Policy Research Working Paper                      10925




  The Division of Revenues from Unexpected
               Demand Shocks
                                Paulo Bastos
                             Natalia P. Monteiro
                             Odd Rune Straume




Development Economics
Development Research Group
September 2024
Policy Research Working Paper 10925


  Abstract
 This paper exploits gaps between observed and recently                             in the form of higher overtime payment and base wage
 forecasted Gross Domestic Product growth in export desti-                          increases. The findings also show that there are significant
 nations to estimate the effects of unexpected demand shocks                        increases in bonus-related pay in firms managed by high-
 on worker compensation. Using employer-employee panel                              skilled managers, and the unequal average distribution of
 data, the paper finds that the revenues from these demand                          unexpected revenues is also mainly driven by wage effects in
 shocks are partly transmitted to workers in the form of                            the same subset of firms. This suggests that the way in which
 higher average wages, especially close to the top of the with-                     revenues from unexpected demand shocks are transmitted
 in-firm wage distribution. These wage responses occur both                         to workers is significantly related to managerial capabilities.




 This paper is a product of the Development Research Group, Development Economics. 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 pbastos@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
The Division of Revenues from Unexpected Demand
                                                Shocks∗
                                              Paulo Bastos†
                                          Natalia P. Monteiro‡
                                          Odd Rune Straume§




Keywords: Unexpected demand shocks, ﬁrm performance, wages, rent sharing, managers. JEL
Classiﬁcation: J2, J6, F16, F66


  ∗
     This paper is ﬁnanced by National Funds of the FCT – Portuguese Foundation for Science and Technology
within the project UIDB/03182/2020. The ﬁndings, interpretations, and conclusions expressed in this paper are
entirely those of the authors. They do not necessarily represent the views of the World Bank and its aﬃliated
organizations, those of the Executive Directors of the World Bank, or those of the governments they represent.
We remain responsible for any errors.
   †
     Development Research Group, World Bank; CEPR; and REM/UECE, pbastos@worldbank.org.
   ‡
     Department of Economics/NIPE, University of Minho, n.monteiro@eeg.uminho.pt
   §
     Department of Economics/NIPE, University of Minho, o.r.straume@eeg.uminho.pt
1        Introduction

Understanding the mechanisms by which ﬁrm-level revenue shocks impact wage setting is impor-
tant for explaining the drivers of wage diﬀerentials and wage premia across ﬁrms and workers,
how ﬁrms adjust their wage policies to absorb shocks, as well as the role played by wage rigidities
in the international transmission of business cycles. Naturally, appreciation is growing in a num-
ber of ﬁelds—macroeconomics, labor, industrial organization, as well as international trade—for
careful empirical work seeking to improve our understanding of these mechanisms.1 However,
research has faced two important challenges. First, it has been diﬃcult to quantify in a sys-
tematic but precise way the unexpected component of demand shocks at the ﬁrm-level, while
distinguishing it from anticipated changes in market conditions. This distinction is important
for establishing a causal link between changes in demand and changes in wages. Second, study-
ing the mechanisms by which these unexpected demand shocks impact wage setting requires
detailed information on the components of worker compensation.
        In this paper, we estimate the eﬀects of unexpected demand shocks on the components of
worker compensation. We also explore if and how the wage eﬀects of such shocks are related
to managerial skill. We propose a new methodology to identify the unexpected component of
demand shocks at the ﬁrm-level, exploiting forecast errors in the GDP growth of export markets.
In each destination, the unexpected component of demand shocks is measured as the diﬀerence
between the GDP growth actually observed and recent forecasts of the International Monetary
Fund. We then aggregate these shocks at the ﬁrm-year level, weighting by the initial share of
destinations in ﬁrms’ total sales. Since ﬁrms initially served diﬀerent destinations to a varying
degree, they were diﬀerentially exposed to these unexpected demand shocks across markets. We
show that these unexpected shocks tend to be uncorrelated with initial ﬁrm attributes, and
account for ﬁrm ﬁxed eﬀects throughout the estimation.
        The empirical analysis draws on an unusually rich collection of data sets for the population
of private sector ﬁrms operating in Portugal during 2006-2018. We combine a yearly ﬁrm census
with information on export transactions and employer-employee panel data. We also use an
auxiliary data set from a management survey for 2016, covering a subset of these ﬁrms, in
order to establish a correlation between managerial skill and observable management practices.
The employer-employee data allow us to distinguish between eﬀects on each of the various
components of worker compensation, including base wages, overtime pay, and other regular and
    1
    Recent contributions to this literature, reviewed in more detail below, include Card, Cardoso, Heining, and
Kline (2018), Kline, Petkova, Williams, and Zidar (2019), Aghion, Akcigit, Hyytinen, and Toivanen (2018),
Grigsby, Hurst, and Yildirmaz (2021), and Fr´  ıas, Kaplan, Verhoogen, and Alfaro-Serrano (2024).



                                                      1
irregular components of compensation. In addition, the information on schooling and detailed
occupation makes it possible to examine heterogeneity in the wage eﬀects of revenue shocks
across workers in diﬀerent positions of the wage distribution, and how these eﬀects are shaped
by the skills of top executives.
      In the empirical analysis, we ﬁnd that unexpected demand shocks have a signiﬁcant impact on
sales, employment, investment and average wages. Using employer-employee panel data, we ﬁnd
that ﬁrm revenue windfalls, in the form of unexpected demand shocks, are partly transmitted
to workers in the form of higher average wages, but in a highly unequal way, with most of the
wage increases occurring close to the top of the within-ﬁrm wage distribution. Interestingly, ﬁrm
adjustments are much larger for negative demand shocks. More speciﬁcally, a demand increase
leads to higher sales, exports, investment, employment and wages when actual demand falls
short of expectations, while we do not ﬁnd similar statistically signiﬁcant responses to demand
increases above the expected level. This suggests the presence of short-run capacity constraints
that make output harder to adjust upwards than downwards, and our results are consistent with
ﬁrms setting short-run capacity according to expected demand.
      Although parts of the aforementioned wage increases consist of increased overtime payments
in exchange for more hours worked, we also ﬁnd that unexpected demand shocks lead to signiﬁ-
cant increases in base wages and other wage components. This suggests that that ﬁrms tend to
share the unexpected changes in revenue with their workers, which is inconsistent with perfectly
competitive labor markets, and also suggests that ﬁrms have some degree of product market
power. Our analysis reveals indications of several diﬀerent channels of rent sharing. First, we
ﬁnd that increases in base wages are much stronger for ﬁrms with ﬁrm-level (as opposed to more
centralized) wage agreements, which we take as an indication of collective bargaining power. Sec-
ond, we also ﬁnd somewhat stronger and more signiﬁcant wage eﬀects for longer-tenured workers,
which might be caused by hiring or training costs that give individual bargaining power to this
type of workers. Finally, we also ﬁnd evidence of explicit proﬁt-sharing arrangements whereby
revenue windfalls are transmitted to workers through increases in performance-based payments.
However, this seems to occur only in ﬁrms with high-skilled managers. This category of ﬁrms is
also found to be the main driver of the aforementioned unequal rent distribution across workers
within the ﬁrm, thus highlighting the importance of managerial skill.
      In addition to the literature mentioned above, this paper relates to several strands of existing
research. A large body of literature seeks to quantify and explain wage diﬀerentials across ﬁrms.2
  2
    Inﬂuential contributions include, among many others, Abowd, Kramarz, and Margolis (1999) (henceforth
AKM), Card, Heining, and Kline (2013), Card, Cardoso, and Kline (2015) and Song, Price, Guvenen, Bloom, and
von Wachter (2019).


