3.1 Model set-up

The framework is a 3-period model combining motive for cash holdings as in Holmström and Tirole (1998) and exporter selection with heterogenous firms as in Melitz (2003). In particular, in period t=0, an incumbent domestic firm has a given level of productivity ϕ₀ and realizes per period profit π(ϕ₀), where π is a non-decreasing continuous function in firm productivity which we will further specify below. The firm decides whether to invest in innovation to upgrade its productivity to a higher level ϕ ≥ ϕ₀ at fixed cost I. Importantly, we abstract from firm entry and exit in the domestic market.

In the intermediate stage t=1, if the firm has invested in innovation, it is exposed to a stochastic liquidity shock in the magnitude ρ which is distributed according to the cdf F (ρ) on the support [0, ∞). This liquidity shock can be seen as a cost over-run or a stochastic reinvestment need. If the firm does not pay ρ, its investment is lost and it reverts to its status quo. If it pays ρ, it survives until the next period. We assume that when the liquidity shock hits, the firm cannot rely on borrowing or raising capital sufficiently fast to entirely cover the cost overrun. Therefore, the funds to cover the liquidity shock need to be hoarded as cash in advance.⁶

Upon surviving the liquidity shock, the firm reaps the benefit of its innovation investment in period t=2 by drawing a new productivity from a Pareto distribution with density function g (ϕ) over the support [ϕ₀, ∞), and shape parameter κ which pins down the dispersion of the distribution. A lower κ represents a thicker upper-tail distribution and hence a higher probability of drawing a high productivity.

After drawing the new productivity, the firm has an opportunity to become an exporter. Exporting requires paying fixed costs f_X and variable (iceberg) costs τ > 1 as in Melitz (2003). Therefore, if the firm’s productivity draw is above a cutoff value, it will serve both the domestic and foreign markets, whereas if it is below the cutoff, it continues to serve the domestic market only, but still operates at higher productivity ϕ ≥ ϕ₀ and realizes higher profit relative to not innovating.

The timing of the events are illustrated in Figure 8.

View Full Size

Timing of events in the baseline 3-period model.

Citation: IMF Working Papers 2019, 017; 10.5089/9781484392164.001.A001

• Download Figure
• Download figure as PowerPoint slide

3.2 Model solution

We solve the model backwards by first considering the exporter selection stage t = 2. As in Melitz (2003), we assume consumers in both the domestic and export markets have the same CES utility over a continuum of substitutable goods with elasticity of substitution σ > 1, and producers (firms) being monopolistically competitive. Conditional on drawing productivity ϕ, the firm’s profit as a function of its new productivity is given by its profit in domestic and export markets (see derivation in Appendix):

with M, M^X being composite terms reflecting total demand in domestic and export markets, taken as given by the firm, and f_X and τ being the fixed and variable cost of exporting as introduced above. Since profits are increasing in ϕ, the firm will only export if its productivity draw is above a cutoff ${\varphi }_X^{*}$ which is pinned down by the zero profit condition for exporting:

Intuitively, the exporting cutoff is lower with lower trade costs and larger export markets. The ex-ante expected profit is therefore:

where g(ϕ) is the density function of the new (post-innovation) productivity distribution with shape parameter κ. Applying the properties and parameters of the Pareto distribution, defining ξ = σ – 1, we can solve for expected profit conditional on innovating to be:

and the return from innovation is therefore:

where we have $\frac{\partial E\left(\pi \right)}{\partial \tau }<0$ and $\frac{\partial E\left(\pi \right)}{\partial M^x}>0$. With lower trade costs and/or larger export markets, the profit for each exporter with given productivity is higher, at the same time as the probability of becoming an exporter is higher, both contributing to higher ex-ante expected profits. At the same time, higher initial productivity ϕ₀ raises expected profit conditional on innovating, as higher ϕ₀, being the lower bound for the new productivity draw, leads on average to a higher level of post-innovation productivity and hence results in higher expected profit in domestic as well as export markets.

Now moving to t = 1, conditional on being hit by a liquidity shock ρ, it immediately follows that the firm should continue whenever ρ is not too high so as to maintain a positive net payoff from the innovation. Denoting this threshold by ρ¹, we can derive its value from the first order condition of:

which yields ρ¹ = E(R(ϕ₀)), that is, the firm maximizes the net present value (NPV) of the innovation project by covering the liquidity shock/cost overrun as long as it does not exceed the expected profit from continuing the innovation. This first-best threshold ρ¹ then is the cash amount the firm needs to hold in t = 0.

In t = 0, if the net present value of the innovation project is positive (which, for given I, effectively requires a minimum productivity level for innovating firms), the firm will want to innovate and pay the upfront amount I plus hold the cash amount ρ¹ to insure against the liquidity shock in t = 1.

We establish the following main results.

Result 1 Only firms above a minimum productivity cutoff will innovate. Innovating firms hold more cash than non-innovating firms. Innovating firms with higher initial productivity hold more cash.

From equation (6) above, we can express the maximized NPV for a firm with given initial productivity ϕ₀ as

Only firms with positive expected MNPV from innovating will do so. In other words, only firms with initial productivity above the cutoff level ${\varphi }_0^1$ will choose to innovate, where the cutoff level is obtained from the break-even condition $MNPV\left({\varphi }_0^1\right)=0$. These firms will pay the upfront cost I and hoard the cash amount ρ¹ to cover subsequent liquidity risk. All else equal, therefore, their cash holdings will be higher by the amount ρ¹ compared to non-innovating firms. Moreover, from equation (5), we know:

so that the innovation-accompanying amount of cash holdings increases with the ex-ante productivity of the firm. This crucial result distinguishes our motive of cash holdings from others in the literature where demand for cash is a result of intangible capital shrinking the firm’s collateral base and external financing capacity (see e.g. Falato et al. 2013). Unlike the collateral channel that generates a stronger motive of cash holdings for smaller firms facing more financial frictions, the innovation channel in our model, encompassing investment in intangible capital, leads to a higher level of cash holdings for larger, more productive firms. This is because more productive firms expect more payoff from productivity enhancing innovation and hence are willing to absorb larger liquidity shocks in the interim to keep their “skin in the game”.

