Policy Research Working Paper 10997 Firm Size and Public Investment Multipliers Micro Evidence from Peru J. Sampi G. Vuletin J. T. Araujo Latin America and the Caribbean Region & Prosperity Vertical December 2024 Policy Research Working Paper 10997 Abstract This paper examines how the concentration of large firms a temporary adjustment occurring two years earlier. The influences the fiscal multiplier effects of public investment multiplier is significantly higher in municipalities with a in transport infrastructure. Using data from 1,891 Peru- greater concentration of large firms, highlighting the role vian municipalities and firm-level information, the analysis of firm concentration in amplifying fiscal shocks. These exploits a quasi-experimental setting stemming from an results suggest that fiscal resources may be more effectively exogenous change in the Municipality Compensation targeted to municipalities with a higher concentration of Fund in 2010, Peru’s primary fiscal transfer mechanism large firms, where the impact of public investment is stron- from the central government to subnational authorities. ger, and that policies promoting firm growth can enhance The findings show that public investment generates a posi- the effectiveness of fiscal policy by increasing the multiplier. tive fiscal multiplier four years after implementation, with This paper is a product of the Office of the Chief Economist, Latin America and the Caribbean Region and the Prosperity Vertical. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http:// www.worldbank.org/prwp. The authors may be contacted at jsampibravo@worldbank.org, gvuletin@worldbank.org, and jaraujo@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Firm Size and Public Investment Multipliers: Micro Evidence from Peru* Sampi, J.a,b , Vuletin, G.a , and Araujo, J. T.a,c a World Bank, U.S. b Vrije Universiteit Amsterdam, The Netherlands cUniversity of Brasilia, Brazil Key words : fiscal multiplier, public investment, stock of public capital, firm size. JEL codes : E22, E32, E62. * Email addresses: jsampibravo@worldbank.org (Sampi, J.), gvuletin@worldbank.org (Vuletin, G.), ja- raujo@worldbank.org (Araujo, J. T.). The findings expressed in this paper are those of the authors and do not necessarily represent the views of the World Bank or its member countries. We would like to thank Aart Kraay, William Maloney, Charl Jooste, Daniel Lederman, and Sergio Correia for valuable comments at the early stages of this paper. We are also grateful to Pilar Ruiz for excellent research assistance. 1 Introduction The economic literature generally agrees that public investment multipliers are positive or at least non-negative in both developing and advanced economies (Ramey, 2011; Kraay, 2014). At the macroeconomic level, several factors have been identified that influence the magnitude of these multipliers, including monetary policy conditions (Ramey and Zubairy, 2018; Christiano et al., 2011), the level of economic development (Ilzetzki et al., 2013), business cycle phases (Riera-Crichton et al., 2015; Berge et al., 2021), the extent of infor- mality (Colombo et al., 2024), and initial capital stock (Izquierdo et al., 2019). A related body of literature examines subnational data to explore fiscal multipliers, often by focus- ing on social security transfers (Pennings, 2021) or leveraging heterogeneity in budgetary processes to predict government spending multipliers (Clemens and Miran, 2012; Dupor and Guerrero, 2017). However, few studies investigate the microeconomic mechanisms that drive these multipliers at the firm level. Preliminary research suggests that market power can amplify fiscal multipliers (Molana and Zhang, 2001), while household credit constraints may dampen their effects (Brinca et al., 2016). Yet, the role of firm size in determining the scale of fiscal multipliers remains largely unexplored. This paper addresses this gap by empirically investigating how the concentration of large firms influences the impact of government transport investment on the value added of Peruvian firms. We leverage a novel dataset combining firm-level census records (2009–2014) with detailed fiscal transfer data from 1,891 Peruvian municipalities. Our analysis exploits a quasi-experimental setting created by an exogenous change in the distribution criteria for fiscal transfers administered by the Ministry of Economy and Finance (MEF) in 2010. The revision of the Municipality Compensation Fund (FONCOMUN)—a primary mechanism for intergovernmental fiscal transfers—provides a natural experiment to assess the effects of public investment on firm-level performance. In February 2010, the Peruvian Ministry of Economy and Finance (MEF) revised the 2 distribution criteria for the Municipality Compensation Fund (FONCOMUN), a key mech- anism for intergovernmental fiscal transfers. This revision introduced significant changes to the allocation of funds through three main channels: (i) the infant mortality rate, previously part of the FONCOMUN distribution formula, was replaced with the percentage of individ- uals with unmet needs; (ii) new incentives were introduced to stimulate local investment; and (iii) territorial extension was incorporated as an additional criterion. These modifi- cations impacted both the volume and composition of funds allocated to municipalities, including those designated for public infrastructure investments. The formula modifications led to a substantial widening of the gap between the Initial Institutional Budget (IIB, or PIA in Spanish) and the Modified Institutional Budget (MIB, or PIM in Spanish) for public investment financed by FONCOMUN resources. The IIB, announced in November each year, contrasts with the MIB, which represents the final allocated budget. Consequently, the disparity between the IIB and the MIB, expressed as a percentage of the IIB, increased from an average of 36 percent between 2007 and 2009 to 113 percent in 2010, as illustrated in Table 1. Table 1: The gap between the local municipalities’ Initial Institutional Budget (IIB) and Modified Institutional Budget (MIB) for public investment financed with FONCOMUN resources, as percentage of IIB 2007 2008 2009 2010 2011 2012 Average 21.8 43.8 43.8 113.3 35.3 68.8 Median 10.8 22.8 14.8 34.0 11.9 16.6 St. Dev. 79.8 148.4 219.0 742.1 436.5 523.2 The table presents the simple mean, median and standard deviation (St. Dev.) for all municipalities in the database. The number of municipalities varies due to recently created or eliminated municipalities, as well as data availability. The sample includes 666, 1,630, 1,630, 1,554, 1,601, and 1,447 municipalities for the years 2007, 2008, 2009, 2010, 2011, and 2012, respectively. The modification of the FONCOMUN distribution formula in 2010 created an exogenous shift for firms operating within affected municipalities, potentially influencing them by altering the availability of public capital relevant to private sector activities. Consequently, 3 this paper evaluates the impact of public investment, as mediated by the FONCOMUN shock, on firm performance. To accomplish this, the paper adopts a novel methodological framework by combining firm-level data with fiscal transfer information. Specifically, it introduces three main strategies: first, by integrating firm-level and fiscal datasets, we create a quasi-natural experiment that allows us to rigorously assess the effects of the FONCOMUN reform on firm performance. Second, we introduce a novel instrumental variable that captures differences in the budgeting process for public investment financed by FONCOMUN resources in 2010, providing an exogenous source of variation in fiscal transfers. Finally, we explore the microeconomic mechanisms underlying public investment multipliers, highlighting how the concentration of large firms amplifies the fiscal multiplier through scale effects. These findings are robust across different definitions of firm size, whether based on employee count or sales volume, and remain consistent when controlling for firm efficiency. Our analysis lies at the intersection of literature on public capital and economic growth, and on public investment multipliers. Building on the seminal contributions of Aschauer (1989) and Barro (1990), the public capital and growth literature has mainly explored the effects on the supply side, focusing on the interaction and complementarity