                                                    2
In a recent review of this literature, Card, Cardoso, Heining, and Kline (2018) conclude that
more research is needed applying (quasi)-experimental research designs that convincingly tease
out the mechanisms through which ﬁrm shocks are transmitted to workers. A few recent papers
have contributed to ﬁll this gap. Exploiting an unexpected devaluation of the Mexican currency,
  ıas, Kaplan, Verhoogen, and Alfaro-Serrano (2024) carefully estimate how diﬀerential export
Fr´
shocks across ﬁrms impact wage premia and worker composition.3 The results reveal that
exports have a signiﬁcant positive eﬀect on wage premia, and that the eﬀect on wage premia
accounts for essentially all of the medium-term eﬀect of exporting on plant-average wages.4
Kline, Petkova, Williams, and Zidar (2019) estimate the eﬀects of innovation on rent-sharing
using employer-employee data and rich patent information for the United States. Under the
identifying assumption that the U.S. Patent Oﬃce’s initial decision on a patent application is
as good as random (conditional on observable attributes of the application and the ﬁrm), the
authors ﬁnd that each patent-induced additional dollar of operating surplus yields a 29-cent rise
in a ﬁrm’s wage bill. Also focusing on patents, but using Finish data and a matching estimator,
Aghion, Akcigit, Hyytinen, and Toivanen (2018) ﬁnd that patents are positively associated with
earnings of the co-workers of inventors.
    Our paper makes several contributions to this literature. First, the employer-employee data
we use make it possible to examine eﬀects of unexpected demand shocks on the various com-
ponents of worker compensation (bases wages, overtime and other components of pay). Recent
work by Grigsby, Hurst, and Yildirmaz (2021) convincingly argues that this distinction is im-
portant for examining and understanding the extent of wage rigidity over the business cycle.5
Second, we exploit the role of top managers’ skills—which we show are correlated with manage-
ment practices observed in data—in shaping these eﬀects. Although a growing body of evidence
reveals that manager attributes are important for management practices and ﬁrm performance
(Bertrand and Mullainathan, 2003; Bertrand and Schoar, 2003; Bastos and Monteiro, 2011;
Bender, Bloom, Card, Van-Reenen, and Wolter, 2018), there is little evidence on whether and
how managers matter for how exogenous ﬁrm shocks are transmitted to workers. Finally, the
   3
     The broader literature on ﬁrm and labor market responses to exchange rate movements includes Revenga
(1992), Bertrand (2004), Verhoogen (2008), Brambilla, Lederman, and Porto (2012), Amiti, Itskhoki, and Konings
(2014) and Bastos, Silva, and Verhoogen (2018), among others. Most closely related to this paper, Araujo and
Paz (2014) and Macis and Schivardi (2016) use unexpected devaluations of the Italian and Brazilian currencies,
respectively, to examine eﬀects on rent sharing and workforce composition.
   4
     A related strand of work examines the relationship between exports and wages, but does not exploit quasi-
experimental variation in exports, including Schank, Schnabel, and Wagner (2007), Munch and Skaksen (2008),
Baumgarten (2013), Irarrazabal, Moxnes, and Ulltveit-Moe (2013), Helpman, Itskhoki, Muendler, and Redding
(2017).
   5
     Using administrative payroll data from the largest U.S. payroll processing company, they provide descriptive
evidence that ﬁrms use base wages to cyclically adjust the marginal cost of their workers, although about one
third of workers receive no base wage change year over year.


                                                       3
methodology we propose for identifying ﬁrm-level demand shocks has some advantages relative
to the shocks previously exploited. Although exchange rate movements and innovations leading
to patents are diﬃcult to forecast, variation over time is likely to reﬂect in part economic funda-
mentals and policy choices. These developments can be monitored and analyzed by ﬁrms, which
may therefore partly respond in anticipation.6 Identifying the eﬀects of unexpected exchange
rate movements is further complicated by the fact that they are subject to incomplete pass-
through, which may be inﬂuenced by the currency in which trade transactions are denominated
and vary across ﬁrms and markets (Amiti, Itskhoki, and Konings, 2014; Gopinath, Boz, Diez,
Gourinchas, and Plagborg-Moller, 2020).7 Unexpected GDP shocks in destinations are arguably
less subject to this concern.8
    Our paper further speaks to a recent strand of work highlighting the role of internationally
active ﬁrms in the international transmission of business cycles. Using French ﬁrm-level data,
di Giovanni, Levchenko, and Mejean (2018) show that trade linkages with a foreign country are
associated with a signiﬁcantly higher correlation between a ﬁrm and that foreign country, which
has signiﬁcant macro implications. In related work, di Giovanni, Levchenko, and Mejean (2020)
document that larger French ﬁrms are signiﬁcantly more sensitive to foreign GDP growth. Using
a quantitative model, they ﬁnd that this granularity accounts for 40 to 85% of the impact of
foreign ﬂuctuations on French GDP. Focusing on exporting ﬁrms, we contribute to this litera-
ture by estimating the response of ﬁrm-level sales, employment, investment and wages to both
expected and unexpected ﬂuctuations in foreign GDP growth. In addition, we use employer-
employee data to estimate how each of these shocks impacts the diﬀerent components of worker
compensation within and across heterogeneous exporters. As Grigsby, Hurst, and Yildirmaz
(2021) emphasize, measuring this nominal wage adjustment in micro data is key for disciplining
macroeconomic theories of nominal wage rigidity.9
   6
     Since there are no oﬃcial forecasts for bilateral exchange rates, it is often diﬃcult to isolate the unexpected
component of exchange rate shocks.
   7
     Furthermore, exchange rate movements may impact not only ﬁrm revenues in export destinations, but also
the prices of materials, components and technologies sourced from those markets (which may complement or
replace workers). It has been diﬃcult to fully distinguish between these eﬀects. Unexpected demand shocks in
destinations are arguably more likely to impact ﬁrm performance primarily through exports.
   8
     In an important contribution, Hummels, Jorgensen, Munch, and Xiang (2014) examine eﬀects of oﬀshoring and
external demand shocks on wages, using employer-employee data from Denmark. Identiﬁcation of demand shocks
exploits variation in ﬁrm-speciﬁc weighted averages of imports of particular goods by the ﬁrm’s trading partners,
using the ﬁrm’s initial shares as weights. We innovate by isolating the unexpected (and idiosyncratic) component
of demand shocks in destinations. In addition, we separately identify eﬀects on the diﬀerent components of pay,
show that wage gains accrue mainly to top earners, and to ﬁrms initially led by highly skilled top executives.
   9
     In concurrent work, Garin and Silv´ erio (2024) examine the eﬀects of idiosyncratic demand shocks associated
with the 2008 global recession on wage setting of Portuguese exporters, and ﬁnd evidence of signiﬁcant wage
adjustments within ﬁrms. The methodology we propose for systematically isolating the unexpected component of
macroeconomic demand shocks in destinations can be applied more generally to assess the eﬀects of unforecasted
(positive and negative) ﬂuctuations in GDP growth. By design, our approach is arguably less likely to capture


                                                         4
    Roadmap The paper is organized as follows. Section 2 describes the data, before Section
3 presents the method for identifying the unexpected component of demand shocks at the ﬁrm-
level. Section 4 describes the econometric model, while 5 reports the corresponding empirical
results. Section 6 concludes the paper.