Result 2 Conditional on being an innovating firm, globalization in terms of lower trade costs τ and/or expanded foreign market size M^X increases the level of the firm’s cash holdings, that is $\frac{\partial {\rho }^1}{\partial \tau }<0,\frac{\partial {\rho }^1}{\partial M^X}>0$. Moreover, more productive firms increase their cash holdings more in response to the same shock than less productive ones.

This result follows from taking the derivative of ρ¹ with respect to M^x, that is:

where Ω(.) > 0 is a composite function of underlying parameters of the model taking on positive values (and similarly for a reduction in τ). A positive export shock increases the returns from innovation and hence makes firms willing to absorb larger liquidity shocks, therefore raising the optimal cash holdings of any firm above the innovation productivity cutoff. In addition, as κ > 0, it follows that this increase in cash holdings from a positive export shock is increasing in the underlying productivity level ϕ₀. Intuitively, a more productive firm is more able to translate the better export opportunity into higher profits, and thus will hold more cash to withstand larger liquidity shocks so as to reap these profits eventually.

Result 3 Globalization in terms of lower trade costs τ and/or expanded foreign market size M^X reduces the productivity cutoff for innovation and thus, for a given distribution of initial productivity across firms, increases innovation activity and the average cash holdings among incumbent firms in the industry.

As derived in result 1, the minimum productivity cutoff for innovating firms is given by the break-even condition $MNPV\left({\varphi }_0^1\right)=0$. At the same time, we know that a positive export shock raises the MNPV for all values of ϕ₀, as per envelope theorem, we have $\frac{\partial MNPV}{\partial M^X}=F\left({\rho }^1\right)\frac{\partial E\left(R\right)}{\partial M^X}>0$. It immediately follows that for given innovation costs I, a positive globalization shock lowers the minimum productivity cutoff for firms to innovate. In other words, a larger M_X and/or lower τ shift down the inverse function MNPV^–1(·) and reduce ${\varphi }_0^1=MNP{V}^{-1}\left(0\right)$ to the lower value ${\varphi }_0^2$.

All three results regarding the schedule of optimal cash holdings as a function of the initial productivity, the set of innovating firms, and the shifted curve triggered by more trade openness (through lower trade costs and/or larger markets) are depicted in Figure 9. As derived above, the schedule of optimal cash holdings is a convex positive function of the underlying productivity level above the minimum productivity cutoff ${\varphi }_0^1$. A trade liberalization that lowers trade cost τ or expands access to foreign markets MX shifts the schedule inward and for any given level of ϕ₀, steepens the slope of the cash holdings schedule. As a result, the increase in cash holdings is disproportionately larger for more productive firms, and zero for firms below the cutoff. Moreover, for any given productivity distribution, export liberalization also expands the set of innovating firms by reducing the cutoff.⁷

View Full Size

Set of innovating firms and schedule of cash holdings as a function of initial productivity

Citation: IMF Working Papers 2019, 017; 10.5089/9781484392164.001.A001

Note: Shift after trade liberalization is sketched in blue. Productivity cutoff for innovators are indexed by 1 and 2 before and after shift.

• Download Figure
• Download figure as PowerPoint slide

3.3 Testable implications

The main testable predictions of the model can be summarized as the following:

• For the within-firm variation in cash holdings, Result 1 and Result 2 predict that a firm’s cash holdings should increase with export opportunities only if the firm has high enough underlying productivity. Using different proxies for a firm’s productivity, we should see that the marginal impact of trade openness on cash holdings is stronger for more productive firms. Moreover, the differential of this marginal impact should be most pronounced when comparing firms with the highest and lowest productivity levels, while comparison with firms of intermediate productivity is less unambiguous in the extended version of the model (see Figure A.3). Importantly, this differential prediction that export opportunities affect firms with higher productivity more also allows us to discriminate between our channel and those relying on innovation interacting with financial constraints such as Falato et al. (2013), where innovation raises cash holdings more in smaller/less productive firms, as they face more external financing constraint.⁸

• Result 2 predicts that, if the firm has high enough productivity, the increase in cash holdings facilitates more innovation activity by making firms more able to absorb liquidity shocks in the interim. Measured spending on innovation (equal $I+{\displaystyle {\int }_0^{{\rho }^1}\rho f\left(\rho \right)d\rho }$ in the model) will therefore also increase with expanding export opportunities.⁹ Similar to the result for cash, this positive effect should be stronger for larger (or more productive/more profitable/export-oriented) firms.

• At the industry-level within a country, Result 3 implies that cash holdings increases on average in an industry when its export opportunities increase. In line with the model, we should also see the average industry-wide R&D spending, a measure of innovation intensity at the industry-level, increase at the same time. However, the average industry-level outcome may be small and hard to be identified due to the heterogeneity in the impact of globalization across firms with varying level of productivity. Moreover, the industry result relies on the prediction that improved market access also allows new firms to become innovators (as the asset threshold falls). However, this extensive margin of innovation is not possible to measure in the R&D data, as missing entries for R&D spending can reflect either a non-reporting of actual spending or zero spending on R&D.¹⁰ We therefore focus on the intensive margin of innovation and associated cash holdings (Results 1 and 2) in the empirical analysis below.

4.1 Data

We use Thomson Reuter Worldscope data spanning the period 1995-2014 for five major industrial countries where coverage of publicly listed firms is among the most comprehensive: the U.S., UK, Germany, Japan, and France. Following the literature, excluded are firms with: negative equity, negative sales and missing value for total assets, as well as firms in the utilities and financial sector (sic 6000-6799, sic 4900-4999, sic 9000-9999, sic 1800-1999).

Table 1 summarizes the median of some key variables for each country in the sample. Overall, firms hold about 10 percent of total assets in cash and short-term investment (cashlike instruments), with Japanese firms having the highest cash ratio of 15 percent for the median firm and being on average the largest in terms of asset size. Among firms that have positive R&D spending, US firms are the most active in terms of investing in innovation, spending on average over twice as much on R&D as a share of their revenues than firms in other countries, while they are also the least likely to pay out dividends.¹¹

Table 1:

Average of yearly medians for each country.

Source: Thomson Reuters Worldscope. CashTA is the ratio of Cash and short-term investment over total book assets; MB is the market-to-book ratio; Size is expressed in Mn US Dollar; CF stands for cash flow in percent of asset, RD2Sales is the median spending on R&D in percent of sales among firms with positive R&D spending (the unconditional median is zero), intangTA is the share of declared intangibles to assets ratio; Dividend is the share of firms that pay dividend; obs is the total number of firm-year observations.