between public and private investment. This body of research has produced largely inconclusive results on the impact of public capital on economic growth, as comprehensively reviewed by Straub (2011). In contrast, the literature on public investment multipliers has predominantly examined the Keynesian demand-side effects of fiscal stimuli, as demonstrated by Christiano et al. (2011), while other studies, such as Ramey (2020) and Ramey and Zubairy (2018), have documented the limited effectiveness of such stimuli. Vagliasindi and Gorgulu (2021) offer a thorough review of this latter body of work. This paper makes a distinctive contribution by introducing a microeconomic perspective to the study of fiscal multipliers, with a focus on firm-level responses to public investment. By highlighting the role of firm size—specifically the concentration of large firms—in de- 4 termining the magnitude of fiscal multipliers, we suggest that fiscal policy measures may be more effective when targeted to municipalities with a larger concentration of large firms. These findings are relevant for policymakers seeking to enhance the effectiveness of fiscal interventions, suggesting that supporting firm growth can indirectly amplify the impacts of public investment. The paper proceeds as follows: Section 2 discusses in detail the Peru Municipality Com- pensation Fund (FONCOMUN) and the 2010 shock. Section 3 describes the data. Section 4 presents the empirical strategy, including the instrumental variable used for identification. Section 5 reports the main empirical results, and Section 6 concludes. 2 Peru’s Municipality Compensation Fund (FONCO- MUN) and the 2010 shock as a quasi-natural exper- iment 2.1 Key features of Peru’s intergovernmental fiscal system Peru’s subnational intergovernmental fiscal system is heavily dependent on fiscal transfers from the central government to regional municipalities. Although municipalities have some autonomy to generate local revenue through mechanisms such as property taxes and service fees, a substantial portion of their funding comes from central government transfers. The primary vehicle for these transfers is the Municipality Compensation Fund (FONCOMUN, on Municipal), which constitutes approximately 20 from the Spanish Fondo de Compensaci´ percent of total municipal revenues. FONCOMUN funds are flexible, allowing allocation across various expenditure categories, including gross fixed capital formation, goods and ser- vices, wages, and salaries, and pensions. In particular, one-third of FONCOMUN resources are directed towards capital formation. Specifically, FONCOMUN contributes 11 percent of the total financing for capital formation, with an even greater role in the transport sector, 5 where it accounts for up to 14 percent of sector-specific capital expenditures. FONCOMUN was established under the 1993 Peru Constitution to improve the decen- tralization of investment promotion and regional development. The Ministry of Economy and Finance (MEF) administers FONCOMUN allocations through a numerical redistri- bution formula, allowing for modifications without congressional approval. This flexibility enables the central government to adapt the allocation criteria in response to evolving fiscal and developmental priorities. According to Peru’s 2005 General Budget Law, once a municipality allocates a por- tion of its FONCOMUN revenues to capital formation, those funds cannot be reallocated for current expenses. This obligation incentivizes subnational municipalities to carefully forecast and plan their FONCOMUN inflows before each fiscal year begins. By ensuring that funds designated for capital formation are not redirected to operational costs, the law encourages municipalities to prioritize long-term investment in infrastructure and develop- ment projects. Consequently, this system aims to promote a more strategic and efficient use of fiscal resources, supporting the broader objective of decentralizing and improving regional investment and development in Peru. Peru’s 2005 General Budget Law also outlines specific timelines and procedures for bud- get preparation and approval. The budget formulation process begins in July of the previous calendar year. During this period, each public entity, including subnational municipalities, prepares and submits its budget proposal to the Ministry of Economy and Finance (MEF) for consolidation and review. The national budget, which encompasses all public entities, is then presented to Congress for approval in November. Following Congressional approval, a preliminary version of the budget, referred to as the Initial Institutional Budget (IIB, or PIA in Spanish), is communicated to subnational entities. Although preliminary, the IIB serves as a critical reference for budget allocations at the beginning of the fiscal year. However, it is subject to further modifications throughout the fiscal year due to additional budgetary adjustments and credits issued by the MEF. This 6 revised version is known as the Modified Institutional Budget (MIB, or PIM in Spanish). Ideally, the discrepancies between the IIB and the MIB should be minimal, reflecting only necessary and anticipated adjustments, as municipalities generally expect a stable proportional increase. However, in practice, external changes, such as midyear modifications in FONCOMUN transfers, can lead to significant and unexpected differences between the IIB and MIB. These unexpected adjustments can impact municipal budget planning and execution, influencing how resources are allocated and used. The FONCOMUN allocation formula relies on a set of predictable and observable fac- tors, including demographic characteristics such as population size. This predictability has resulted in a high correlation between the Initial Institutional Budget (IIB) and the Modified Institutional Budget (MIB) for FONCOMUN revenues from year to year, as illus- trated in Figure 1, panel (a). This high correlation indicates that municipalities generally anticipate deviations in their FONCOMUN inflows with a considerable degree of accuracy. The FONCOMUN formula also accounts for the political and territorial organization of Peru by distinguishing between provinces and districts, both classified as subnational governments. Districts are grouped into provinces, with the latter playing a central role in coordinating economic policies among the constituent districts. Consequently, provinces receive specific budget allocations to manage this coordination function effectively.1 The FONCOMUN allocation mechanism operates in three phases:2 First, FONCOMUN funds are distributed among the 196 provinces based on a simple population index; second, the provinces retain 20 percent of the funds, with the remaining allocated among districts within the province, according to the percentage of people living in rural areas; finally, each district receives the greater of either the estimated amount based on the formula or 8 tax units.3 The estimated funds are then transferred in monthly installments over the fiscal 1 According to the 1981 Organic Law of Municipalities, provinces are also responsible for ensuring that subnational economic development plans align with national objectives. 2 See Article 87 of the Municipal Taxation Law. 3 The tax unit is a reference value set annually by the Ministry of Economy and Finance (MEF) to determine taxes, penalties, fines, processing fees, deductions, and other fiscal obligations. 7 year. Given the simplicity of the formula, only exogenous changes in the FONCOMUN funding sources were unpredictable for subnational entities. Figure 1: Predictability of FONCOMUN revenues (a) Before the 2010 formula change (b) During the 2010 formula change The figure plots the ratio between the Modified Institutional Budget (MIB) and the Initial Institutional Budget (IIB) in logarithms for 1834 municipalities for the years 2008, 2009 and 2010. 2.2 The FONCOMUN shock: Modification of the formula The FONCOMUN allocation formula is subject to discretionary modifications by the Min- istry of Economy and Finance (MEF), requiring only approval from the Prime Minister. In February 2010, a newly appointed Minister of Finance4 enacted a revision to the FONCO- 4 Mercedes Araoz was designated as Minister of Economics and Finance on December 22, 2009, and served as minister until September 12, 2010. 