2     Data

The empirical analysis in this paper combines and examines several sources of panel data from
Portugal spanning the period 2006-2018. We provide a brief description of each data source in
this section and give further details in Appendix A.3.
    Employer-employee data: Quadros de Pessoal (QP) [Personnel Records] is a high-quality
compulsory census run by the Ministry of Employment covering the population of ﬁrms with
wage earners in manufacturing and services. Each ﬁrm is required by law to provide information
on an annual basis about its characteristics and those of each individual that comprises its
workforce. Firm-level information includes annual sales, number of employees, industry code,
geographical location, date of constitution and share of capital that is foreign-owned. The set of
worker characteristics includes wages (monthly base wage, overtime pay, and other components
of pay), gender, age, schooling, date of starting, detailed occupation and hours worked. A
worker may also be matched to the ﬁrm in which he is employed. Extensive checks have been
performed to guarantee the accuracy of worker and ﬁrm data. After these checks, we kept for
analysis full-time wage earners working at least 100 hours a month, and aged between 20 and
60 years old.
    Firm census: Using common unique ﬁrm identiﬁers, we supplement the ﬁrm-year data
from QP with information from Sistema de Contas Integradas das Empresas (SCIE) [Enterprise
Integrated Accounts System], a yearly census of ﬁrms run by National Statistics Institute (INE).
The main objective of SCIE is to characterize the economic and ﬁnancial behaviour of ﬁrms.
This data set includes information on total sales, investment, employment, value added, wage
bill, industry, location, among several other variables.
    International trade statistics: We merge the above data sets with yearly data on ﬁrms’
                              ısticas do Com´
export transactions from Estat´             ercio Internacional (ECI) [Foreign Trade Statistics]
from INE. This is the country’s oﬃcial information source on imports and exports. It comprises
common shocks associated with a single global recession (e.g. ﬁnancial factors), which may plausibly have hetero-
geneous impacts across ﬁrms. The focus of our analysis is also distinct in that we separately identify the eﬀects of
these unexpected shocks on the diﬀerent components of pay, show that wage eﬀects apply mainly to top earners,
and to ﬁrms initially led by highly skilled top executives. This enable us to make further progress in identifying
speciﬁc mechanisms whereby ﬁrm-speciﬁc unexpected demand shocks impact wage setting.



                                                         5
the export ﬂows of virtually all exporting ﬁrms, and provides detailed information on the product
exported, the destination market, and the value and quantity exported. Export values in these
data are free-on-board, thus excluding any duties or shipping charges.
                                                             erito `
   Management practices survey: We further use data from Inqu´          aticas de Gest˜
                                                                   as Pr´             ao
(IPG) [Management Practices Survey] for 2016. IPG is a non-periodical survey conducted by
INE, which collects information on the perceptions of top executives about the management
practices of their ﬁrms. The 2016 survey was the ﬁrst and only of its kind collected in Portugal.
It seeks to evaluate the importance of management practices for ﬁrm productivity, as well
as other key indicators that make it possible to evaluate diﬀerences in productivity between
Portuguese ﬁrms. IPG employed a stratiﬁed sample of ﬁrms operating in Portugal covering
the whole non-ﬁnancial private sector in 2016, excluding micro ﬁrms (with fewer than ﬁve
employees). The sample is representative by sector (20 sectors corresponding of aggregations of
the 2-digit level of the CAE), ﬁrm size and age, as well as belonging (or not) to a conglomerate.
The IPG survey includes questions seeking to evaluate management practices in three main
areas: (1) Strategy, monitoring and information; (2) Human Resources; and (3) Management
and social responsibility systems. We selected 18 variables that are closely related to those
adopted in Bloom and Reenen (2007). Following their approach, our measure of management
quality was constructed by z-scoring (normalizing to mean 0, standard deviation 1) the 18
individual questions in IPG, taking averages, and then z-scoring the average. This process
yields a management practice score with mean 0 and standard deviation 1.
   Actual and forecasted GDP growth: We further use yearly information on actual and
recently forecasted GDP growth from the World Economic Outlook (WEO) of the Interna-
tional Monetary Fund (IMF). WEO is usually published twice a year (in April and Septem-
ber/October). It presents IMF staﬀ economists’ analyses of global economic developments dur-
ing the near and medium term. Every April and October, the WEO provides year-ahead and
current-year GDP growth forecasts. We refer to the year for which the forecast is being made
as the target year. Forecasts made in the Fall WEO before the target year are called year-ahead
forecasts and those made during the Spring target year are called current-year forecasts. During
our sample period, forecast data are available for 195 countries. As shown by An, Jalles, and
Loungani (2018), IMF growth forecasts are virtually identical to other growth forecasts from
the private sector, and can therefore be regarded as consensus forecasts from a wide variety of
market analysts from both the oﬃcial and private sectors. After merging these IMF data with
ECI we were left with 174 destinations, which account for 99.7% of all exports in 2006. Table
A2 reports the export shares to the main destinations in 2006, both in the full ECI data and in


                                               6
the estimation sample.


3    Methodology for identifying unexpected revenue shocks

In this section, we propose a new methodology to identify the unexpected component of demand
shocks at the ﬁrm-level, which exploits forecast errors in the GDP growth of export markets.
In each destination, the unexpected component of demand shocks is measured as the diﬀerence
between the GDP growth actually observed and the current-year growth forecast for that country
published in the Spring edition of the World Economic Outlook of the International Monetary
Fund. Speciﬁcally, the forecast error for a destination-year is deﬁned as:


      F Edt = Gdt − F Gdt ,                                                                     (1)


where F Edt denotes the forecast error for destination d in year t, Gdt denotes the GDP growth
rate of destination d in year t and F Gdt denotes the current-year growth forecast for country d in
year t. We then aggregate these destination-year forecast errors at the ﬁrm-year level, weighting
by the share of destinations in total sales of ﬁrm i in the initial year:

                  D
      W F Eit =         sdi0 F Edt ,                                                            (2)
                  d=0


where sd0 is the share of exports to destination d in total sales of ﬁrm i in 2006 (the ﬁrst year of
our data) and D is the set of destinations for which data on growth forecasts are available. Since
ﬁrms initially served diﬀerent destinations to a varying degree, they were diﬀerentially exposed
to these unexpected demand shocks across markets. Using the same weights, we also aggregate
the forecast growth (WFG) at the ﬁrm-year level.
    By its construction, the weighted forecast error constitutes an unexpected shock in export
demand and is therefore plausibly exogenous to the ﬁrm. However, one potential concern is
that diﬀerent ﬁrms might select themselves into diﬀerent export environments in a way that
make them diﬀerentially exposed to forecast errors. For example, growth-seeking ﬁrms might
choose to operate more in export markets where the expected growth is higher but also more
uncertain, thus exposing themselves to higher forecast errors. In order to check for any indica-
tions of potential selection eﬀects of this kind, we present a series of scatter plots showing the
correlation between weighted forecast errors and each of a set of baseline ﬁrm characteristics
(sales, exports, export share, employment, investments in tangible and intangible assets, value
added, value added per worker and average worker pay) based on observations in 2006 (or in the

                                                 7
ﬁrst year of observation in case of ﬁrms that entered the market later than 2006). These plots
are displayed in Figure A1 in the Appendix and do not reveal any noticeable correlation between
weighted forecast errors and any of the baseline ﬁrm characteristics. We take this as a reassuring
indication that our methodology for deﬁning unexpected demand shocks is not compromised by
ﬁrm selection eﬀects.


4     Econometric method

We now describe the econometric strategy for examining the eﬀects of unexpected demand
shocks on ﬁrm performance and worker compensation. Our baseline speciﬁcation is:


      ∆Yip = α∆W F Eip + β ∆W F Gip + ηXSi0 ∗ ιp + ζi + γjp + τrp +                  ip ,                      (3)


where Yip denotes the log of the outcome variable of interest in ﬁrm i in period p; W F Eip is the
weighted forecast error in ﬁrm i in period p; W F Gip is the weighted forecast growth in ﬁrm i
in period p; ηXSi0 ∗ ιp is the interaction between the initial export share of sales of ﬁrm i and
a period dummy; ζi is a ﬁrm ﬁxed eﬀect; γjp denotes an industry-period eﬀect; τrp denotes a
region-period eﬀect; and       ip   is the error term.
                ıas, Kaplan, Verhoogen, and Alfaro-Serrano (2024), we take 3-year period av-
    Following Fr´
erages of the corresponding ﬁrm-year variables. This makes it possible to obtain a measure
of unobserved worker skills by estimating person eﬀects (using AKM regressions) within each
3-year period, and to examine the medium-term eﬀects of more permanent unexpected demand
shocks. Furthermore, the period deﬁnition of all the independent variables is lagged one year
compared to the period deﬁnition of the dependent variable Y .10 This enables us to capture
potential lagged responses to unexpected shocks. Allowing for potentially lagged responses to
more permanent unexpected shocks is especially important for wage determination, as wage
negotiations tend to occur at a particular point in the calendar year.
    The ∆ operator denotes the linear change of a variable between each period p and period
p − 1 throughout the paper. Since the ﬁrm-level weights in (2) are ﬁxed in the initial period,
variation over time in the forecast error and forecast growth stems only from variation over
time in actual and forecasted GDP growth across destinations. This is therefore a shift-share or
Bartik-type setting in which aggregate shocks are combined with measures of shock exposure.
In our application, the sum of the exposure weights across destinations is diﬀerent from 1 and
  10
     If period p is deﬁned for the independent variables as the three years from t to t + 2, the dependent variable
Yip measures the (log of the) outcome variable of interest averaged over the years t + 1 to t + 3.



                                                         8
varies across ﬁrms, which corresponds to the incomplete shift-share case described in Borusyak,
Hull, and Jaravel (2022). We therefore adopt their proposed correction, by controlling for the
interaction between this sum (the export share of sales of the ﬁrm in the initial period) and a
period dummy in all our regressions, deﬁned as XSi0 ∗ ιp .
         The ﬁrm ﬁxed eﬀects account for the role of any time-invariant (observable or unobservable)
ﬁrm attribute. The industry-period eﬀects absorb common shocks to all ﬁrms in an industry
in each period, while the region-period eﬀects capture the impacts of common shocks across
ﬁrms operating in the same region in a given period. We report standard errors clustered by
ﬁrm. Since the forecast error is deﬁned as the diﬀerence between actual and forecasted GDP
growth in export destinations, in order to isolate its eﬀect on the dependent variable we need
to control for the weighted forecast growth, which would be expected to have a direct impact
on ﬁrm-level outcomes. Our identiﬁcation assumption is that after controlling for the weighted
forecast growth and these ﬁxed eﬀects, subsequent variations in weighted forecast errors are
uncorrelated with ﬁrm-speciﬁc shocks to ﬁrm performance and worker compensation.11


5         Results

5.1        Summary statistics

Before turning to the econometric analysis, we report descriptive statistics on several variables
underlying our empirical strategy. Our ﬁrm-level baseline estimation sample is composed of
manufacturing ﬁrms that exported in the initial year, and for which it is possible to link infor-
mation from all the data sets described above (except the IPG survey, which is available only
for a subset of ﬁrms in 2016). Table 1 reports summary statistics on ﬁrms from the estimation
sample, both in levels and in changes. These statistics reveal that there exists considerable
variation across ﬁrms and over time with regard to the weighted average of actual and forecast
growth. Table A1 in the Appendix provides summary statistics on ﬁrms in the estimation sam-
ple for each 3-year period considered in the econometric analysis. Once again, these descriptive
statistics show considerable variation across ﬁrms in the main variables of interest, both within
and across periods. In Table A2 in the Appendix we also show summary statistics for ﬁrm age,
    11
    Since isolating the eﬀects of the weighted forecast error on ﬁrm-level exports and sales requires controlling for
the weighted forecast growth, our empirical methodology is not amenable to an instrumental variable approach.
If both variables were included in the ﬁrst stage in order to predict changes in ﬁrm-level exports or sales, in
the second stage we could only identify the total eﬀect of both expected and unexpected GDP growth in export
destinations on wage setting. We therefore adopt a more reduced-form approach in which both the weighted
forecast errors and weighted forecast growth are regressed directly on wages, which allows us to isolate the eﬀect
of the unexpected component of demand shocks across destinations on wage setting.