Table 1:

Average of yearly medians for each country.

country	CashTA	Size	MB	Leverage	CF	RD2Sales	intangTA	Dividend	obs
FRANCE	.108	103	1.238	.178	.072	.033	.112	.538	13121
GERMANY	.094	125	1.263	.152	.076	.032	.06	.469	12602
JAPAN	.15	273	1.01	.205	.052	.012	.006	.81	65323
UK	.089	80	1.38	.131	.073	.03	.118	.559	24598
USA	.115	202	1.547	.151	.075	.078	.108	.228	83684

Source: Thomson Reuters Worldscope. CashTA is the ratio of Cash and short-term investment over total book assets; MB is the market-to-book ratio; Size is expressed in Mn US Dollar; CF stands for cash flow in percent of asset, RD2Sales is the median spending on R&D in percent of sales among firms with positive R&D spending (the unconditional median is zero), intangTA is the share of declared intangibles to assets ratio; Dividend is the share of firms that pay dividend; obs is the total number of firm-year observations.

View Table

Table 1:

Average of yearly medians for each country.

country	CashTA	Size	MB	Leverage	CF	RD2Sales	intangTA	Dividend	obs
FRANCE	.108	103	1.238	.178	.072	.033	.112	.538	13121
GERMANY	.094	125	1.263	.152	.076	.032	.06	.469	12602
JAPAN	.15	273	1.01	.205	.052	.012	.006	.81	65323
UK	.089	80	1.38	.131	.073	.03	.118	.559	24598
USA	.115	202	1.547	.151	.075	.078	.108	.228	83684

Source: Thomson Reuters Worldscope. CashTA is the ratio of Cash and short-term investment over total book assets; MB is the market-to-book ratio; Size is expressed in Mn US Dollar; CF stands for cash flow in percent of asset, RD2Sales is the median spending on R&D in percent of sales among firms with positive R&D spending (the unconditional median is zero), intangTA is the share of declared intangibles to assets ratio; Dividend is the share of firms that pay dividend; obs is the total number of firm-year observations.

Although the dataset contains consolidated accounts of mostly publicly listed firms, evidence in Dao and Maggi (2018) confirms that the combined cash holdings and net lending rates of these large firms, when added up, track well the aggregate evolution of corresponding variables from official flow of funds and sectoral national accounts data. This finding on aggregate representativeness is consistent with other studies focusing exclusively on the U.S. corporate sector, which also established that the corporate net lending as well as cash holdings is extremely concentrated among large firms (see Armenter, 2012 and references therein). Understanding drivers of cash holdings and saving by listed firms can therefore shed light on forces that drive the evolution of overall private saving (to which the corporate sector contributes substantially), real interest rates and the current account.

Worldscope provides two main four-digit-level standard industry classifications (SIC) for each firm, accompanied by the amount of sales of each category. Using an initial year’s product segment sales shares, we can construct each firm’s exposure to export opportunities and import competition by combining this firm-level data with industry-level trade and tariff data. The bilateral trade data, sector output data, and MFN tariff data come from the UN Comtrade database, World Input Output Database (WIOD), and TRAINS database, respectively, all of which are then matched to the SIC code of each firm in an initial year to derive sector- or firm-specific weighted average exposure to export and import shocks (more details below).

Figure 10 illustrates a rapidly growing role of China in trade with advanced economies in our sample. It also highlights substantial variation across countries and industries, providing a source of identification for our econometric analysis below. Similarly, Figure 11 describes the extent of tariff liberalization in our sample economies and their trading partners. It is worth noting that import tariff liberalization has been stagnant over the last two decades in our sample countries, mainly because their tariff rates had already reached low levels by the mid 1990’s. In contrast, their trading partners, many of them emerging market economies, whose tariff rates were relatively high, have undergone substantial tariff liberalization over the same period. In this regard, as far as tariff rates are concerned, firms in our sample appear to have experienced not so much import shocks as positive export shocks. This fact underscores the importance to look at both the export and import margins when analyzing the effect of trade.

View Full Size

The evolving role of China in global trade.

Citation: IMF Working Papers 2019, 017; 10.5089/9781484392164.001.A001

Source: World Input Output Database (WIOD) and authors’ calculations; Country-2digit sector level distribution.

• Download Figure
• Download figure as PowerPoint slide

View Full Size

Global tariff liberalization.

Citation: IMF Working Papers 2019, 017; 10.5089/9781484392164.001.A001

Source: UN Comtrade, TRAINS, and authors’ calculations; Country-2digit sector level distribution.

• Download Figure
• Download figure as PowerPoint slide

Our empirical strategy essentially relies on a difference-in-difference identification of the heterogeneous effect of export shocks across firms with differential initial productivity level ϕ₀. In a multi-factor production environment, the variable ϕ should correspond to firm-level total factor productivity (TFP). Crucial elements for computing firm-level TFP are however not available in our dataset (such as intermediate inputs, prices etc). Thus, we will have to rely on simple proxies following the literature such as size, labor productivity, export intensity, or profitability.¹²

Another important aspect of our model is the firm’s decision to engage in long-term investment in innovation. We understand innovation here as any activity that requires long-term financial commitment with uncertain outcomes, and exposes the firm to liquidity risk in the interim. R&D is one category of such innovation activities, and evidence exists that firms indeed use cash holdings to smooth R&D expenditure, which is costly to adjust, in response to financing shocks (Brown and Petersen, 2011). But similar risk applies to other activity of the firm that increases its intangible capital stock, in particular investment in its human, organizational and social capital (see Lev and Radhakrishnan, 2005; Eisfeldt and Papanikolaou, 2013). The reason is because spending on such activities as internal training, brand development, development of distribution networks etc. require steady and long-term outlays, increasing the share of quasi-fixed costs and decoupling the firm’s revenues from its operating expenditure, hence increasing its exposure to liquidity shortage (see Srivastava and Tse, 2016). However, we do not have data on such activities by firms. The closest measure encompassing such intangible capital enhancing spending are spending on so-called SG&A (selling, general and administration costs), which however also contain non-intangible enhancing expenses such as social security taxes, pension costs, and other overhead costs. We therefore consider both R&D and SG&A spending as our measure of innovation activity when investigating the underlying channel through which globalization leads to an increase in cash holdings.