8 MUN distribution criteria through a ministerial decree.5 This modification occurred after the budget for fiscal year 2010 had been finalized and approved by Congress in November 2009. Consequently, the revised formula had an immediate and unavoidable impact on monthly transfers to subnational municipalities throughout 2010. The modifications introduced in 2010 led to a substantial increase in the unpredictability of the differences between the Initial Institutional Budget (IIB) and the Modified Institu- tional Budget (MIB) for FONCOMUN revenues, compared to the relative predictability observed in 2009 (see Figure 1). The revised FONCOMUN formula introduced three sig- nificant revisions: (i) replacing the infant mortality rate with the percentage of individuals with unmet needs as a criterion; (ii) incorporating municipal management performance metrics to incentivize local investment;6 and (iii) including territorial extension as a new consideration within the district framework (see Figure 5 in Appendix A). Although there is a notable correlation between the percentage of individuals with un- met needs and the infant mortality rate, the inclusion of municipal management perfor- mance metrics has introduced significant variability in the allocation formula. Specifically, municipalities that allocated their 2009 FONCOMUN revenues to capital formation were incentivized under the revised criteria. Figure 2 illustrates the percentage of FONCOMUN revenues directed toward capital formation in various municipalities, revealing a broad spec- trum ranging from nearly 0 to 100 percent, with few municipalities showing comparable ratios. This raises concerns that the revision of the FONCOMUN formula may have favored more productive municipalities, particularly under the assumption that high-performing firms are concentrated in these localities. However, this assumption is challenged by the significant fluctuation in public investment execution rates over time. For example, the average execution rate for public investments financed by FONCOMUN was 63.5 percent 5 Ministerial Decree 060-2010-EF 6 This criterion aims to incentivize the use of FONCOMUN resources for capital formation as a comple- ment to private investment, therefore fostering local development through public investment. 9 in 2008 but decreased to 57.2 percent in 2009. In particular, municipalities that fully executed their budgets, such as the Municipality of Salamanca in 2008, managed to execute only one-third of it in 2009. In general, municipalities with higher execution rates in 2008 did not consistently maintain these levels in 2009, as illustrated in Figure 3. Consequently, potential beneficiaries of the FONCOMUN formula revision are not necessarily those with a consistent history of high execution rates, given their temporal variability. In addition, firms typically base their location decisions not only on municipal execution rates but also on factors such as public services and product demand. As shown in Figure 6 in the Appendix, execution rates in 2009 exhibit a notable lack of correlation with firm value- added growth for that year, suggesting that there were no specifically targeted beneficiaries in the modification of the FONCOMUN formula. A similar pattern emerges when analyzing the value added per worker (see Figure 7 in the appendix). This evidence indicates that the revisions to the FONCOMUN formula did not align effectively with the promotion of growth in high-value firms, challenging the notion of targeted advantages based on execution performance. As a result of these revisions, the discrepancy between the IIB and MIB, expressed as a percentage of the IIB, rose significantly from an average of 36 percent during the 2007-2009 period to 113 percent in 2010. Further metrics, including standard deviation and median, also show a substantial and unexpected widening of this gap in 2010, as detailed in Table 1. For subnational municipalities and the firms operating within them, the 2010 modifi- cations to the FONCOMUN distribution formula constituted an unanticipated exogenous shock. This shock provides a valuable opportunity to conduct a quasi-natural experiment to examine firms’ responses to the resulting changes in public investment at the subna- tional level. The exogeneity of the modifications of the FONCOMUN rule relative to firm performance is supported by the fact that these modifications were implemented uniformly across the country and were not designed to selectively benefit high-performing munic- 10 Figure 2: Percentage of FONCOMUN revenues used to finance capital formation, 2009 The chart presents FONCOMUN expenditure allocated to capital formation as percentage of total FON- COMUN revenues per each subnational municipality in 2009. Municipalities names are codified by location identifiers available at: Peru Ubigeos. Figure 3: FONCOMUN public investment execution rates in 2008 and 2009 The chart presents the execution rates of FONCOMUN capital formation for each municipality in 2009 and 2008. Execution is defined as the payment obligation arising from previously recorded committed expenditure, regardless of whether the payment has been disbursed. 11 ipalities. The inherent randomness of the modifications, along with the impracticality of orchestrating nearly 2,000 individual negotiations with local mayors, makes targeted adjust- ments highly improbable. Thus, the exogenous nature of the FONCOMUN reform allows for a robust analysis of its impact on firm outcomes, without substantial concern for the confounding influence of deliberate favoritism or strategic manipulation. To this end, we propose an instrumental variable defined as the differences between the budgeted (IIB) and modified (MIB) public investment financed with FONCOMUN resources in 2010. The instrumental variable is specified as follows: M 1 Sj,2010 = (M IBj,2010 − IIBj,2010 ) − (M IBj,2010−t − IIBj,2010−t ) (1) M t=1 where j represents each municipality, and M describes the number of years used to calculate the historical average gap between the MIB and IIB budget numbers. Note that, since differences between the IIB and the MIB often arise from factors unrelated to the above shock – e.g. due to poor forecasting capabilities at the subnational level – we subtract the two-year average (M = 2) of such differences prior to 2010 to isolate the effects of the shock.7 The focus of this analysis on public investments in the transport sector is motivated by two main considerations. First, investments in transport infrastructure are more likely to have a direct and substantial impact on firms. Improved transportation networks can enhance logistical efficiency, reduce costs, and facilitate access to markets, thus benefiting businesses operating in affected municipalities. Second, transport investments are uniformly recorded across all municipalities, unlike other types of public investments, such as those in hospitals or schools, which are not consistently undertaken by every municipality. The uneven distribution of investments in other sectors could introduce additional variability and noise into the analysis, complicating the interpretation of the results. Consequently, the 7 Data limitations prevent us from going beyond 2 years in the past, for example, recorded information in 2007 declines to less than half of what was recorded in 2008. 12 critical differentiator among the municipalities in this study is their capacity to implement and manage strategic infrastructure projects in the transport sector. By concentrating on transport investments, the analysis minimizes potential confounding factors and allows for a clearer assessment of how variations in public investment directly influence firm performance and municipal development. Infrastructure projects can include the construction or improvement of roads connecting one or more neighborhoods, as in the case of the municipality of Villa Maria del Triunfo in the capital state (Lima), which registered an increase of 4.2 million nuevos soles8 (roughly 1.1 million USD) to improve three main avenues connecting several neighborhoods. How- ever, these projects are not limited to roads; they also include the construction or improve- ments of sidewalks that are largely absent in locations close to the jungle or mountains, which can enhance business dynamics. For example, the municipality of Iquitos saw an increase of 3.4 million nuevos soles (roughly 0.9 million USD) to improve sidewalks and on Libertad. Likewise, small bridges were roads along one of the city’s main avenues, Jir´ also included among the infrastructure projects that benefited from the increase in FON- COMUN financing. This was the case for the municipality of Chanchamayo in the state of Junin, which received 235,000 nuevos soles (63,400 USD) to build a pedestrian bridge over on La Merced. However, not all municipalities one of the city’s main avenues, Circunvalaci´ received an increase in funds. In fact, nearly half of the municipalities were negatively affected by the change in the FONCOMUN formula, requiring them to either postpone project execution or re-allocate their budgets using alternative financing sources. 