                                                         9
location and corporate structure. Unsurprisingly, most of the observations (more than 90 per-
cent) are recorded in the three most populous regions in Portugal (Norte, Centro and Lisboa),
and the overall degree of domestic ﬁrm ownership is relatively high (almost 90 percent). Our
data refer exclusively to economic activity that takes place in Portugal, which constitutes all
economic activity for the vast majority of ﬁrms. There are some ﬁrms in our data that also have
plants located abroad, but these are very few.12

                                                 [ Table 1 here ]

      Table A3 in the Appendix reports key moments on the distribution of export destinations
of Portuguese manufacturing ﬁrms in 2006, both in the full customs data and in our estimation
sample. The main export destinations are other EU member states that are part of the eurozone
(Spain, Germany, France), but also include countries outside the eurozone and/or the EU,
notably the United Kingdom, United States, Angola and Singapore. For all destinations, export
shares in the estimation sample are remarkably similar to those in the full customs data.
      Figure A2 in the Appendix shows the variation of forecasted and actual GDP growth in
each of the top 18 destinations for Portuguese exports. We observe signiﬁcant variation across
destinations with regard to both these variables. Since ﬁrms initially served diﬀerent destinations
to a varying degree, they were diﬀerentially exposed to these unexpected demand shocks across
markets.
      Figure 1 shows that there exists signiﬁcant variation across ﬁrms in the estimation sample
with regard to weighted actual and forecast growth. The range of forecast and actual growth
is often greater than 10 percentage points, which is considerably higher than the averages for
both these variables. The range of forecast growth is especially wide (and considerably larger
than that of actual growth) in the initial years of the sample period, but remains sizeable over
the whole period.

                                                [ Figure 1 here ]


5.2      Average eﬀects on ﬁrm performance and worker compensation

We now turn to the main focus of the empirical analysis: the impacts of unexpected demand
shocks on worker compensation. Table 2 reports the point estimates on the eﬀects of the weighted
forecast error and weighted forecast growth on various measures of ﬁrm performance and average
labor costs, using data from SCIE. Columns (1) and (2) reveal that unexpected demand shocks
 12
      A total of 18 ﬁrms, four of which are foreign owned, have plants located outside Portugal.


                                                        10
in export destinations lead to increased exports and sales, with the former showing stronger
responses than the latter. The stronger eﬀects on exports would be expected given the source
of variation we are exploiting: unforeseen growth shocks in export destinations. We take these
results as reassuring conﬁrmation that our strategy for identifying unexpected export shocks is
valid. We also observe that the magnitude of the eﬀects of the forecast error on both exports
and sales is somewhat larger than that of the forecast growth. In Table A4 of the appendix, we
split ﬁrm-level export values into averages export quantities (in kilograms) and average export
prices per kilogram exported. The results reveal that the eﬀect of the demand shock on exports
is entirely a quantity eﬀect, so our estimates should be interpreted with this evidence in mind.
   Columns (3) to (8) of Table 2 also show signiﬁcant eﬀects on investment (in both tangible
and intangible assets), employment, value added, value added per worker, and average labor
costs. For all these variables, the eﬀects of the weighted forecast error are consistently stronger
than those of the weighted forecast growth.

                                         [ Table 2 here ]

   In Table 3, we use the employer-employee data to examine the eﬀects on the various compo-
nents of worker compensation and on the skill composition of the workforce. Regarding worker
compensation, the total monthly wage consists of three components: base wage, overtime pay,
and other pay (e.g., various types of bonuses). We also have information on the number of hours
worked per month, which allows us to calculate the total hourly wage. We report eﬀects on
ﬁrm-level averages of each of these variables.

                                         [ Table 3 here ]

   The estimates in columns (1) and (2) of Table 3 indicate positive and signiﬁcant eﬀects of
the forecast error on both monthly and hourly wages. Columns (3)-(4) show that this positive
average eﬀect is driven by signiﬁcant increases in both base wages and overtime pay. The other
pay coeﬃcient in column (5) also has a sizeable positive point estimate, though the eﬀect on this
residual wage component is not statistically signiﬁcant. In line with the employment responses
documented in Table 2, column (6) shows a positive impact on total hours worked.
   We also investigate whether an unexpected demand shock has any eﬀects on workers’ skill
composition. We measure skill composition in two diﬀerent ways: (i) the share of workers
with a degree, and (ii) average person eﬀects estimated through AKM models. The results are
somewhat mixed. While the estimate in column (7) does not show any signiﬁcant impact on
the share of workers with a degree, the estimate in column (8) suggests a signiﬁcantly negative

                                                 11
composition eﬀect, indicating that a positive demand shock leads to a reduction in average
worker skills. However, the estimation of person eﬀects leads to a sizable reduction in the
number of observations, and hence some caution is warranted in interpreting this result.
       Overall, our results suggest that ﬁrms respond to an unexpected demand shock partly by
asking existing workers to work longer hours and partly by employing more workers.13 On
average, workers beneﬁt from such a shock in the form of higher wage payments. Whereas
part of these wage increases accrue in the form of higher overtime payments, the positive and
signiﬁcant eﬀects on average base wages, despite some indication of reduced average worker skills,
suggest that the unexpected increase in revenue due to higher demand is partly transmitted to
workers through permanent rises in their compensation.
       As previously explained, our identiﬁcation strategy relies on creating a measure of the unex-
pected component of demand shocks at the ﬁrm-level, in order to establish a causal link between
changes in revenue and changes in wages. The importance of this strategy can be assessed by
comparing our estimates in Tables 2 and 3 with the corresponding estimates from a regression
            ıvely) measure the demand shock by the actual GDP growth in export destinations.
where we (na¨
Such estimates are presented in Tables A5 and A6, respectively, in the Appendix. We see that
these estimates are consistently smaller in magnitude than the corresponding estimates in Table
2 and 3. Although an increase in actual GDP growth in export destinations has a signiﬁcantly
positive eﬀect on sales, exports and employment, the magnitudes of these eﬀects are between 38
and 66 percent smaller. Furthermore, the estimated coeﬃcients for monthly, hourly and base
wages are not only much smaller in magnitude, but they are also statistically insigniﬁcant. This
suggests that an analysis using actual GDP growth instead of GDP forecast errors as the main
explanatory variable would strongly underestimate the wage eﬀects of changes in demand.


5.3      The division of revenues from unexpected demand shocks

We proceed by examining whether the division of unexpected demand-induced revenues inside
the ﬁrm accrues disproportionately to some groups of the workforce. In particular, we verify
if the demand shocks beneﬁt mainly the ﬁrms’ high earners, thereby contributing to increased
intra-ﬁrm wage inequality, or if the increased revenues are more evenly distributed among the
entire workforce.

                                              [ Table 4 here ]
  13
    The result in Column (5) in Table 2 shows that a positive demand shock has a signiﬁcantly positive eﬀect on
the number of workers in the ﬁrm, whereas the result in Column (6) in Table 3 shows that such a demand shock
also signiﬁcantly increases the average number of hours worked per worker.



                                                      12
   In Table 4 we show the estimated eﬀects of unexpected demand shocks on the average
monthly wages of high versus low earners within each ﬁrm, using three diﬀerent earnings thresh-
olds. Whereas the coeﬃcients are statistically signiﬁcant for both groups of workers, the magni-
tudes of these estimates display a very clear and consistent pattern. First, the estimated eﬀect
is consistently larger for the high earners than for the rest of the workforce. Furthermore, the
estimated wage eﬀect for the high earners, and the diﬀerence in the eﬀects for high and low
earners, are both monotonically increasing in magnitude as we move the high earner threshold
further towards the top end of the distribution. This suggests that, on average, the division of
revenues from unexpected demand shocks is quite unequal and disproportionately beneﬁtting
the ﬁrms’ top earners.


