approach of Ferrando and Ruggieri (2018), we use the classification scheme described in the previous subsection to estimate an ordered probit regression. In this regard, we run a regression with the classification scheme as the dependent variable controlling for firms’ characteristics such as age, size, industry, and financial outcomes. The predicted outcome of the regression can be interpreted as the conditional probability of being financially constrained and will be the index used in the remainder of this paper.

The ordered probit is specified as follows for a given firm i at time t: ¹¹

where I_i,t is the classification of firm i at time t according to Table 2. j can take values 0 or 1 with j = 0 referring to firms that are categorized as unconstrained, and j = 1 shows the constrained group. X_i,t-1 are firm level characteristics. To account for firm specific characteristics, we include size dummies based on employment for micro, small, medium, and large firms.¹² We also include dummies for age, industry as well as year to control for business cycles. X_i,t-1 also contains variables on the financial characteristics of firms such as leverage, cash holdings and their interaction with firm size and firm age. All variables relating to firm specific characteristics are lagged one period to avoid simultaneity. Finally, c_i controls for possible correlation between unobserved and time invariant characteristics of firms and X_i,t-1. In doing so we follow Chamberlain (1979) and take steps similar to Ferrando and Ruggieri (2018). The outcome is reported in the next section.

We then use the estimated the model to form the index of financial constraint denoted by FC_it in the remainder of the paper. To verify the validity of our measure we make use of interest payment data on firms’ balance sheets. This information is only available after 2016, and therefore was not originally used to classify firms into different categories. In particular, we expect firms that face higher interest rates have a higher likelihood of being constrained. To test, we find the median interest rate for each sector and each year and calculate the average index

3.3 Financial Constraints, Labor Productivity, and Investment

Financial constraints affect the ability of a firm to borrow in order to invest, improve its production capacity and increase its productivity. This section studies how these constraints relate to firms’ investment and their implications for the firm-level labor productivity.

The measure of labor productivity is calculated as value added of a given firm, deflated, and divided by the number of employees.¹³ To understand the response of firm-level productivity to our measure of financial constraints, we run a regression with the natural logarithm of the firm specific labor productivity as the dependent variable, including a vector of controls for firm characteristics denoted by X^Ch and a vector of controls for firm’s financial characteristics shown by W^FC. The main specification is written as:

where ln(prodv_i,t) is the natural log of labor productivity of firm i and year t, the specification includes firm and sector-year fixed effects denoted by δ_t and δ_st respectively. Standard errors are clustered the level of fixed effect and all explanatory variables are lagged to reduce simultaneity bias. $W_{i,t-1}^{FC}$ includes debt ratio and cash to sales ratio while X^ch contains variables on firm size, firm age, and their interaction with the financial constraint index.

Next, we explore the relationship between financial constraint and firms’ decisions to invest.¹⁴ Investment at time t is defined as the change in fixed assets from time t + 1 to time t plus depreciation. The main specification is given by:

The specification contains firm and sector-year fixed effects and standard errors are clustered as before.

Note that we include in this equation ln(prod_i,t-1) instead of ln(investment_i,t-1) on the right hand side to mitigate attenuation biases, as investment usually is quite volatile.¹⁵ The coefficient of interest is β , estimates the association between financial constraints and investment decision of firms. W^FC includes debt and cash, and X^ch contains information on firm size, firm age, and their interaction with the financial constraint index.¹⁶ Results are discussed in the next section.

4.1 Index for Financial Constraints

Table 4 presents the results. The estimation is based on an ordered probit model using the classification with two outcomes as the explanatory variable. Standard errors are provided in parenthesis and are robust to heteroskedasticity. We report the outcome of the estimation for four different cases, however, the index used throughout the paper is based on the results of column (3) as it has more controls. Columns (1) – (3) exclude health, education, and agriculture sector, as our sample is not representative of the Lithuanian economy for these sectors as discussed in the Data section of this paper. Recall that 0 referred to firms categorized as unconstrained, and 1 to constrained firms. In all three estimations the coefficient on financial leverage is positive and highly significant suggesting that higher debt ratio increases the probability of belonging to the constrained group. On the other hand, the coefficient on the relative cash holding is negative as expected since higher access to liquidity lowers the likelihood of being constrained. In section 4.3 we discuss in details the relationship of age and size to financial constraints.