8 The Peruvian currency. 13 3 Data 3.1 Firm census data The analysis is based on firm-level data from the Annual Economic Survey (EEA) collected by Peru’s National Institute of Statistics (INEI) for the 2009-2014 period. The EEA sample is drawn from a directory of formal firms, based on administrative tax records, and includes firms that exceed certain sales thresholds to ensure a robust representation of economic activity in sectors such as agriculture, manufacturing, utilities, construction, trade, trans- port, communication, and other services. In addition, the survey also considers firms below the threshold, which are randomly selected each year and are representative of each sector. These smaller firms do not significantly impact aggregate figures compared to larger firms. The threshold is set at 1,700 tax units (2 million USD), as defined by the tax authority and adjusted annually for inflation. The EEA provides unique identifiers for each firm, which allows us to follow them over the years. In addition, the data include information related to firms’ location, sector classification (ISIC Rev4), value added, capital stock, wage bill, and number of employees. While most firms in Peru are single-establishment, INEI also provides establishment-level data for a subset of variables within three sectors that are more likely to include multi-establishment firms: agriculture, manufacturing, and hotels and restau- rants. Finally, the data do not provide output quantities for individual firms. Therefore, we deflate all monetary production function variables using constant price indices at the most disaggregated level available (2-digit). 3.2 Public investment information at the municipality level Information on public expenditures for each subnational municipality is provided by the Ministry of Economy and Finance (MEF) in its website portal. The database provides detailed information for each spending category and its financing sources across different economic sectors, such as transport. This data enables us to identify how each municipality 14 allocates FONCOMUN resources. In addition, the information includes specific location codes for each municipality, preventing confusion from similar names in different states. Fiscal data is subsequently deflated by using constant price indices at the state level. 3.3 The matching process The analysis is carried out at the municipal level. This implies that firm-level information is aggregated at the municipal level by adding data from individual firms. Therefore, municipal value added (Y ) is given by Yj,t = vai,j,t (2) i ∈j where vai,j,t is the value added of firm i that operates in the municipality’s location j at time t. Finally, we match the information related to transport public investment in 2010 at the municipal level with aggregate firm information using the location codes in the firm database. After merging both datasets, the sample is reduced to 242 municipalities, where the FONCOMUN shock strategy can be applied. Notice that the number of observations in the analysis is determined by the firm database, where formal firms are mainly located in few locations. Table 2 presents the summary statistics for the main variables. In addition, we present the fiscal statistics divided by the number of municipalities that experienced a positive shock and a negative one. In particular, approximately half of the municipalities in the sample experienced a positive shock, which, unsurprisingly, led to an increase in public investment in the transport sector. Meanwhile, a negative shock mainly led to a decline in public investment. Note that the increase in firm’s value added is higher in municipalities that experienced a positive shock, suggesting possible positive multipliers. 15 Table 2: Descriptive statistics for the main firm and fiscal variables as recorded in 2010 Obs. Mean St. Dev Min Max Annual pct. change in value added 242 0.3 1.0 -0.8 7.5 FONCOMUN shock (Sj,2010 /Yj,2009 ) 242 0.6 5.2 -27.5 49.6 Delta in trans. pub. investment / Yj,2009 242 0.0 12.6 -140.7 88.9 Only if Sj,2010 >0 Annual pct. change in value added 125 0.5 1.2 -0.8 7.5 FONCOMUN shock (Sj,2010 /Yj,2009 ) 125 1.8 6.5 0.0 49.6 Delta in trans. pub. investment / Yj,2009 125 1.7 11.7 -50.3 88.9 Only if Sj,2010 <=0 Annual pct. change in value added 117 0.2 0.5 -0.8 2.2 FONCOMUN shock (Sj,2010 /Yj,2009 ) 117 -0.7 2.7 -27.5 0.0 Delta in trans. pub. investment / Yj,2009 117 -1.8 13.4 -140.7 6.5 The firm statistics are constructed by aggregating the individual information at municipal level. The FONCOMUN shock requires fiscal information for the years 2008 and 2009 for each municipality. 4 Empirical specification The integration of subnational public investment data with firm-level census information enables the estimation of fiscal multipliers at the municipal level through the application of a two-stage instrumental variable (IV) regression framework. This methodological approach is in line with the natural experiment initially proposed by Barro (1981) and subsequently refined by Ramey and Shapiro (1998) and Ramey (2011). The IV framework addresses po- tential endogeneity issues in estimating fiscal multipliers, specifically the problem of public expenditure being endogenously correlated with GDP fluctuations, which could introduce bias into the coefficient estimates. Recent advances in this area of research include the novel IV techniques introduced by Kraay (2014), who developed an approach tailored to estimate fiscal multipliers in develop- ing economies. This method leverages the temporal gaps between the commitment and ac- tual disbursement of loans by official creditors to these governments. Similarly, Sheremirov and Spirovska (2022) examined the use of military expenditures as an exogenous instru- ment to estimate fiscal multipliers in developed and developing economies. These innovative 16 methods contribute to a more nuanced understanding of fiscal multipliers by mitigating en- dogeneity concerns and providing more robust estimates of the impact of public investment on economic activity. Our methodology adopts a similar strategy using budgetary data to propose an innova- tive instrumental variable to explain variations in public investment. This approach builds on established practices in the literature, where budget information is leveraged to develop new instruments that address endogeneity concerns in fiscal analysis. The regression framework employed in this paper is specified as follows: Yj,2010 − Yj,2009 Gj,2010 − Gj,2009 = c + cs + β + ϵj,2010 (3) Yj,2009 Yj,2009 where Y represents the firm’s value added, and G the transport public investment informa- tion of municipality j in 2010. The constant cs represents the state dummies.9 Furthermore, the instrument is specifically designed to account for variations in public investment in the transport sector financed by FONCOMUN resources for the fiscal year 2010 (Gj,2010 − Gj,2009 ). This instrument is not applicable for subsequent years because, beginning in 2011, local municipalities were aware of the modifications to the allocation formula. Consequently, the nature of the change transitioned from being an exogenous shock to a foreseeable adjustment, thereby compromising the exogeneity of the instrument for those years. It is important to note that both sides of the regression equation are normalized by dividing each variable by the aggregate firm value added in 2009. This normalization allows for a relative comparison of the impact of changes in public investment. The model, specified in terms of levels differences, interprets the coefficient β as the fiscal multiplier. However, it is crucial to