5.4     The importance of heterogeneity in managerial skill

We now examine if and how the distribution of rents generated by these unexpected shocks
vary systematically with the skill-level of the ﬁrms’ top managers (measured at the beginning
of the sample period), which proxies for the exogenous heterogeneity in ﬁrm capabilities that
features in Melitz (2003)-type trade models. The underlying assumption (which we explore and
discuss below) is that diﬀerences in managerial skill are systematically linked with the adoption
of diﬀerent management practices, which in turn aﬀect worker behavior and eﬀort. For manage-
ment practices that are likely to interact with worker behavior (e.g., monitoring, goal setting,
and incentive schemes), it seems reasonable to assume that the eﬀects of such practices depend
on the characteristics of the workforce. Thus, we would expect that the adoption of diﬀerent
management practices is systematically related to diﬀerences in both the skill composition of
the ﬁrm’s workforce and the structure of its pay system. Indeed, a growing body of evidence
suggests that the skills of top executives are important for management practices, employee
selection and ﬁrm performance (Bertrand and Mullainathan, 2003; Bertrand and Schoar, 2003;
Bastos and Monteiro, 2011; Bender, Bloom, Card, Van-Reenen, and Wolter, 2018). However,
there is little evidence on whether and how managers matter for how revenues from exogenous
demand shocks are transmitted to workers.


5.4.1    Measuring managerial practices and managerial skills

Following Bender, Bloom, Card, Van-Reenen, and Wolter (2018), we ﬁrst use the 2016 man-
agement survey to compute ﬁrm-level management z-scores —an index of adoption of advanced
management practices. We then link these data to the other data sets on workers and ﬁrms.
This allows us to relate measured management quality to worker and ﬁrm observables, including

                                               13
worker pay at previous employers in 2011-2016, which we use to estimate worker eﬀects in order
to infer ability. The worker eﬀects allow us to measure the quality of workers’ skills at each ﬁrm
as well as the relative quality of top managers versus other workers.
       Table A7 in the Appendix provides descriptive statistics on worker and ﬁrm attributes for
ﬁrms with management z-scores above and below the median. To proxy for worker ability, we
consider estimates of individual eﬀects from AKM models using data for the period 2011-2016.
We use a similar approach to estimate managers’ ability. Furthermore, we consider a direct
measure of ﬁrm management skills, namely the share of a ﬁrm’s managers holding a university
degree.
       Managers are identiﬁed in two alternative ways. Our main approach is to identify managers
directly by occupational category in the data (using the categories CEO and other top managers).
Although allowing for more precise identiﬁcation, the downside of this approach is a loss of
observations due to missing data. We therefore complement it by an alternative identiﬁcation
strategy which is in line with previous literature (Juhn, McCue, Monti, and Pierce, 2018), namely
to identify managers by their position in the ﬁrm’s wage distribution. In this vein, we classify
the ﬁve percent highest earners in the ﬁrm as top managers.14
       The set of ﬁrm-level attributes is composed of ﬁrm size, the share of foreign and state
capital, ﬁrm age, the percentage of female employees, export status and export share. The
summary statistics reported in Table A7 reveal that ﬁrms with higher management z-scores
tend to be larger, to have a greater share of foreign and privately-owned capital, and are slightly
more likely to be exporters and tend to employ a larger share of female workers. Turning
to worker and manager attributes, the statistics reveal that ﬁrms with above-median z-scores
tend to have a higher share of workers and managers with a degree, as well as larger average
estimates of employee and managers’ ability (as revealed by person eﬀects from AKM models).
The relationship between z-scores and observable manager characteristics is further explored
in Figure A3 in the Appendix. In this ﬁgure we show the distribution of z-scores across ﬁrms
with high- and low-skilled managers, respectively, where the former (latter) are deﬁned as ﬁrms
with a share of managers with a degree above (below) the median. These distributions are
shown when managers are identiﬁed according to occupational category (Panel A) and when
managers are identiﬁed as the top ﬁve percent earners in the ﬁrm (Panel B). In both cases,
we see that the z-score distribution of ﬁrms with high-skilled managers lies systematically to
the right of the corresponding distribution of ﬁrms with low-skilled managers. This pattern is
  14
    We have also used a stricter threshold of one percent highest earners. The results are very similar and are
available upon request.




                                                      14
generally consistent with the evidence in Bender, Bloom, Card, Van-Reenen, and Wolter (2018)
for Germany, and suggests that observed skills of top executives are systematically associated
with advanced management practices.


5.4.2      Managerial skill and the division of revenue shocks

Having shown that management z-scores tend to be higher among ﬁrms with a greater share
of highly-skilled top executives, we now examine whether and how managers matter for the
distribution of exogenous shifts in revenue. To do so, we split our estimation sample according
to whether the proportion of managers with a degree was above or below the median in 2006
(or in the ﬁrst year of observation in case of ﬁrms that entered the market later than 2006).15 , 16
       Tables A8 and A9 in the Appendix provide some summary statistics for each of the two
subsamples, with the sample split based on a deﬁnition of top managers according to occupa-
tional category in Table A8 and as the 5 percent highest earners in Table A9. In either case,
we see that, on average, ﬁrms with high-skilled managers are larger, as measured by sales and
employment, and also export and invest more, than ﬁrms with low-skilled managers. The for-
mer category of ﬁrms also tend to employ more skilled workers and pay them better along all
dimensions of worker compensation.
       When estimating (3) on each of the two subsamples, the results reported in Tables A10 and
A11 in the Appendix show that the baseline results on the eﬀects of unexpected demand shocks
on ﬁrm performance generally apply to both subsamples. Regardless of the method used for
identifying top managers, the forecast error has a signiﬁcant eﬀect on exports and sales for both
categories of ﬁrms, which suggests that our strategy for identifying unexpected demand shocks
is valid also for these more restricted subsamples.
       Having established the general validity of our identiﬁcation strategy, our main objective in
this part of the analysis is to re-estimate the results in Table 4 using the above explained sample
partition, in order to examine whether and how the within-ﬁrm distribution of the revenues
from unexpected demand shocks is related to managerial skills. The resulting estimates for the
two alternative sample partitions are shown in Tables 5 and 6, respectively.

                                           [ Table 5 and 6 here ]
  15
     When identifying managers by occupational category, the classiﬁcation of ﬁrms as being managed by high-
skilled vs low-skilled managers is made according to the ﬁrst year in which a managerial occupation is observed.
  16
     Whether we identify managers by occupational category or by their position in the wage distribution, the
share of managers with a degree is zero for more than half of the ﬁrms in our sample in 2006, and is thus zero
also for the median ﬁrm. This implies that all ﬁrms in which no manager has a degree are classiﬁed as ﬁrms with
low-skilled managers, whereas all ﬁrms in which at least one manager has a university degree are classiﬁed as
ﬁrms with high-skilled managers.