4.2 Empirical Strategy

The main testable hypothesis from the model in this paper is that rising globalization, particularly in terms of expanded export opportunities, would lead to an increase in cash holdings, and particularly more so for innovating firm with higher productivity. Moreover, the model suggests that it occurs through boosting incentive for innovation, resulting in an increase in innovation spending, more so in firms with higher initial productivity.

Accordingly, at the firm-level, we use the following baseline specification to test the main hypotheses:

where Y_ijct is the dependent variable of interest, primarily the ratio of cash holdings to total assets in log for firm i in two-digit-level sector j and country c at year t. $SHOC{K}_{jct}^{exp}$ captures the country-sector-year-level potential export opportunities that could stem from, for instance, improved foreign market access or increased demand abroad. Likewise, $SHOC{K}_{jct}^{imp}$ denotes the degree of country-sector-year-level import competition that domestic firms face. As such, export and import shocks are separately estimated. Z_ijct is a set of other relevant control variables such as a firm’s total sales volume, operating cash flow, etc. The baseline regression includes firm fixed effects to explore within-firm variation over time in response to changing degrees of globalization. In addition, country-year fixed effects would absorb any other macroeconomic factors that could affect a firm’s cash holding decision.

As spelled out above, the model predicts that the effect of globalization should be more pronounced for firms with higher initial productivity, as they are more likely to undertake innovation-related investment (due to selection into innovation) and more able to translate the better export opportunity into higher profits (due to the post-innovation productivity distribution). This prediction can be tested by including additional interaction terms to capture heterogeneous responses along different levels of productivity proxied by, for instance, size (measured by total assets), productivity (measured by labor productivity), export intensity (measured by the share of foreign sales), or profitability (measured by net income per employee). Specifically, we assign each firm a tercile dummy variable encoding its relative position in the distribution of the respective productivity proxy in a given country and year. The corresponding specification can then be expressed as:

where I_ijct,l is a tercile dummy variable, whose stand-alone level is also included in Z_ijct.

A typical empirical challenge in identifying the causal effect of globalization on firm-level decisions is likely to prevail in this setting, not least because our sample is composed of publicly listed firms, some of which are large enough to influence potential globalization measures. In an effort to alleviate such endogeneity concerns, among several possible candidate variables for $SHOC{K}_{jct}^{exp,imp}$, we employ measures of exports to, and imports from, China—both scaled by the two-digit country-sector-level output—as a measure for export opportunities and import competition, respectively:

and

To the extent that much of the recent rise in trade with China is driven by supply-side and demand-side shocks from China—productivity shocks for China’s exports and unilateral trade liberalization for China’s imports—, these measures are expected to embody exogenous shocks from the perspective of a firm in any given G7 partner country. Still, however, we acknowledge that such measures are not entirely immune to potential en-dogeneity biases. We thus follow Autor et al. (2013) and Autor et al. (2016) to further instrument them with the corresponding average values of other advanced economies. This instrumental variable approach allows us to extract China-made exogenous shocks in each industry, which in turn underlie the rising role of China in global trade.¹³ Such China-specific shocks common to all third partner countries should therefore be strongly correlated with the change in export opportunities and import competition in a given partner country, but would not be directly related to a firm’s cash holdings and innovation spending once their overall performance and other macroeconomic specific shocks are controlled for.

Past evidence, as discussed in detail in Autor et al. (2013), tends to lend support for the validity of our identification strategy in that demand and/or technology shocks common to major advanced economies played only a minor role in explaining the recent surge in China’s trade. Nevertheless, caution is warranted in interpreting the estimated coefficients below, in case our instrumental variables are still contaminated by any remaining correlated demand and supply shocks across countries, such as sector-year-level common trend in technological growth.¹⁴ As robustness checks, we thus also consider alternative firm-level trade shock measures, thereby ensuring sufficient variation across firms even with additional sector-year fixed effects to absorb any such concerns.¹⁵

The first alternative measure is constructed as the weighted average of four-digit country-sector-level shocks:

for exports, and

for imports, respectively, where the share of exports to (imports from) China in total exports (imports) is used as a four-digit country-sector-level export (import) shock measure and the weights are calculated using the initial year’s sales share of the primary and secondary four-digit SIC codes, j′, for each firm. That is, the country-sector level export/import exposure to China is converted to a firm-level one using weights ω_ij′c, which corresponds to the respective share of a given firm’s sales in each four-digit sector in the initial year. As above, we further consider instrumental variables for 2SLS estimation by taking the corresponding average values from other advanced economies while keeping the firm-level weight.

The second alternative measure is constructed by replacing China-based trade shock measures with tariff-based measures:

for k = {exp, imp}, where the weight ω_ij′c is defined as above, while ${\tau }_{c{j}^{\prime }}^{imp}$ denotes country c′s most favored nation (MFN) tariff rate imposed on imported goods from industry j′, and ${\tau }_{c{j}^{\prime }}^{exp}$ is a weighted average of MFN tariff rates that a country c′s exporters in industry j′ would face in their destination countries.¹⁶

4.3 Estimation Results

Baseline estimation results Table 2 and Table 3 present firm-level estimation results from our baseline specification using OLS and 2SLS, respectively, where the dependent variable is cash-to-asset ratio in log (multiplied by 100 for ease of interpretation). Column 1 reports estimation results from the baseline specification without any interaction terms, while columns 2, 3, and 4 summarize estimation results from the augmented specification that includes interaction terms with tercile dummy variables based on total asset size.

Table 2:

Globalization and Cash Holding: China Shocks; OLS

Notes: This table presents results of panel OLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 2:

Globalization and Cash Holding: China Shocks; OLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	4.386***	4.228***	-2.851***	-2.386***
	(0.581)	(0.536)	(0.785)	(0.756)
expSHOCK_X2		-0.186	1.393***	0.918*
		(0.498)	(0.519)	(0.516)
expSHOCK_X3		0.419	4.389***	3.548***
		(0.719)	(0.857)	(0.814)
impSHOCK				-0.146**
				(0.0642)
impSHOCK_X2				0.210***
				(0.0789)
impSHOCK_X3				0.407***
				(0.0918)
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
R2	0.720	0.720	0.732	0.732
N	57172	57172	57172	57172

Notes: This table presents results of panel OLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 2:

Globalization and Cash Holding: China Shocks; OLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	4.386***	4.228***	-2.851***	-2.386***
	(0.581)	(0.536)	(0.785)	(0.756)
expSHOCK_X2		-0.186	1.393***	0.918*
		(0.498)	(0.519)	(0.516)
expSHOCK_X3		0.419	4.389***	3.548***
		(0.719)	(0.857)	(0.814)
impSHOCK				-0.146**
				(0.0642)
impSHOCK_X2				0.210***
				(0.0789)
impSHOCK_X3				0.407***
				(0.0918)
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
R2	0.720	0.720	0.732	0.732
N	57172	57172	57172	57172

Notes: This table presents results of panel OLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 3:

Globalization and Cash Holding: China Shocks; 2SLS

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 3:

Globalization and Cash Holding: China Shocks; 2SLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	8.209***	7.229***	-5.435***	-5.187***
	(0.758)	(0.987)	(1.224)	(1.646)
expSHOCK_X2		-0.264	3.131**	2.769
		(1.280)	(1.342)	(1.967)
expSHOCK_X3		2.752*	8.177***	8.042***
		(1.483)	(1.456)	(2.098)
impSHOCK				-0.117
				(0.193)
impSHOCK_X2				0.101
				(0.244)
impSHOCK_X3				0.0212
				(0.268)
First stage	Dep. var: expSHOCK
expSHOCK (IV)	1.248*** [0.058]
F-stat.	40.9
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
N	57172	57172	57172	57172

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 3:

Globalization and Cash Holding: China Shocks; 2SLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	8.209***	7.229***	-5.435***	-5.187***
	(0.758)	(0.987)	(1.224)	(1.646)
expSHOCK_X2		-0.264	3.131**	2.769
		(1.280)	(1.342)	(1.967)
expSHOCK_X3		2.752*	8.177***	8.042***
		(1.483)	(1.456)	(2.098)
impSHOCK				-0.117
				(0.193)
impSHOCK_X2				0.101
				(0.244)
impSHOCK_X3				0.0212
				(0.268)
First stage	Dep. var: expSHOCK
expSHOCK (IV)	1.248*** [0.058]
F-stat.	40.9
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
N	57172	57172	57172	57172

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

A first look at the OLS estimation results in Table 2 suggests a significant within-firm increase in cash holdings, on average, in response to expanded export opportunities (column 1). As we add interaction terms with tercile dummy variables, no differential effects across firms are found (column 2). However, as country-year fixed effects are further added to control for macroeconomic shocks, the effect turns out to be concentrated among bigger firms (column 3). The results continue to hold after taking into account the effect of import competition. Interestingly, the export channel tends to dominate the import channel in that the point estimate on the former is about 10 times bigger than that of the latter (column 4).

Turning to Table 3 for 2SLS estimation results whereby the country-sector-level trade shock measures are instrumented by the average values in the other advanced economies, qualitatively identical results to those from OLS estimation are found. The first stage regression result in the bottom panel supports the strength of the instrumental variable—the substantial predictive power of instrumental variables from the peer country group’s trade with China for a given country’s export shock is confirmed. The second stage regression results in the upper panel further reveals that the coefficient estimates on the export shock variable from 2SLS estimator tend to be significantly greater in absolute terms than those from OLS estimator, suggesting that our instrumental variable strategy could partly correct for attenuation bias—due to measurement errors particularly in export shock variables—as well as omitted variable bias that would push the relationship negatively—reflecting for instance forces that discourage investment of domestic firms for precautionary reasons and hence, reduce their exports to China but boost their cash holdings.

The size of the estimated coefficient in column (4) implies that, conditional on all other macroeconomic and firm-level factors affecting cash holdings proportionately across firms, a 1 percentage point increase in export opportunities to China raises cash-to-asset ratio by around 8 percent more for largest firms compared to average firms in that country and year, whereas smallest firms actually reduce cash-to-asset ratio by around 5 percent relative to average firms in a given country-year. A simple back-of-the-envelope calculation suggests that globalization-led growth in export opportunities explains around 25 percent of the differential growth in cash holdings between bottom and top tercile firms over the period 2000-2011 for US firms, with the magnitude ranging from 10 percent for French firms to 86 percent for German firms. On the other hand, the import channel yields much smaller and nosier estimates, possibly reflecting offsetting forces of increased foreign competition on innovation activity such as the ‘Schumpeterian’ force versus the ‘escape competition’ force proposed in Aghion, Bloom, Blundell, Griffith and Howitt (2005).

Figure 12 illustrates this differential effect stemming from the export channel (as reported in column 2 in Table 3), by plotting the reduced-form correlation between cash ratio and the China export shock for large and small firms, conditional on other explanatory variables in the baseline regression. While the positive correlation is also found for firms in the bottom tercile of the asset size distribution, the slope is steepest for the top tercile firms. To the extent that firm size is a valid proxy for firm-level productivity, such heterogeneity in the degree of the export channel on cash holdings is consistent with the model prediction.¹⁷ Since the regressions control for firm fixed effects as well as sales and operating cash flows, we stress that the impact on cash holdings is not mechanically driven by increased profitability.

View Full Size

cash holdings increases with higher export demand, more so for larger firms.

Citation: IMF Working Papers 2019, 017; 10.5089/9781484392164.001.A001

Notes: higher export demand is associated with higher cash holdings, but only significantly so for large firms, i.e. those within the top tercile of asset size distribution of a given country-year. The dots represent the average within each of the 25 quantiles of the China shock (measured as change in export share to China from third countries by industry, weighted by the firm’s sales share in its 2 main industries), absorbing the firm fixed effects, plotted against the corresponding quantile-average of log cash holdings, conditional on the firm fixed effect. The conditional correlation is plotted separately for firms in the top and bottom tercile of the country-specific asset size distribution.

• Download Figure
• Download figure as PowerPoint slide

Alternative trade shock measures One potential concern about our baseline estimation strategy that exploits sector-level trade shocks stemming from China is that trade shock measures might be contaminated by other types of sector-year-level variation such as common technological growth trends. To check robustness of our baseline results to such concerns, we first tweak the underlying original trade shock measure (that was scaled by total output) to one scaled by total exports in each country-sector-year. This basically allows us to construct trade shock measures at the four-digit sector level, which in turn enables us to construct firm-level trade shocks by taking a weighted average across the two main four-digit sectors for each firm, with weights being the firm’s sales share in its two main product segments. This alternative trade shock measures the role of China in each country-sector’s overall trade, rather than overall production and absorption as in the baseline. With these more granular trade shock measures now varying across firms in a given country-sector, additional sector-year fixed effects can be added to control for common sector-year shocks across countries, effectively isolating trade shocks from other sources of sector-level shocks. Table 4 confirms that results are qualitatively identical to the baseline estimation results.