emphasize that the magnitude of this coefficient should not be directly compared to estimates from macroeconomic studies. This is primarily because 9 There are 25 states in Peru. 17 our analysis focuses on a specific subset of formal firms, rather than the broader economic activity that includes the informal sector. As a result, the fiscal multiplier derived from this model reflects the impact within a more narrowly defined economic context. In contrast, it is plausible that the impact of public investment may take longer than a year to fully manifest. To address this temporal dimension, we employ an approach similar to the local projection method introduced by Jorda (2005). This process allows us to track the dynamic effects of public investment on firms over time, using annual firm-level data to observe changes. In this dynamic framework, we extend the static equation (3) to incorporate a temporal dimension by applying the cumulative percentage increase in the firm’s value added, fol- lowing a strategy similar to that in Dupor and Guerrero (2017). This leads to the following specification: Yj,2010+h − Yj,2009 Gj,2010 − Gj,2009 = c + cs + β h + ϵj,2010+h (4) Yj,2009 Yj,2009 Note that when h = 0, the model simplifies to the basic equation (3). The coefficient β h traces out the impulse response function of the change in Y at time 2010 + h to a change in transport public investment in 2010, while β h captures the cumulative impact of an additional nuevo sol (in real 2010 values) invested in the transport sector at the level of Y after h periods. 5 Results In this section, we assess the impact of public investment in transport infrastructure on the value added of firms. We compare the fiscal multiplier obtained from the Ordinary Least Squares (OLS) estimation of equation (3) with the multiplier derived from the Instrumental Variables (IV) framework. Additionally, we present the dynamic fiscal multiplier as outlined 18 in equation (4). It is important to note that we do not need to report F-statistics for different years, as the instrument is only valid for 2010. Firstly, we underscore the importance of our instrument. Panel C of Table 3 presents the estimated coefficient for the FONCOMUN shock in relation to the annual changes in real public transport investment for the year 2010. This estimate is derived from a straightforward Ordinary Least Squares (OLS) regression model. To standardize our results, we normalize the variables by the aggregate value added of firms in 2009. This normalization does not alter the underlying results, but rather aligns our analysis with the conventional fiscal multiplier framework. The regression model incorporates a dummy variable for the 25 Peruvian states and uses clustered standard errors at the state level. Furthermore, we consider the exclusion of municipalities in the capital city, Lima, to demonstrate that our instrument maintains its relevance across all municipalities within the national territory. To evaluate the robustness of the FONCOMUN shock as an instrument, we conduct a weak instrument test, which involves examining the F statistic corresponding to the null hypothesis that the coefficient of the FONCOMUN shock is zero. The F statistics obtained from both regressions exceed the threshold of ten, as proposed by Staiger and Stock (1997), and the associated coefficient is highly significant, which confirms that the FONCOMUN shock qualifies as a strong instrument. The estimated fiscal multiplier is presented in panels A and B of Table 3. We consider both the OLS and IV regression results. First, we confirm that the public investment multiplier is positive but below one as in Ramey and Zubairy (2018), and that OLS estimates slightly underestimate the effects of the multiplier. A fiscal multiplier below one is not new in the literature on subnational fiscal multipliers. Clemens and Miran (2012) attribute low multiplier values to government debt crowding out current private consumption and investment. Similarly, Dupor and Guerrero (2017) find a cumulative fiscal multiplier below one using subnational information for the US. Second, we find that the coefficients are not statistically significant in either the OLS or IV regression frameworks. This lack of 19 significance may be attributed to the fact that infrastructure projects often require an extended period to translate into tangible economic effects. Table 3: Benchmark municipal fiscal multipliers in 2010 Full sample Excluding Lima Panel A. OLS estimates ∆ transport public investment (∆G/Y ) 0.00369 0.00383 (0.00932) (0.00937) Panel B. 2SLS estimates ∆ transport public investment (∆G/Y ) 0.0109 0.0108 (0.0183) (0.0185) Panel C. First stage estimates (dependent variable: ∆ transport public investment) FONCOMUN shock 1.291*** 1.292*** (0.336) (0.340) First-stage F-statistic on excluded instrument 33.93 32.90 Observations 242 197 Clustered standard errors at state level in parentheses, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010, and include state dummies. Panels A and B reports OLS and 2SLS estimates of equation (3). In addition to the static analysis, we also examine the dynamic effects of public invest- ment over time. Table 4 presents the findings from estimating equation (4) for horizons h ranging from zero to four.10 Specifically, h = 1 refers to a regression that uses firm-level data from 2011, while keeping fiscal data from 2010 constant. This allows us to explore how the initial impact of public investment evolves over time. Our analysis reveals that Ordinary Least Squares (OLS) estimates tend to underestimate fiscal multipliers, with notable exceptions at h = 3 and h = 4. This discrepancy underscores the necessity for more nuanced estimation methodologies to accurately capture the effects of public investment. In particular, a substantial instrumental variable (IV) fiscal multiplier is observed at h = 2 and h = 4, indicating a temporary negative effect on firm value added 10 The case of h = 0 is also reported in Table 3. 20 before transitioning to a positive multiplier at h = 4. This temporary negative effect may be attributed to several factors. Initially, firms may face adjustment costs associated with reallocating resources in response to public investment. These costs can manifest as operational disruptions, reduced productivity, or shifts in labor allocation, ultimately leading to a decline in value added during the early phases of implementation. Furthermore, as highlighted by Ramey (2020), infrastructure projects often encounter implementation delays that can hinder immediate positive impacts, contributing to this observed dip. During this interim period, we observe a pronounced positive multiplier in municipalities with a higher concentration of large firms, a phenomenon that we will discuss in greater detail in the subsequent section. Furthermore, firms’ adaptation processes, such as investing in new technologies or retraining employees, can also explain the temporary negative effect observed at h = 2. These adaptation processes often require time before their benefits are fully realized, suggesting that the negative impact may reflect a transitional phase as firms align with the opportunities presented by public investment. In fact, these effects typically fully materialize in subsequent years, reinforcing the notion that the benefits of public investment in the transport sector may take a minimum of four years to be fully realized by firms. This finding highlights the importance of considering the temporal dynamics of fiscal multipliers and the broader economic context in which these investments occur. The estimated fiscal multiplier at h = 4 is 0.875 and statistically significant. This implies that, over the period from 2010 to 2014, an additional nuevo sol invested in infrastructure in the transport sector translates to an approximate increase of one nuevo sol in firms’ value added after four years. This result underscores the delayed yet substantial impact of public investment on firm performance. 