                                                      15
   In the case where managers are identiﬁed according to occupational category (Table 5), the
results are quite striking. In ﬁrms with high-skilled managers, we ﬁnd a strong and signiﬁcant
eﬀect of an unexpected demand shock on the wages of the top earners in the ﬁrm, but no
signiﬁcant eﬀect on the wages of the remaining workforce (with point estimates much closer to
zero). Thus, for this subset of ﬁrms, the wage eﬀects appear to be even much more unequal
than what is reported in Table 4 using the full sample of ﬁrms. For the remaining subset
of ﬁrms, however, the picture is very diﬀerent. In ﬁrms with low-skilled managers, there are
no indications of the top earners beneﬁting more from an unexpected demand shock than the
remaining workers in the ﬁrm. In fact, all the point estimates are negative, and more so for the
top earners, though none of these eﬀects are statistically signiﬁcant.
   If we instead identify managers by their position in the wage distribution (Table 6), the point
estimates are once more much larger for high earners than for the remaining workforce in ﬁrms
with high-skilled managers, while these diﬀerences are much smaller in ﬁrms with low-skilled
managers. However, the main diﬀerence from Table 5 is that the wage eﬀects in the former
category of ﬁrms are now much less precisely estimated and thus not signiﬁcantly diﬀerent from
zero. Nevertheless, when seen in conjunction, the results in Tables 5 and 6 give some indications
that the distribution of revenues from unexpected demand shocks is related to managerial skills,
with the overall unequal distribution of rents seemingly being driven by ﬁrms with high-skilled
managers, where unexpected demand shocks to a much larger extent beneﬁts workers close to
the top end of the earnings distribution.
   In order to further explore the role of managerial skill in the division of the revenues from
unexpected demand shocks, we decompose the total wage eﬀects reported in Tables 5 and 6,
showing the eﬀects on each wage component for diﬀerent types of workers (top earners versus
the remaining workers) across the two categories of ﬁrms (managed by high-skilled versus low-
skilled managers). These results are displayed in Table 7 (where managers deﬁned according to
occupational category) and Table 8 (where managers are deﬁned as the top ﬁve percent earners).
We show these results for the intermediate wage distribution threshold (top 15 percent earners
versus the rest of the workforce), but the estimates are fairly similar if we use diﬀerent thresholds.

                                       [ Table 7 and 8 here ]

   Although there are some diﬀerences in the estimates depending on the strategy for identify-
ing top managers, the results in Tables 7 and 8 suggest once more that managerial skill might
play a key role for the way in which revenues from unanticipated demand shocks are transmitted
to workers. In ﬁrms with high-skilled managers, there is little or no evidence that rents are dis-


                                                 16
tributed via increases in base wages. Instead, wage increases apparently occur through increases
in overtime and/or other pay. In particular, an increase in other pay, which includes various
types of bonuses, is a consistent ﬁnding for this category of ﬁrms, regardless of how managers are
identiﬁed. In contrast, there is no indication that this type of pay increases plays an important
role in ﬁrms with low-skilled managers. For these ﬁrms, the signiﬁcant wage increases seem to
come in the form of either overtime payment or higher base wages.
   Overall, these results suggest that high-skilled managers to some extent adopt diﬀerent
remuneration schemes for workers than what low-skilled managers do, and that the way in
which revenue windfalls are distributed to workers is therefore partly related to managerial skill.


5.5   Discussion

The main result of our analysis is that revenues from unexpected demand shocks are partly
transmitted to workers in the form of higher average wages, but seemingly in a highly unequal
way, with most of the wage increases occurring close to the top of the within-ﬁrm wage distri-
bution. A ﬁrst basic observation is that this result is inconsistent with the notion of perfectly
competitive labor markets. If ﬁrms are wage takers in a market where wages reﬂect workers’
skills, unexpected ﬁrm-speciﬁc demand shocks would aﬀect wages only if they lead to changes
in the skill composition of the ﬁrm’s workforce. Since we ﬁnd no evidence of an increase in
average worker skills in response to a demand shock (on the contrary, we ﬁnd some indications
of the opposite), our results suggest that wages to some extent reﬂect some kind of rent sharing
between ﬁrms and their workers.
   The evidence of rent sharing in wage determination found in our analysis adds to an already
sizeable empirical rent-sharing literature (see Card, Cardoso, Heining, and Kline (2018), for
an overview). The exact mechanisms by which rents are shared with workers are potentially
numerous, though, and not necessarily mutually exclusive. First, rent sharing could be a result
of worker bargaining power, either collectively or individually. Collective bargaining power
results from the presence and inﬂuence of trade unions in wage determination, while individual
bargaining power could result from labor market frictions created by hiring and/or training
costs, which create rents that can be captured by incumbent workers (see, e.g., Kline, Petkova,
Williams, and Zidar (2019)). Second, rent sharing could also be the result of explicit proﬁt-
sharing arrangements, for example in the form of performance-based pay contracts, which may
be used to increase productivity (Lazear, 1986, 2000). Third, wage determination might partly
result from ﬁrms’ incentives to induce the desired amount of eﬀort from its workers, in line
with the fair wage hypothesis of Akerlof and Yellen (1990). If workers’ notion of a fair wage


                                                17
is based on an internal reference which reﬂects the ﬁrm’s ability to pay, such as revenues per
worker, the revenue windfall from an unexpected demand shock would be partly transmitted
to workers through an increase in what is considered to be a fair wage.17 Fourth, rent sharing
could be due to ﬁrms having some monopsony power in the labor market, for example because
of market concentration on the demand side or heterogeneous job preferences on the supply side
(Manning, 2021). If each ﬁrm faces an upward sloping labor supply curve, the revenues from an
unexpected demand shock will be (partly) transmitted to workers in the form of higher wages
via higher labor demand.
       We would argue that each of the above suggested mechanisms is plausible in our context,
despite the importance of industry-level collective bargaining in Portugal. Although trade union
density is very low, a high share of wage contracts are determined by collective agreements at
industry level, which might suggest that there is limited room for wage adjustments in response
to ﬁrm-speciﬁc shocks.18 However, it is worth emphasizing that wage determination in Portugal
is characterized by a two-tiered wage setting system where ﬁrm-speciﬁc arrangements result in a
mark-up, often of considerable magnitude, on top of the bargained wage ﬂoor.19 Thus, although
the presence of collective bargaining might result in some downward wage rigidity, the two-tiered
wage setting system still leaves considerable room for ﬁrm-speciﬁc adjustments to ﬁrm-speciﬁc
shocks.
       The diﬀerent suggested mechanisms are also, to some extent, related to diﬀerent wage com-
ponents. In particular, explicit proﬁt-sharing agreements in the form of performance-based pay
would be captured by the wage component labeled other pay, which includes various types of
bonus payments. In our analysis we do ﬁnd some evidence that this is a relevant mechanism
for rent sharing, but also that the existence of this mechanism crucially relies on managerial
skill. More speciﬁcally, the previously discussed results in Table 7 and 8 suggest that ﬁrms with
high-skilled managers to a larger extent adopt performance-based payment schemes and that
this is a signiﬁcant channel for rent distribution in such ﬁrms.
       In the following we exploit our data to look for indications of the relevance of some of
the other rent sharing mechanisms mentioned above, and which are not necessarily related to
managerial skill. In particular, we examine whether our results could be explained by individual
  17
     For fair wage models based on a ﬁrm-internal point of reference, see, e.g., Danthine and Kurman (2006),
Bastos, Monteiro, and Straume (2009) and Egger and Kreickemeier (2009).
  18
     Portugal and Vilares (2013) report a union coverage rate of more than 90 percent despite a union density
rate of only 11 percent.
  19
     See Cardoso and Portugal (2005) and Bastos, Monteiro, and Straume (2009). In the latter study, using data
for the period 1991 to 2000, wages are found to be more than 25 percent higher than the bargained wage ﬂoor,
on average.