^{Table 3:}

Summary Statistics: Constrained and Unconstrained Firms

¹High skilled service sectors include education, entertainment, health, and professional services.

²The final column, Other, contains agriculture, mining, electricity, construction, and utilities.

³Assets are presented as log of total real assets of firms.

^{Table 3:}

Summary Statistics: Constrained and Unconstrained Firms

		Manufacturing	IT	Transport	Retail	Services High Skilled1	Services Low Skilled	Other2
Unconstrained
Observation		10,528	1,843	3,958	14,701	5,396	5,606	10,482
Firms		4,660	716	2,035	6,204	2,287	2,628	4,213
Employees
	Mean	72.3	56.8	85.7	61.4	38.1	53.6	58.2
	Median	30	21	22	20	19	21	28
Assets
	Mean	8.9	8.9	9.0	9.0	8.0	7.8	8.9
	Median	8.8	8.8	8.8	8.9	7.8	7.6	8.7
Small		7,123	1,448	3,076	12,283	4,609	4,401	7,411
Medium		2,856	355	716	2,129	712	1,021	2,758
Large		549	40	166	289	75	184	313
Age		10.1	10.0	10.4	10.3	9.6	8.9	9.9
Constrained
Observation		18,401	1,555	9,542	16,628	4,764	6,689	11,808
Firms		4,709	5,42	2,471	5,515	1,881	2,357	3,869
Employees
	Mean	97.7	74.1	78.8	50.4	37.1	57.0	71.7
	Median	37	21	24	25	19	22	31
Assets3
	Mean	8.6	8.6	8.9	9.0	8.2	8.7	7.5
	Median	8.4	8.5	8.8	8.9	8.0	7.5	8.5
Small		10,947	1,125	7,091	13,439	4,078	5,335	7,786
Medium		5,987	353	2,070	2,888	622	1,190	3,516
Large		1,467	77	381	301	64	274	506
Age
	Mean	10.0	10.1	10.5	9.9	9.1	8.7	9.5

¹High skilled service sectors include education, entertainment, health, and professional services.

²The final column, Other, contains agriculture, mining, electricity, construction, and utilities.

³Assets are presented as log of total real assets of firms.

View Table

^{Table 3:}

Summary Statistics: Constrained and Unconstrained Firms

		Manufacturing	IT	Transport	Retail	Services High Skilled1	Services Low Skilled	Other2
Unconstrained
Observation		10,528	1,843	3,958	14,701	5,396	5,606	10,482
Firms		4,660	716	2,035	6,204	2,287	2,628	4,213
Employees
	Mean	72.3	56.8	85.7	61.4	38.1	53.6	58.2
	Median	30	21	22	20	19	21	28
Assets
	Mean	8.9	8.9	9.0	9.0	8.0	7.8	8.9
	Median	8.8	8.8	8.8	8.9	7.8	7.6	8.7
Small		7,123	1,448	3,076	12,283	4,609	4,401	7,411
Medium		2,856	355	716	2,129	712	1,021	2,758
Large		549	40	166	289	75	184	313
Age		10.1	10.0	10.4	10.3	9.6	8.9	9.9
Constrained
Observation		18,401	1,555	9,542	16,628	4,764	6,689	11,808
Firms		4,709	5,42	2,471	5,515	1,881	2,357	3,869
Employees
	Mean	97.7	74.1	78.8	50.4	37.1	57.0	71.7
	Median	37	21	24	25	19	22	31
Assets3
	Mean	8.6	8.6	8.9	9.0	8.2	8.7	7.5
	Median	8.4	8.5	8.8	8.9	8.0	7.5	8.5
Small		10,947	1,125	7,091	13,439	4,078	5,335	7,786
Medium		5,987	353	2,070	2,888	622	1,190	3,516
Large		1,467	77	381	301	64	274	506
Age
	Mean	10.0	10.1	10.5	9.9	9.1	8.7	9.5

¹High skilled service sectors include education, entertainment, health, and professional services.

²The final column, Other, contains agriculture, mining, electricity, construction, and utilities.

³Assets are presented as log of total real assets of firms.

