Assets prices can amplify economic cycles through their effects on firm investment. Furthermore, these effects vary depending on firm size, with larger firms being better at self-financing investments. This column uses firm-level data from France to show significant heterogeneity by firm size in the sensitivity of investment to real estate prices. Using variation in firm size distributions, it then estimates different collateral channels at the aggregate level across countries. These findings are relevant for policymakers whose decisions impact diverse regions, such as the ECB.
Asset prices significantly amplify economic cycles. When firms use assets as collateral, the value of these assets determines their ability to invest. During economic expansions, asset values increase, boosting investment and further fuelling growth. Conversely, in recessions, asset prices plummet, reducing investment and deepening the downturn. These dynamics have long been formalised by Bernanke and Gertler (1989) and Kiyotaki and Moore (1997). 1
A key takeaway is that the magnitude of this mechanism hinges on how much investment depends on the value of assets. This relationship, in turn, is influenced by a firm’s capacity to self-finance versus its reliance on its holdings of assets eligible as collateral. Evidently this dependency varies by firm: large firms are well-known to be much better at self-financing their investments, through market finance or retained earnings, compared to smaller firms.
It follows immediately that the distribution of firm sizes — or more broadly, firms’ ability to self-finance or to directly access financial markets — influences the extent to which aggregate investment (and therefore economic fluctuations) depends on asset prices. Now, firm size distributions vary greatly across countries. For example, there is extant heterogeneity in firm-level employment distributions across EU countries, as documented in Figure 1. Firms in Italy or Portugal tend to be smaller on average compared to those in Denmark or Finland.
Therefore, there are fundamental reasons why a shock affecting asset prices can lead to heterogeneous responses in business cycles across countries, based on the observed firm distribution. Such predictable heterogeneity is immediately relevant for policymakers whose decisions impact diverse regions, especially if they prefer uniform effects. A prime example is the ECB. Its monetary policy decisions affect countries with vastly different firm distributions, which means a single monetary policy shock could lead to divergent business cycles among member countries, thereby increasing the very cost of maintaining a single currency.
Figure 1 Employment share by employment category
Notes: The bars show, for each country, the share in total employment of firms belonging to each of the five employment classes. CHE = Switzerland, GBR = United Kingdom, DNK = Denmark, FIN = Finland, FRA = France, BEL = Belgium, SWE = Sweden, NLD = Netherlands, ESP = Spain, PRT = Portugal, ITA = Italy, CZE = Czech Republic.
Source: Author’s calculations, based on the CompNet database (9th Vintage)
In Héricourt et al. (2024), we explore the quantitative relevance of this mechanism. Our objective is to create a methodology using detailed firm-level data from one country to assess how responsive aggregate investment is to financing constraints in other countries. We want to achieve this using only summary characteristics of the firm size distribution and without detailed firm-level data in those countries.
We begin by estimating the response of investment to changes in collateral value at the firm level in France, focusing on quantiles of French firms from 1994 to 2015. The quantiles are chosen to split the universe of French firms into categories with heterogeneous credit constraints. The estimation results are then imputed to other countries where we have data on firm distributions but lack individual balance sheets. We assume that firms in similar positions within their respective distributions exhibit comparable investment responses to changes in collateral values. The assumption allows us to approximate the distribution of investment responses in countries without firm-level balance sheet data and to estimate the aggregate collateral channel.
There are several challenges to address before achieving this objective. First, we need firm-level data on investment and collateral values in one country, which here is France. A common measure of collateral is the value of real estate assets, as used by Chaney et al. (2012) in a sample of listed US companies. However, listed companies are likely the least affected by financial constraints, suggesting that those data provide a lower bound to the actual investment responses. Here, we combine the balance sheet data on the universe of French firms with information on local real estate prices.
Secondly, there are well-known issues of simultaneity and endogeneity between investment and real estate prices, which we address following Chaney et al. (2012). Their approach was heavily criticised by Welch (2021), and we incorporate his comments in a battery of alternative specifications that ascertain the existence of a causal and heterogeneous relationship going from real estate prices to investment in the universe of French firms.
Thirdly, to impute the effects estimated from French data to other countries, we need information on firm distributions in those countries. Specifically, we require the distribution of firms across the different quantiles used in the French data estimations, which we obtain from CompNet. We can impute the heterogeneous effects from the French data to other countries with similar financial systems, particularly those with a comparable reliance on bank finance. Beyond France, our analysis includes nine Western European countries (Belgium, Denmark, Finland, Germany, Italy, the Netherlands, Portugal, Spain, and Sweden). For comparison purposes, we also consider the Czech Republic, a former transition country where credit constraints are likely more severe than in Western Europe; Switzerland, a Western European country outside the EU; and the UK, known for its significant market finance sector.
We reach two key conclusions. First, the average sensitivity of investment to changes in real estate prices among French firms is approximately 0.2. Importantly, the estimates vary considerably with firm size: the smallest firms are at least three times more responsive to changes in collateral value compared to the largest firms. This heterogeneity is salient for a variety of measures of size, such as employment and real value added. In Figure 2, we plot the elasticities estimated by firm employment classes.
These estimates are robust across various data sources and controls. Although the estimated effects are smallest for large firms, which are presumably the least credit constrained, they remain statistically significant.
Figure 2 Investment sensitivity to changes in collateral value in France
Notes: The graph plots the estimated coefficients from regressing investment on the value of firms’ real estate holdings. All coefficients are significant at the 1% level.
Source: Héricourt et al. (2024)
The distribution of firm sizes varies across countries, as shown in Figure 1. The elasticity estimates vary along the firm size distribution, as shown in Figure 2. These two sources of heterogeneity combine to create potentially significant differences in collateral channels at the aggregate level between countries. This reasoning leads to our second main finding. We identify significant cross-country heterogeneity in the estimated reactions to collateral shocks, attributable to differences in firm distributions (see Figure 3).
Because of substantial cross-country variation in firm distributions, the sensitivity of aggregate investment to collateral shocks ranges from 0.16 in Switzerland to 0.25 in the Czech Republic. Focusing on euro area members, the estimates range from 0.18 in Finland to 0.25 in Belgium. These results are robust across various aggregation exercises using different proxies for firm size. Practically, this means that an identical shock to real estate prices (e.g. a monetary policy shock) elicits a 1.3 to 1.4 times greater investment response in Belgium or the Czech Republic compared to Finland or France.
Figure 3 Heterogeneity in the aggregate collateral channel in Europe
Notes: The bars represent country-level estimates of the collateral channel, computed by combining the results obtained from French data shown in Figure 2 with country-specific information on firm distributions provided by CompNet (9th Vintage).
Source: Héricourt et al. (2024)
One of the main sources of this heterogeneity is the significant role of small firms in determining the magnitude of the aggregate investment response to a collateral shock. This finding is somewhat unexpected, given the disproportionate influence of large firms on aggregate dynamics, as highlighted in a growing body of literature pioneered by Gabaix (2011). Even though small firms have a relatively minor impact on aggregate dynamics, they are crucial in determining the magnitude of the response of aggregate investment to collateral shocks.
The finding has interesting policy implications. The pronounced heterogeneous effects of collateral shocks across firms should be a concern for national authorities. The resulting heterogeneous effects of collateral shocks across countries should be a concern for international policymakers, particularly the ECB.
Source : VOXeu