Table 4:

Globalization and Cash Holding: Firm-level China Shocks; 2SLS

Notes: This table presents results of panel 2SLS regressions examining the effect of firm-level export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 4:

Globalization and Cash Holding: Firm-level China Shocks; 2SLS

	(1	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	5.001***	4.467***	-0.729	-0.815
	(0.853)	(1.019)	(1.213)	(1.376)
expSHOCK_X2		-0.119	1.418	1.424
		(0.929)	(0.895)	(0.930)
expSHOCK_X3		1.259	4.221***	3.856***
		(1.035)	(1.052)	(1.054)
impSHOCK				-0.335
				(0.247)
impSHOCK_X2				-0.0444
				(0.193)
impSHOCK_X3				0.302
				(0.284)
First stage	Dep. var: expSHOCK
expSHOCK(IV)	0.861***[0.131]
F-stat.	43.4
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
Sector-year FE	N	N	Y	Y
N	63600	63600	63578	63567

Notes: This table presents results of panel 2SLS regressions examining the effect of firm-level export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 4:

Globalization and Cash Holding: Firm-level China Shocks; 2SLS

	(1	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expSHOCK	5.001***	4.467***	-0.729	-0.815
	(0.853)	(1.019)	(1.213)	(1.376)
expSHOCK_X2		-0.119	1.418	1.424
		(0.929)	(0.895)	(0.930)
expSHOCK_X3		1.259	4.221***	3.856***
		(1.035)	(1.052)	(1.054)
impSHOCK				-0.335
				(0.247)
impSHOCK_X2				-0.0444
				(0.193)
impSHOCK_X3				0.302
				(0.284)
First stage	Dep. var: expSHOCK
expSHOCK(IV)	0.861***[0.131]
F-stat.	43.4
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
Sector-year FE	N	N	Y	Y
N	63600	63600	63578	63567

Notes: This table presents results of panel 2SLS regressions examining the effect of firm-level export and import shocks vis-à-vis China on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are inlcuded in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

Alternatively, we also compute trade shock measures based on tariff rates at the country-industry level. In particular, we compute a firm-specific measure of export tariff by using the lagged MFN tariff rate for each product (averaged across countries using trade weights at the beginning of the sample period), which is then weighted across products using the firm’s sales share in its two main product segments defined at the four-digit level. A change in MFN tariffs in trading partner countries is arguably exogenous to an individual firm incorporated in the exporting country and hence presents a useful robustness check of our main results. To control for the effect of increased import competition resulting from domestic tariff liberalization, we also compute a corresponding measure using the import MFN tariff in each product category, weighted in a similar manner.

Estimation results for cash holdings using these alternative MFN tariff-based trade shocks are summarized in Table 5. Consistent with the baseline results using the China trade shock, we find that a decline in export tariffs faced by firms is associated with higher cash holdings for firms in the top tercile of the asset size in each country, but lower cash holdings for firms in the bottom tercile, both relative to average firms in a given country-year. Once again, the impact is stronger through the export expansion rather than the import competition channel.

Table 5:

Globalization and Cash Holding: Tariff; OLS

Notes: This table presents results of panel OLS regressions examining the effect of export and import tariff changes on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are included in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 5:

Globalization and Cash Holding: Tariff; OLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expMFN	-3.106***	-1.753***	1.330***	1.008**
	(0.602)	(0.541)	(0.467)	(0.476)
expMFN_X2		-1.393**	-1.320***	-0.837
		(0.531)	(0.476)	(0.622)
expMFN_X3		-2.058***	-2.063***	-1.655**
		(0.761)	(0.610)	(0.675)
impMFN				0.636
				(0.820)
impMFN_X2				-1.788*
				(0.965)
impMFN_X3				-1.735*
				(0.911)
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
Sector-year FE	N	N	Y	Y
R2	0.713	0.713	0.731	0.730
N	45873	45873	45854	44162

Notes: This table presents results of panel OLS regressions examining the effect of export and import tariff changes on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are included in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 5:

Globalization and Cash Holding: Tariff; OLS

	(1)	(2)	(3)	(4)
	No interaction	X=SIZE	X=SIZE	X=SIZE
expMFN	-3.106***	-1.753***	1.330***	1.008**
	(0.602)	(0.541)	(0.467)	(0.476)
expMFN_X2		-1.393**	-1.320***	-0.837
		(0.531)	(0.476)	(0.622)
expMFN_X3		-2.058***	-2.063***	-1.655**
		(0.761)	(0.610)	(0.675)
impMFN				0.636
				(0.820)
impMFN_X2				-1.788*
				(0.965)
impMFN_X3				-1.735*
				(0.911)
Firm FE	Y	Y	Y	Y
Country-year FE	N	N	Y	Y
Sector-year FE	N	N	Y	Y
R2	0.713	0.713	0.731	0.730
N	45873	45873	45854	44162

Notes: This table presents results of panel OLS regressions examining the effect of export and import tariff changes on cash holding. Dependent variable is cash holding-to-assets ratio in log. All columns include firm fixed effects as well as other firm-level controls such as total sales, operating cash flow. Additional country-year and sector-year fixed effects are included in columns 3-4. Tercile dummy variables based on total assets are also included but not reported in columns 2-4. Standard errors in parentheses are clustered at two levels (country-year and sector-year). (* p < 0.10, ** p < 0.05, *** p < 0.01)

Confirming the innovation channel Next, we turn to testing the underlying mechanism of the model: firms raise cash holdings because of an increased incentive for R&D and intangible investment in response to globalization shocks. This implies that the effect of globalization on cash holdings should be observed only among firms that do invest in innovation activity, which can be tested by checking the baseline estimation results for innovative and non-innovative firms separately. We define innovative firms as those who have ever spent on R&D, and non-innovative firms as those who have never spent on R&D over the sample period. Column 1 in Table 6 reports the estimation results for innovative firms which are somewhat stronger than the baseline estimation results reported in column 4 in Table 3. On the other hand, column 2 in Table 6 shows that there is no significant effects from export or import shocks across all types of firms, which can be interpreted as a placebo test.¹⁸