21 Table 4: Cumulative municipal fiscal multipliers OLS OLS 2SLS 2SLS Observations (i) (ii) (iii) (iv) h=0 Transp. Pub. Investment (G/Y) 0.00353 0.00369 0.0109 0.0109 242 (0.00941) (0.0183) (0.00932) (0.0184) D.Large*(G/Y) -0.742 0.0121 (0.888) (0.0205) h=1 Transp. Pub. Investment (G/Y) 0.0484 0.192 -0.212 -0.175 159 (0.281) (0.139) (0.224) (0.244) D.Large*(G/Y) -2.636*** 0.0714 (0.376) (0.0804) h=2 Transp. Pub. Investment (G/Y) -0.430*** -0.442*** -0.842*** -0.834*** 176 (0.0813) (0.0733) (0.284) (0.287) D.Large*(G/Y) 2.950*** 0.384 (1.002) (0.367) h=3 Transp. Pub. Investment (G/Y) -0.157 -0.192 0.300 0.193 157 (0.149) (0.114) (0.262) (0.208) D.Large*(G/Y) 7.353*** 2.181** (1.428) (0.811) h=4 Transp. Pub. Investment (G/Y) 0.777 0.438 0.875* 0.947* 184 (0.523) (0.601) (0.431) (0.469) D.Large*(G/Y) 13.17** 2.034 (6.044) (1.655) Clustered standard errors at state level in parentheses, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010 + h, and include state dummies following equation (4). 5.1 The importance of firm size Firm characteristics, particularly firm size, significantly influence how the FONCOMUN shock impacts the private sector. The findings suggest that the effectiveness of the FON- COMUN shock in generating benefits through public investment is closely related to a firm’s size. This aligns with the concept proposed by Izquierdo et al. (2019), which posits that complementarity between public and private capital requires a minimum firm size threshold, often associated with an initial stock of capital. Large firms are better positioned to take advantage of newly available infrastructure compared to smaller firms, primarily due to their scale. An important advantage of the new transport infrastructure is its ability to reduce the cost of intermediate goods by im- proving access to more affordable inputs. Larger firms, due to their scale of operations, can 22 more effectively capitalize on these cost reductions, thus reducing their production expenses without needing to make additional investments. In addition, improved transport networks facilitate recruitment from surrounding areas. This expanded access to a broader labor pool allows larger firms to improve their pro- ductivity and operational efficiency. In addition, improved infrastructure supports market expansion for larger firms by reducing logistical costs and increasing market accessibility, enabling them to enter and compete in new markets more effectively. However, the introduction of new transport infrastructure can increase competition for smaller firms. Increased competition may arise from new entrants attracted by the improved infrastructure and the potential to capture new consumer segments. Smaller firms, particularly those with lower tangible capital, might be incentivized to relocate or adjust their operations to benefit from the expanded market opportunities, even if the overall national demand for their services remains unchanged. To illustrate this concept, we introduce a dummy variable specifically designed to capture the presence of large firms within municipalities. This variable is constructed by calculating the ratio of large firms to the total number of firms in each municipality. We assign a value of one to this dummy variable if the ratio exceeds the national median, thus identifying municipalities where large firms are notably more prevalent. We define large firms as those with 100 or more employees. Subsequently, we incorporate this dummy variable into our analysis by estimating equa- tion (4), concentrating on the interaction term created by multiplying the dummy variable with the public investment variables. The results are presented in columns (ii) and (iv) of Table 4. In particular, we observe that the coefficient associated with the interaction term is consistently positive, suggesting that the presence of large firms amplifies the fiscal effects of public investment, potentially due to their greater capacity to leverage public investment for increased value creation. Specifically, the estimated fiscal multiplier for large firms in h = 3 is approximately two times higher than the overall fiscal multiplier in h = 4. This 23 implies that the presence of large firms amplifies the economic benefits of public investment, highlighting their crucial role in maximizing the impact of such investments over time. Concerns may arise regarding the concentration of large firms, primarily in the capital city, which could potentially bias the results due to sample selection. To address this issue, we examine the validity of our instrument by excluding Lima from the analysis, as shown in Table 3. This sample exclusion approach allows us to replicate our results in a manner that mitigates the influence of the capital city. The findings from this narrowed sample are detailed in Table 7 in the appendix. In particular, the positive multiplier associated with municipalities characterized by a higher proportion of large firms remains of a magnitude similar to that in the main results, with strong statistical significance at h = 3. Furthermore, consistent with our main findings, we observe a temporary reduction in cumulative firm value added at h = 2. This reduction is followed by a complete internalization of the positive multiplier effect by h = 4. These results reinforce the robustness of our conclusions, highlighting the enduring impact of large firms on regional economic dynamics, even when accounting for potential biases arising from capital-city concentration. An additional concern arises from the arbitrary cutoff used to determine whether a municipality has a larger proportion of large firms. The results in Table 4 based on the assumption that the key threshold is a proportion of large firms above the national median. To address this concern, we present results considering alternative thresholds: firms below the median, those between the median and the 75th percentile, and those above the 75th percentile. These findings are presented in Table 8 of the appendix. The results indicate that for municipalities with a proportion of large firms below the median, the coefficient is negative and statistically significant at h = 3. In contrast, for municipalities between the median and the 75th percentile, the coefficient is consistently positive and significant at h = 3. However, for municipalities above the 75th percentile, the coefficient reverses to negative, though it is not significant. This suggests that while a higher proportion of large 24 firms can enhance the fiscal multiplier, an excessive concentration of large firms may have a detrimental effect. The varying impact of firm size on fiscal multipliers can be attributed to several factors, including differences in efficiency and market structure. For instance, previous research has shown that market power can amplify the size of fiscal multipliers (Molana and Zhang, 2001), whereas credit constraints tend to limit their effectiveness (Brinca et al., 2016). Figure 4 provides an overview of firm characteristics aggregated at the municipal level, distinguishing between large firms, and medium, small and micro firms, for the year 2010. As anticipated, the efficiency of firms, measured by the value added per employee, is markedly higher in municipalities dominated by large firms compared to those dominated by medium, small, and micro firms. Similarly, the capital-to-employee ratio also reflects this disparity, suggesting that large firms achieve greater efficiency, which could skew the observed fiscal multipliers. This efficiency advantage may not necessarily stem from the effectiveness of public investment itself, but rather from inherent efficiencies in large firms, such as economies of scale facilitated by key infrastructure investments. Furthermore, the figure reveals that both intermediate costs per worker and average wages are higher in large firms. This finding challenges the notion that the fiscal multiplier might be biased upwards due to lower variable costs in large firms. Instead, it suggests that the higher operational costs associated with large firms may counterbalance any potential biases in the multiplier estimation that could arise from cheaper variable costs. Overall, these insights indicate that firm efficiency gains could indeed bias fiscal multi- pliers. The observed differences in efficiency and cost structures across firm sizes suggest that the impact of public investment on firm performance is not uniform and can be affected by their underlying characteristics. 25 Figure 4: Firms’ characteristics at the municipal level, 2010 (a) Value Added per Worker (b) Capital Stock per Worker (c) Intermediate Cost per Worker (d) Avg. Wage The figure plots the kernel density of aggregated firm information at the municipal level in 2010. 