^{Table 4:}

Index of Financing Constraints

Notes: Standard errors are in parentheses and are robust to heteroskedasticity. The dependent variable is the index created according to table 2. All specifications exclude firms with fewer than 10 employees and specifications (1) -(3) exclude health, education, and agriculture sectors. Time and sector dummies are included for all specifications, and we use Chamberlain (1979) to control for unobserved characteristics. Age groups are defined as 0–5, 6–10, 11–15, 15+ years. Size groups are defined as 10– 49, 50–249, 250+ employees.

^{Table 4:}

Index of Financing Constraints

	1)	2)	3)	4)
L.debt ratio	0.850***	0.853***	0.751***	0.532***
L cash/sales	(0.0267)	(0.0267)	(0.0384)	(0.0274)
	-1.983*** (0.103)	-1.976*** (0.103)	-1.570*** (0.199)	-1.029*** (0.120)
L.age group	✓	✓	✓	✓
Lsi:ze	✓	✓	✓	✓
L.si:ze× Lage group		✓	✓	✓
Lsi:ze × L.debt ratio			✓	✓
Lsi:ze × L.cash/sales			✓	✓
Lage group × L.debt ratio			✓	✓
Lage group × L.cash/sales			✓	✓
Obs	77636	77636	77636	107917
* p < .10, **p < .05, *** p < .01.

Notes: Standard errors are in parentheses and are robust to heteroskedasticity. The dependent variable is the index created according to table 2. All specifications exclude firms with fewer than 10 employees and specifications (1) -(3) exclude health, education, and agriculture sectors. Time and sector dummies are included for all specifications, and we use Chamberlain (1979) to control for unobserved characteristics. Age groups are defined as 0–5, 6–10, 11–15, 15+ years. Size groups are defined as 10– 49, 50–249, 250+ employees.

View Table

^{Table 4:}

Index of Financing Constraints

	1)	2)	3)	4)
L.debt ratio	0.850***	0.853***	0.751***	0.532***
L cash/sales	(0.0267)	(0.0267)	(0.0384)	(0.0274)
	-1.983*** (0.103)	-1.976*** (0.103)	-1.570*** (0.199)	-1.029*** (0.120)
L.age group	✓	✓	✓	✓
Lsi:ze	✓	✓	✓	✓
L.si:ze× Lage group		✓	✓	✓
Lsi:ze × L.debt ratio			✓	✓
Lsi:ze × L.cash/sales			✓	✓
Lage group × L.debt ratio			✓	✓
Lage group × L.cash/sales			✓	✓
Obs	77636	77636	77636	107917
* p < .10, **p < .05, *** p < .01.

Notes: Standard errors are in parentheses and are robust to heteroskedasticity. The dependent variable is the index created according to table 2. All specifications exclude firms with fewer than 10 employees and specifications (1) -(3) exclude health, education, and agriculture sectors. Time and sector dummies are included for all specifications, and we use Chamberlain (1979) to control for unobserved characteristics. Age groups are defined as 0–5, 6–10, 11–15, 15+ years. Size groups are defined as 10– 49, 50–249, 250+ employees.

4.2 Financial Constraints, Labor Productivity, and Investment

Table 5 presents the relationship between financial constraints, productivity and investment of firms using the specification discussed in the previous section. Table 9 and Table 10 in the appendix include the full regression outcomes for productivity and investment respectively. As expected, financial constraints limit firms’ investment and lead to lower productivity. This is in line with findings of Gatti and Love (2008) for Bulgarian firms, and Chen and Guariglia (2013) for Chinese firms.

^{Table 5:}

Financial Constraints, Productivity, and Investment

Standard errors in parentheses. * p < .10, ** p < .05, *** p < .01.All specifications include time-sector FE and control for financial characteristics (firm debt and cash) and firm unobserved characteristics with firm FE.

^{Table 5:}

Financial Constraints, Productivity, and Investment

	(1)	(2)
	ln (productivity)	ln (investment)
LFC	-0.0585*** (0.0203)	-0.821*** (0.0818)

Standard errors in parentheses. * p < .10, ** p < .05, *** p < .01.All specifications include time-sector FE and control for financial characteristics (firm debt and cash) and firm unobserved characteristics with firm FE.

View Table

^{Table 5:}

Financial Constraints, Productivity, and Investment

	(1)	(2)
	ln (productivity)	ln (investment)
LFC	-0.0585*** (0.0203)	-0.821*** (0.0818)

Standard errors in parentheses. * p < .10, ** p < .05, *** p < .01.All specifications include time-sector FE and control for financial characteristics (firm debt and cash) and firm unobserved characteristics with firm FE.