Table 6:

The Underlying Channel of Globalization and Cash Holding: China Shocks; 2SLS

Notes: This table presents results of panel 2SLS regressions confirming the innovation channel through which export and import shocks vis-à-vis China affect cash holding. Dependent variable in columns 1 and 2 is cash holding-to-assets ratio in log, whille depedent variable in columns 3 and 4 is R&D spending-to-sales ratio and SG&A spending-to-sales ratio, respectively, both in log. Column 1 includes innovative firms only and column 2 includes non-innovative firms only. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables based on total assets are also included but not reported. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 6:

The Underlying Channel of Globalization and Cash Holding: China Shocks; 2SLS

	(1)	(2)	(3)	(4)
	X=SIZE	X=SIZE	X=SIZE	X=SIZE
	innovative firms	non-innovative firms	DV:R&D spending	DV:SG&A spending
expSHOCK	-6.977***	3.537	-11.06***	-7.220***
	(1.760)	(5.963)	(2.960)	(1.489)
expSHOCK_X2	2.957	8.067	11.74***	8.940***
	(2.376)	(7.624)	(3.519)	(1.642)
expSHOCK_X3	8.747***	17.45	19.35***	11.56***
	(2.164)	(10.80)	(3.710)	(1.839)
impSHOCK	-0.0276	-0.0784	1.409**	0.494**
	(0.248)	(0.349)	(0.561)	(0.194)
impSHOCK_X2	-0.0926	0.0996	-0.843	-0.434*
	(0.400)	(0.428)	(0.715)	(0.223)
impSHOCK_X3	-0.230	0.0977	-2.056***	-0.630**
	(0.353)	(0.505)	(0.751)	(0.259)
Firm FE	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y
N	43782	13390	36782	51474

Notes: This table presents results of panel 2SLS regressions confirming the innovation channel through which export and import shocks vis-à-vis China affect cash holding. Dependent variable in columns 1 and 2 is cash holding-to-assets ratio in log, whille depedent variable in columns 3 and 4 is R&D spending-to-sales ratio and SG&A spending-to-sales ratio, respectively, both in log. Column 1 includes innovative firms only and column 2 includes non-innovative firms only. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables based on total assets are also included but not reported. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 6:

The Underlying Channel of Globalization and Cash Holding: China Shocks; 2SLS

	(1)	(2)	(3)	(4)
	X=SIZE	X=SIZE	X=SIZE	X=SIZE
	innovative firms	non-innovative firms	DV:R&D spending	DV:SG&A spending
expSHOCK	-6.977***	3.537	-11.06***	-7.220***
	(1.760)	(5.963)	(2.960)	(1.489)
expSHOCK_X2	2.957	8.067	11.74***	8.940***
	(2.376)	(7.624)	(3.519)	(1.642)
expSHOCK_X3	8.747***	17.45	19.35***	11.56***
	(2.164)	(10.80)	(3.710)	(1.839)
impSHOCK	-0.0276	-0.0784	1.409**	0.494**
	(0.248)	(0.349)	(0.561)	(0.194)
impSHOCK_X2	-0.0926	0.0996	-0.843	-0.434*
	(0.400)	(0.428)	(0.715)	(0.223)
impSHOCK_X3	-0.230	0.0977	-2.056***	-0.630**
	(0.353)	(0.505)	(0.751)	(0.259)
Firm FE	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y
N	43782	13390	36782	51474

Notes: This table presents results of panel 2SLS regressions confirming the innovation channel through which export and import shocks vis-à-vis China affect cash holding. Dependent variable in columns 1 and 2 is cash holding-to-assets ratio in log, whille depedent variable in columns 3 and 4 is R&D spending-to-sales ratio and SG&A spending-to-sales ratio, respectively, both in log. Column 1 includes innovative firms only and column 2 includes non-innovative firms only. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables based on total assets are also included but not reported. Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 7:

Globalization and Cash Holding: China Shocks; 2SLS; Alternative firm proxies

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding for innovative firms. Dependent variable is cash holding-to-assets ratio in log. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on labor productivity (columns 1-2), the share of foreign sales in total sales (columns 3-4), and net income per employee (columns 5-6). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 7:

Globalization and Cash Holding: China Shocks; 2SLS; Alternative firm proxies

	(1)	(2)	(3)	(4)	(5)	(6)
	X=LP	X=LP	X=EXP	X=EXP	X=NIPE	X=NIPE
expSHOCK	-3.513***	-4.436***	-10.89***	-13.34***	-7.726***	-5.763**
	(1.030)	(1.716)	(2.357)	(3.420)	(1.959)	(2.907)
expSHOCK_X2	2.410*	2.919	8.122***	9.867**	5.244	-2.783
	(1.421)	(2.251)	(2.894)	(4.044)	(3.982)	(6.850)
expSHOCK_X3	12.70***	15.39***	9.039***	12.40***	19.04***	20.54***
	(2.177)	(3.209)	(2.197)	(3.509)	(2.854)	(3.875)
impSHOCK		0.291		0.341		-0.383
		(0.380)		(0.331)		(0.517)
impSHOCK_X2		-0.237		-0.287		1.740
		(0.420)		(0.340)		(1.078)
impSHOCK_X3		-0.819*		-0.678		-0.289
		(0.496)		(0.418)		(0.560)
Firm FE	Y	Y	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y	Y	Y
N	54929	54929	40059	40059	54911	54911

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding for innovative firms. Dependent variable is cash holding-to-assets ratio in log. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on labor productivity (columns 1-2), the share of foreign sales in total sales (columns 3-4), and net income per employee (columns 5-6). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 7:

Globalization and Cash Holding: China Shocks; 2SLS; Alternative firm proxies

	(1)	(2)	(3)	(4)	(5)	(6)
	X=LP	X=LP	X=EXP	X=EXP	X=NIPE	X=NIPE
expSHOCK	-3.513***	-4.436***	-10.89***	-13.34***	-7.726***	-5.763**
	(1.030)	(1.716)	(2.357)	(3.420)	(1.959)	(2.907)
expSHOCK_X2	2.410*	2.919	8.122***	9.867**	5.244	-2.783
	(1.421)	(2.251)	(2.894)	(4.044)	(3.982)	(6.850)
expSHOCK_X3	12.70***	15.39***	9.039***	12.40***	19.04***	20.54***
	(2.177)	(3.209)	(2.197)	(3.509)	(2.854)	(3.875)
impSHOCK		0.291		0.341		-0.383
		(0.380)		(0.331)		(0.517)
impSHOCK_X2		-0.237		-0.287		1.740
		(0.420)		(0.340)		(1.078)
impSHOCK_X3		-0.819*		-0.678		-0.289
		(0.496)		(0.418)		(0.560)
Firm FE	Y	Y	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y	Y	Y
N	54929	54929	40059	40059	54911	54911