5.2 Robustness to firm efficiency In this section, we aim to validate our findings by incorporating firm efficiency into the analysis. As discussed previously, if large firms were indeed more productive than their medium, small, and micro counterparts at the time of the FONCOMUN shock, this would likely have led to a significant increase in firm efficiency over time. Such efficiency improve- ments could skew the fiscal multiplier upwards, especially in the period following the initial shock (h > 0). To address this potential bias, we use the value added per worker (Yj /Lj ) as a proxy for firm efficiency. Here, Yj denotes the total value added in the municipality j , and Lj represents the total number of employees in that municipality. By incorporating this efficiency measure, we can better assess whether the observed fiscal multiplier effects are 26 influenced by differences in efficiency levels across firms of varying sizes. This approach helps us isolate the impact of the FONCOMUN shock from the underlying productivity differences among firms, providing a more accurate estimate of the fiscal multiplier, and revealing whether scale effects are indeed present due to the presence of large firms. Therefore, we extend equation (4), as follows Yj,2010+h − Yj,2009 Gj,2010 − Gj,2009 Yj,2010+h = c + cs + β h + α log + ϵj,2010+h (5) Yj,2009 Yj,2009 Lj,2010+h where G is the transport public investment in municipality j in 2010. Firm productivity Y is represented by the value added per worker, log Lj, 2010+h j,2010+h . The constant cs represents the state dummies.11 The results are summarized in Table 5. First, our analysis indicates that Ordinary Least Squares (OLS) estimates tend to underestimate the size of fiscal multipliers in most cases, with an exception observed at h = 2. As described before, there is not a significant immediate impact of public investment on firm’s value added. Second, there is a predominantly positive correlation between improvements in firm ef- ficiency and their growth in value added, with a significant coefficient observed at h = 0. This finding indicates that increased efficiency is strongly related to enhanced value-added output for firms. Furthermore, we note a temporary negative multiplier at h = 2, which is substantially outweighed by the large positive multiplier at h = 4, confirming that the ben- efits of public investment in the transport sector require longer horizons to fully materialize. It should be noted that multipliers, adjusted for firm productivity, are magnified compared to those reported in Table 4. This suggests that controlling for productivity provides a clearer picture of the fiscal impact on value-added growth, highlighting the critical role of efficiency in maximizing the benefits of public investment. Third, the positive coefficients associated with the interaction term confirm that the 11 There are 25 states in Peru. 27 presence of large firms significantly amplifies the fiscal effects. The multiplier for large firms at h = 1 in IV results is nearly twice as high as the overall positive multiplier observed at h = 4, reflecting a trend similar to that reported in Table 4. In particular, the multiplier for large firms is significantly larger than those found in the baseline results. These findings corroborate the main conclusions outlined in Table 3. Specifically, they highlight two key insights: (i) the benefits of public investment in the transport sector need a longer time horizon to fully materialize, and (ii) the fiscal multiplier associated with large firms is significantly higher than that observed for the overall economy, even after controlling for firm productivity. This discrepancy underscores the importance of scale effects in the transmission of fiscal shocks, emphasizing that firm size is a critical determinant of the impact and effectiveness of fiscal policies. 5.3 Robustness to firm size definition In this section, we validate our findings regarding firm size by employing an alternative definition to classify large firms. The principal results presented in Table 3 are based on the criterion that firm size is determined by the number of employees, with large firms de- fined as having 100 or more employees. However, it is important to recognize that national definitions of firm size can vary and may be influenced by tax benefits and other considera- tions unique to each country. This variability highlights the importance of examining how different definitions impact firm size classification and, consequently, the associated results. In this analysis, we adopt the Peruvian legislative definition of firm size, which catego- rizes firms based on their gross sales. According to Peruvian law, firms with annual sales exceeding 2,300 tax units are classified as large. The tax authority sets the value of the tax unit annually, which means the threshold for defining large firms may vary each year. Firms with sales below this threshold are categorized as medium, small, or micro enterprises, depending on their sales volume. Following the approach in Section 5.1, we introduce a dummy variable that indicates the presence of large firms within municipalities. 28 Table 5: Cumulative municipal fiscal multipliers conditional on firm efficiency OLSOLS 2SLS 2SLS Observations (i)(ii) (iii) (iv) h=0 Transp. Pub. Investment (G) 0.00443 0.00460 0.0159 0.0159 242 (0.00816) (0.0199) (0.00806) (0.0199) D.Large*(G/Y) -0.700 0.00180 (0.975) (0.0256) VA/L, logs 0.218* 0.216* 0.234** 0.234** (0.107) (0.107) (0.108) (0.105) h=1 Transp. Pub. Investment (G) 0.0488 0.205 -0.814 -0.400 159 (0.282) (0.125) (0.875) (0.848) D.Large*(G/Y) -2.693*** 3.854*** (0.326) (1.214) VA/L, logs 0.0113 0.276 -0.227 0.0992 (0.420) (0.422) (0.517) (0.541) h=2 Transp. Pub. Investment (G) -0.433*** -0.443*** -1.031*** -1.033*** 176 (0.0835) (0.0766) (0.345) (0.349) D.Large*(G/Y) 2.934*** 2.851 (0.981) (2.312) VA/L, logs -0.315 -0.136 -0.311 -0.354 (0.635) (0.510) (0.593) (0.583) h=3 Transp. Pub. Investment (G) -0.165 -0.217* 0.707 0.675 157 (0.165) (0.118) (0.629) (0.606) D.Large*(G/Y) 7.614*** 8.198 (0.930) (10.59) VA/L, logs 0.678 1.905 0.369 0.572 (4.198) (3.869) (3.981) (4.024) h=4 Transp. Pub. Investment (G) 0.868 0.544 2.520* 2.347* 184 (0.536) (0.652) (1.225) (1.277) D.Large*(G/Y) 13.26** 26.23 (6.093) (20.19) VA/L, logs 3.921 4.746 3.946 5.286 (4.375) (4.209) (4.371) (4.028) Clustered standard errors at state level in parenthesis, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010 + h under the IV regression framework, and include state fixed effects following equation (5). Table 6 presents the results obtained using this alternative definition of firm size. Similar to previous results, the estimates suggest that Ordinary Least Squares (OLS) tends to un- derestimate fiscal multipliers compared to those derived from the Two-Stage Least Squares (2SLS) method. Furthermore, our analysis confirms the presence of a positive fiscal mul- 29 tiplier associated with large firms in municipalities, with the positive impact specifically observed at h = 3. This finding confirms the main conclusion that firm size is crucial for understanding fiscal multipliers, an angle not widely explored in the literature. Table 6: Cumulative municipal fiscal multipliers by firm size defined by sales OLS 2SLS Observations (i) (ii) h=0 Transp. Pub. Investment (G) 0.00354 0.0107 242 (0.00942) (0.0182) D.Large*(G/Y) -0.389 0.0188 (0.444) (0.0127) h=1 Transp. Pub. Investment (G) 0.246** -0.489 159 (0.0953) (0.287) D.Large*(G/Y) -3.091*** 0.135 (0.418) (0.185) h=2 Transp. Pub. Investment (G) -0.445*** -0.690* 176 (0.0713) (0.359) D.Large*(G/Y) 2.514** -0.844 (1.049) (1.073) h=3 Transp. Pub. Investment (G) -0.178 0.197 157 (0.127) (0.250) D.Large*(G/Y) 3.708 1.684* (5.206) (0.883) h=4 Transp. Pub. Investment (G) 0.723 0.899* 184 (0.502) (0.450) D.Large*(G/Y) 1.748 -4.099 (5.408) (2.663) Clustered standard errors at state level in parenthesis, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010 + h under the IV regression framework, and include state dummies following equation (4). 