Table 5 suggests that the elasticity of natural logarithm of labor productivity to the measure of financial constraints is -0.058. To get a better sense of the estimated elasticity, we use the point estimate to study a counterfactual scenario in which firms’ financial constraints are removed. This scenario is equivalent to setting the index of financial constraints FC for all firms to the average value of the index calculated for firms categorized in the unconstrained group. In particular, we find that the index is on average 0.15 points higher for the constrained class with respect to the unconstrained class. Further, 59 percent of firms are categorized as constrained. We use these values to calculate the response of labor productivity to removal of financial constraints, given firm characteristics, in the Lithuanian firms:

0.59 * 0.15*0.058 ~ 0.51%

Therefore, we find that under a scenario in which firms are not financially constrained, average firm level investment is expected to increase significantly by roughly 7.2 percent.

Besides these counterfactual scenarios, as shown in Table 9 and Table 10 in the appendix, past productivity is positively related to current productivity and investment, higher debt, and higher cash flow both increase investment and productivity. Investment is increasing in firm size and firm age, though the estimates for are not significant at 10 percent significance level, while size has a high explanatory power.

4.4 Financial Constraints, Firm Age, and Size

Firm size and firm age are negatively related to the measure of financial constraints in line with the finding of literature that smaller and younger firms have more difficulty getting access to liquidity to finance their investments. To explore this further, specification (2) of Table 4 includes the interaction of age with size and Table 6 summarizes these values. The number reported in each cell is the sum of the coefficients when the respected dummies are equal to 1. For example, the value reported in size large and age group 0–5 years, is the sum of dummies for large=1, age group 0–5 =1, and the interaction of the two terms. Note that no value is reported for small and age group 0–5 as these are the reference groups. Next, Table 6 suggests that for each size class, age and financial constraints are negatively related, i.e., younger firms of a given size group are more likely to be financially constrained.

As in Hadlock and Pierce (2010) our paper highlights the importance of firms’ age and size in predicting the probability of being constrained. Additionally, the coverage of our dataset of SME and young firms allows us to uncover further trends that to the best of our knowledge have not been identified by the literature. We find that size, age, and financial constraints display a non-linear relationship. In particular, for young firms age group 0- 5), large firms are more likely to be financially constrained. This result is directly related to our classification scheme which captured both the desire and ability of firms to invest. Among young firms, while those that are large may be able to borrow more, it is likely that they are not getting the amount they desire to reach their optimal size while smaller firms may not be able to borrow as much but it is possible that their desire to borrow is not as strong. As firms grow older in particular for ages above 10 smaller firms are more likely to be financially constrained. Finally, the interaction of debt ratio with size suggests that higher debt leads to a higher probability of being constrained for smaller firms, while old firms having higher debt ratio have a higher probability of being constrained. The interaction of cash with age and size does not have a high explanatory power.

^{Table 6:}

Effect of Age and Size on the Likelihood of Being Constrained

p < .10, ** p < .05, *** p < .01. This table shows the coefficients on firm age and firm si:1e as presented is specification (2) of Table 4. The reference group is small and young firms. Size is based on number of employees, and definitions are presented in parenthesis.

^{Table 6:}

Effect of Age and Size on the Likelihood of Being Constrained

	0–5	6–10	11–15	15+
Small (10–49)	-	0.031*	-0.005	-0.057*
Med (50–249)	0.108*”	0.102	0.028***	-0.082”
Large (250+)	0.091	0.070	-0.038*	-0.223*”

p < .10, ** p < .05, *** p < .01. This table shows the coefficients on firm age and firm si:1e as presented is specification (2) of Table 4. The reference group is small and young firms. Size is based on number of employees, and definitions are presented in parenthesis.

View Table

^{Table 6:}

Effect of Age and Size on the Likelihood of Being Constrained

	0–5	6–10	11–15	15+
Small (10–49)	-	0.031*	-0.005	-0.057*
Med (50–249)	0.108*”	0.102	0.028***	-0.082”
Large (250+)	0.091	0.070	-0.038*	-0.223*”

p < .10, ** p < .05, *** p < .01. This table shows the coefficients on firm age and firm si:1e as presented is specification (2) of Table 4. The reference group is small and young firms. Size is based on number of employees, and definitions are presented in parenthesis.