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash holding for innovative firms. Dependent variable is cash holding-to-assets ratio in log. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on labor productivity (columns 1-2), the share of foreign sales in total sales (columns 3-4), and net income per employee (columns 5-6). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Alternatively, noting that our model implies that an increase in cash holdings would be eventually translated to an increase in R&D spending, we can check the prediction by replacing the dependent variable in the baseline specification above with R&D spending-to-sales ratio in log (multiplied by 100 for ease of interpretation). Column 3 in Table 6 presents the estimation results from our baseline specification using 2SLS estimator, which confirm the model-implied mechanism: the estimated heterogeneity of the export effect on innovation spending across firms mirrors the model’s prediction for cash holdings. We find the presence of significant and sizable export channel effects particularly for bigger firms, whereas smallest firms actually reduce their R&D investment relative to average firms in a given country-sector in response to positive export shocks.

In fact, our model’s mechanism should apply to any activity by the firm that enhances its productivity in the long run but exposes it to higher liquidity risk in the short to medium run. Although R&D is important category of such of innovation spending, it has been argued that a range of other types of expenditure represent investment in intangible organizational, social, and human capital (see Eisfeldt and Papanikolaou, 2013). These are for example spending in employee training, marketing, advertising, branding, IT upgrading etc., which tend to enhance profits in the long run, but due to their nature of being quasi-fixed cost, in the short-run introduce more disconnect between cost and revenues and thus, heighten liquidity risk. Data from income statement contain such expenditure (in addition to R&D spending) in the variable SG&A (selling, general and administration), although these may also capture other kinds of non-innovation related spending such as contribution to employee’s social security funds. Column 4 in Table 6 reports the baseline regression results using the share of SG&A spending (in percent of sales) as the dependent variable (multiplied by 100 for ease of interpretation), confirming the main result of positive and stronger effect of export shocks on broader innovation spending for bigger firms.

Confirming the liquidity risk channel We have shown that larger/more productive firms tend to increase cash holdings and spending on innovation and other quasi-overhead outlays when experiencing a positive export shock, consistent with our model’s prediction. One final missing link is whether such firms indeed face increased liquidity risk subsequent to incurring such innovation spending to boost productivity, consistent with the model. We test this prediction in our empirical framework by estimating the differential effect of export shocks on the forward-looking volatility of cash flow from operations. That is, for each firm and year, we compute the standard deviation of cash flow from operation (scaled by assets) over the subsequent 4 years. The results are summarized in Table 8.

Table 8:

Cash flow volatility and globalization shocks

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash flow volatility, defined as the standard deviation of cash flow/asset rate over the subsequent 5-year window. Dependent variable is standard deviation of cash flow/asset ratio over the subsequent 5 years. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on firm’s total assets (size). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

Table 8:

Cash flow volatility and globalization shocks

	(1)	(2)	(3)	(4)
	All firms	SG&A< median	SG&A> median	SG&A> 75thpctile
exp_SHOCK	-3.696**	-0.401**	-7.409*	-31.26*
	(1.862)	(0.172)	(4.132)	(17.53)
exp_SHOCK_X2	3.506**	0.192	7.143*	29.29*
	(1.785)	(0.210)	(3.942)	(16.99)
exp_SHOCK_X3	3.440*	0.152	7.043*	30.72*
	(1.883)	(0.193)	(4.218)	(17.62)
imp_SHOCK_X1	0.477*	0.0361*	1.045	4.879*
	(0.282)	(0.0197)	(0.671)	(2.941)
imp_SHOCK_X2	-0.498*	0.00941	-1.091	-4.616
	(0.290)	(0.0327)	(0.694)	(2.894)
imp_SHOCK_X3	-0.498*	-0.0232	-1.065	-5.069*
	(0.297)	(0.0429)	(0.688)	(3.047)
Firm FE	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y
N	36430	16997	18901	10845

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash flow volatility, defined as the standard deviation of cash flow/asset rate over the subsequent 5-year window. Dependent variable is standard deviation of cash flow/asset ratio over the subsequent 5 years. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on firm’s total assets (size). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)

View Table

Table 8:

Cash flow volatility and globalization shocks

	(1)	(2)	(3)	(4)
	All firms	SG&A< median	SG&A> median	SG&A> 75thpctile
exp_SHOCK	-3.696**	-0.401**	-7.409*	-31.26*
	(1.862)	(0.172)	(4.132)	(17.53)
exp_SHOCK_X2	3.506**	0.192	7.143*	29.29*
	(1.785)	(0.210)	(3.942)	(16.99)
exp_SHOCK_X3	3.440*	0.152	7.043*	30.72*
	(1.883)	(0.193)	(4.218)	(17.62)
imp_SHOCK_X1	0.477*	0.0361*	1.045	4.879*
	(0.282)	(0.0197)	(0.671)	(2.941)
imp_SHOCK_X2	-0.498*	0.00941	-1.091	-4.616
	(0.290)	(0.0327)	(0.694)	(2.894)
imp_SHOCK_X3	-0.498*	-0.0232	-1.065	-5.069*
	(0.297)	(0.0429)	(0.688)	(3.047)
Firm FE	Y	Y	Y	Y
Country-year FE	Y	Y	Y	Y
N	36430	16997	18901	10845

Notes: This table presents results of panel 2SLS regressions examining the effect of export and import shocks vis-à-vis China on cash flow volatility, defined as the standard deviation of cash flow/asset rate over the subsequent 5-year window. Dependent variable is standard deviation of cash flow/asset ratio over the subsequent 5 years. All columns include firm- and country-year fixed effects as well as other firm-level controls such as total sales, operating cash flow. Tercile dummy variables included but not reported in this table is based on firm’s total assets (size). Standard errors in parentheses are clustered at country-sector-year level. (* p < 0.10, ** p < 0.05, *** p < 0.01)