6 Conclusions In this study, we use a newly acquired dataset from Peru to examine the effects of public in- vestment, specifically in transport infrastructure, on the value added of firms. This analysis leverages an exogenous modification in fiscal transfers to subnational municipalities, stem- 30 ming from a 2010 adjustment to the Municipal Compensation Fund (FONCOMUN). This policy change provides a quasi-experimental framework, offering a relevant and exogenous instrument for evaluating the influence of governmental investment on firm performance within the context of a single country’s institutional environment. Our empirical findings show that public investment in transport infrastructure generates a positive fiscal multiplier, with significant effects emerging four years after implementation and a temporary adjustment two years earlier. Importantly, we observe a significantly larger multiplier in municipalities with a higher concentration of large firms, suggesting that scale effects play a crucial role in amplifying the transmission of fiscal shocks. These results are robust across various checks, including controls for firm-level efficiency and alternative firm size classifications. The findings emphasize the importance of firm concentration in determining the effectiveness of fiscal interventions. For policymakers, this implies that the efficacy of fiscal policies may vary depending on the distribution of firm sizes within the economy. In areas with a higher concentration of large firms, fiscal measures are likely to have a stronger impact, making it critical to consider firm concentration when designing and targeting fiscal policies. Moreover, policies that support firm growth could indirectly enhance the effectiveness of fiscal interventions by increasing the fiscal multiplier. Overall, this study highlights the importance of firm concentration in the economic transmission of fiscal shocks and underscores the need for policymakers to tailor fiscal interventions to the structural characteristics of their economies. 31 References Aschauer, D. A. (1989). Is public expenditure productive? Journal of Monetary Economics, 23(2):177–200. Barro, R. J. (1981). Output effects of government purchases. Journal of Political Economy, 89(6):1086–1121. Barro, R. J. (1990). Government Spending in a Simple Model of Endogenous Growth. Journal of Political Economy, 98(5):103–126. Berge, T., De Ridder, M., and Pfajfar, D. (2021). When is the fiscal multiplier high? a comparison of four business cycle phases. European Economic Review, 138:103852. Brinca, P., Holter, H. A., Krusell, P., and Malafry, L. (2016). Fiscal multipliers in the 21st century. Journal of Monetary Economics, 77:53–69. “Inequality, Institutions, and Redistribution”held at the Stern School of Business,New York University,April 24-25, 2015. Christiano, L., Eichenbaum, M., and Rebelo, S. (2011). When is the government spending multiplier large? Journal of Political Economy, 119(1):78–121. Clemens, J. and Miran, S. (2012). Fiscal policy multipliers on subnational government spending. American Economic Journal: Economic Policy, 4(2):46–68. Colombo, E., Furceri, D., Pizzuto, P., and Tirelli, P. (2024). Public expenditure multipliers and informality. European Economic Review, 164:104703. Dupor, B. and Guerrero, R. (2017). Local and aggregate fiscal policy multipliers. Journal of Monetary Economics, 92:16–30. Ilzetzki, E., Mendoza, E., and Vegh, C. (2013). How big (small?) are fiscal multipliers? Journal of Monetary Economics, 60(2):239–254. 32 Izquierdo, M. A., Lama, M. R., Medina, J. P., Puig, J., Riera-Crichton, D., Gramont, M. C. A. V., and Vuletin, G. J. (2019). Is the Public Investment Multiplier Higher in Developing Countries? An Empirical Exploration. IMF Working Papers 2019/289, International Monetary Fund. Jorda, O. (2005). Estimation and inference of impulse responses by local projections. American Economic Review, 95(1):161–182. Kraay, A. (2014). Government spending multipliers in developing countries: Evidence from lending by official creditors. American Economic Journal: Macroeconomics, 6(4):170–208. Molana, H. and Zhang, J. (2001). Market structure and fiscal policy effectiveness. The Scandinavian Journal of Economics, 103(1):147–164. Pennings, S. (2021). Cross-region transfer multipliers in a monetary union: Evidence from social security and stimulus payments. American Economic Review, 111(5):1689–1719. Ramey, V. A. (2011). Can government purchases stimulate the economy? Journal of Economic Literature, 49(3):673–85. Ramey, V. A. (2020). The Macroeconomic Consequences of Infrastructure Investment, pages 219–268. University of Chicago Press. Ramey, V. A. and Shapiro, M. D. (1998). Costly capital reallocation and the effects of government spending. Carnegie-Rochester Conference Series on Public Policy, 48:145–194. Ramey, V. A. and Zubairy, S. (2018). Government spending multipliers in good times and in bad: Evidence from us historical data. Journal of Political Economy, 126(2):850–901. Riera-Crichton, D., Vegh, C. A., and Vuletin, G. (2015). Procyclical and countercyclical fiscal multipliers: Evidence from oecd countries. Journal of International Money and 33 Finance, 52:15–31. Adjustment in the Aftermath of the Global Crisis 2008-09: New Global Order? Sheremirov, V. and Spirovska, S. (2022). Fiscal multipliers in advanced and developing countries: Evidence from military spending. Journal of Public Economics, 208:104631. Staiger, D. and Stock, J. H. (1997). Instrumental variables regression with weak instruments. Econometrica, 65(3):557–586. Straub, S. (2011). Infrastructure and development: A critical appraisal of the macro-level literature. The Journal of Development Studies, 47(5):683–708. Vagliasindi, M. and Gorgulu, N. (2021). What Have We Learned about the Effectiveness of Infrastructure Investment as a FiscalStimulus ? A Literature Review. Policy Research Working Paper Series 9796, The World Bank. 34 A Appendix A.1 Additional charts Figure 5: FONCOMUN’s transfer mechanism. Post-2010 reforms Dashed lines refer to additional considerations introduced in 2010, while the gray shadow refers to compo- nents dropped by the 2010 reforms. 35 Figure 6: Relationship between firm’s value added growth and FONCOMUN investment execution rates 36 Figure 7: Relationship between firm’s value added per worker and FONCOMUN investment execution rates 37 Table 7: Cumulative municipal fiscal multipliers by firm size excluding the capital city, Lima OLS OLS 2SLS 2SLS Observations (i) (ii) (iii) (iv) h=0 Transp. Pub. Investment (G/Y) 0.00383 0.00382 0.0108 0.0108 197 (0.00944) (0.00937) (0.0185) (0.0186) D.Large*(G/Y) -0.0268 0.0109 (0.803) (0.0211) h=1 Transp. Pub. Investment (G/Y) 0.0559 0.210 -0.230 -0.156 115 (0.291) (0.137) (0.231) (0.271) D.Large*(G/Y) -2.625*** 0.0846 (0.376) (0.0801) h=2 Transp. Pub. Investment (G/Y) -0.428*** -0.442*** -0.844*** -0.833** 132 (0.0845) (0.0753) (0.290) (0.296) D.Large*(G/Y) 3.668*** 0.359 (0.289) (0.369) h=3 Transp. Pub. Investment (G/Y) -0.153 -0.188 0.250 0.168 114 (0.156) (0.124) (0.286) (0.227) D.Large*(G/Y) 6.580*** 1.729** (1.836) (0.748) h=4 Transp. Pub. Investment (G/Y) 0.707 0.438 0.785* 0.841 140 (0.550) (0.619) (0.451) (0.494) D.Large*(G/Y) 10.88** 1.491 (5.010) (1.393) Clustered standard errors at state level in parenthesis, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010 + h under the IV regression framework, and include state dummies following equation (4). 38 Table 8: 2SLS Cumulative municipal fiscal multipliers by firm size using different thresholds <50th >50th and <=75th >75th Observations (i) (ii) (iii) h=0 Transp. Pub. Investment (G/Y) 0.0231 0.0110 0.0109 242 (0.0296) (0.0183) (0.0185) D.Large*(G/Y) -0.0121 0.0214 -0.0195 (0.0205) (0.0151) (0.0274) h=1 Transp. Pub. Investment (G/Y) -0.104 -0.230 -0.434* 159 (0.281) (0.217) (0.232) D.Large*(G/Y) -0.0714 0.0208 -0.484 (0.0804) (0.0247) (0.382) h=2 Transp. Pub. Investment (G/Y) -0.450 -0.839*** -0.866*** 176 (0.466) (0.286) (0.273) D.Large*(G/Y) -0.384 0.320 -0.785 (0.367) (0.320) (0.758) h=3 Transp. Pub. Investment (G/Y) 2.373*** 0.338 0.219 157 (0.803) (0.252) (0.222) D.Large*(G/Y) -2.181** 5.206* 0.564 (0.811) (2.826) (1.718) h=4 Transp. Pub. Investment (G/Y) 2.981 0.833* 0.340 184 (1.789) (0.428) (0.759) D.Large*(G/Y) -2.034 0.554 -9.258 (1.655) (0.970) (8.386) Clustered standard errors at state level in parenthesis, with *** p<0.01, ** p<0.05, * p<0.1. All regressions are estimated using pooled municipal data in 2010 + h under the IV regression framework, and include state dummies following equation (4). The 50th and 75th indicate the cutoff percentiles for which the threshold on the proportion of large firms within a municipality location is defined. 39