                                                     18
or collective bargaining power. If rent sharing is caused by hiring or training costs that gives
individual bargaining power to longer-tenured workers, we would expect that the wage eﬀects of
a demand shock are stronger for such workers than for newly hired workers. This hypothesis is
explored in Table 9, where we show the decomposed wage eﬀects across these two categories of
workers, and where we have classiﬁed newly hired workers as workers with less than four years
of tenure in the ﬁrm.20

                                                 [ Table 9 ]

       The results in Table 9 show that the wage eﬀects (which all have positive point estimates)
tend to be statistically signiﬁcant for longer-tenured workers but not for the newly hired ones,
and the point estimate for the base wage increase is also larger in magnitude for longer-tenured
workers. These results give some support to the hypothesis that rent sharing is related to worker
tenure.
       An alternative hypothesis is that rent sharing is caused by collective bargaining power. This
hypothesis can be pursued by exploiting some heterogeneity in the type of wage agreements
that exist in our data. Whereas most of the collective wage agreements in Portugal are made at
industry or sectorial level, as previously mentioned, there is also a small prevalence of ﬁrm-level
wage agreements. If trade unions are able to extract some of the rents related to ﬁrm-speciﬁc
demand shocks, we would expect that the resulting wage eﬀects are stronger for workers whose
wage contracts are bargained at ﬁrm level, all else equal. We explore this hypothesis in Table 10,
where we show the estimated wage eﬀects for workers with ﬁrm-speciﬁc collective agreements
relative to workers whose wage contracts result from collective bargaining at a more centralized
level.21

                                             [ Table 10 here ]

       Interestingly, despite a very low number of observations in the group of ﬁrms with ﬁrm-
speciﬁc collective agreements, we see that the eﬀect of a demand shock on base wages is statis-
tically signiﬁcant and much larger in magnitude for this category of ﬁrms than for ﬁrms whose
wages are bargained at a more centralized level, which we take as an indication of collective bar-
gaining power. It is also worth noticing the large and signiﬁcant eﬀect of an unexpected demand
shock on other pay in ﬁrms with ﬁrm-level wage agreements, which, besides indicating collective
bargaining power, suggests that performance-based pay is imbedded in the wage contracts in
this type of ﬁrms.
  20
    A similar deﬁnition is used by Kline, Petkova, Williams, and Zidar (2019).
  21
    According to our classiﬁcation, ﬁrms with ﬁrm-level wage agreements are those in which at least 50 percent
of the workers have this type of wage contracts.


                                                     19
5.6     Extensions

We now extend our main analysis by exploring two diﬀerent aspects of the demand shocks that
might be potentially relevant, namely the persistence and the sign of the shocks. In doing so,
we restrict attention to our main results presented in Tables 2 and 3.


5.6.1    Persistent versus transitory shocks

How strongly ﬁrms respond to a demand shock will likely depend on the persistence of the shock,
particularly in the presence of non-trivial adjustment costs. We explore the importance of shock
persistence by estimating a version of (3) where we interact the forecast error and forecast growth
variables with dummy variables indicating whether a given ﬁrm-period observation is classiﬁed
as a persistent shock or not. We deﬁne a persistent shock as a ﬁrm-period observation for which
the three yearly forecast errors satisfy the following two conditions: (i) the forecast error has
the same sign in at least two consecutive years, and (ii) the sign of the errors referred to in (i)
is equal to the sign of the 3-year period average.

                                     [ Table 11 and 12 here ]

   The results are presented in Tables 11 and 12 and show a very clear and consistent pattern.
Unexpected demand shocks have signiﬁcantly positive eﬀects on sales, exports, investments,
employment and wages only when these shocks are persistent (according to our deﬁnition). The
importance of shock persistence is particularly pronounced for the wage eﬀects, as we ﬁnd that
persistent demand shocks have signiﬁcantly positive eﬀects on monthly, hourly and base wages,
with coeﬃcient estimates that are markedly larger in magnitude than the baseline estimates
reported in Table 3. Overall, these results are quite intuitive, and the consistency between the
results in Table 11 and Table 12 is reassuring.


5.6.2    Positive versus negative demand shocks

We also explore whether ﬁrms respond diﬀerently to positive and negative demand shocks, as
deﬁned by the sign of the 3-year period average of the forecast error. We do so by estimating
a version of (3) where we interact the forecast error and forecast growth variables with dummy
variables indicating whether the 3-year period average of the forecast error is positive or negative.
The resulting coeﬃcient estimates are displayed in Tables 13 and 14.

                                     [ Table 13 and 14 here ]



                                                  20
   These results show a quite consistent pattern, as most of the statistically signiﬁcant eﬀects
are due to variations in the negative demand shocks. In other words, variations in demand
have statistically signiﬁcant eﬀects on sales, exports, investments, employment and wages only
if actual demand falls short of expectations. On the other hand, variations in demand above the
expected level have few signiﬁcant eﬀects on our dependent variables. These results strongly sug-
gest the presence of short-run capacity constraints that make it harder to adjust output upwards
than downwards. One plausible interpretation is that ﬁrms set their short-run capacity levels
according to expected demand, which allows for short-run adjustments to demand ﬂuctuations
below, but not above, the forecasted level.


5.7     Robustness

In the following we present some robustness checks to the baseline estimates, once more focusing
on the results in Tables 2 and 3.


5.7.1    Alternative forecast error weights

In our main analysis we have identiﬁed the unexpected demand shocks by creating a weighted
forecast error variable where destination-year forecast errors at the ﬁrm-year level are weighted
by the share of destinations in total sales (in the initial year). By using weights that are based on
total sales, we implicitly assume that export-related demand shocks have a larger impact on more
export-intensive ﬁrms, all else equal. Although this seems like a highly reasonable assumption,
our approach might be less appropriate if ﬁrms make separate strategic decisions (regarding
wage contracts, hiring decisions, investment, etc.) for the export-oriented part of the business.
Thus, as a robustness check, we have redone our baseline analysis using alternative forecast error
weights, where destination-year forecast errors are weighted by the share of destinations in total
exports (again in the initial year). The results from these alternative estimations are presented
in Tables A12 and A13 in the Appendix.
   Although there are some diﬀerences compared with the baseline results in Tables 2 and 3,
these new estimates paint the same overall picture. An unexpected positive demand shock leads
to signiﬁcantly higher sales, exports and employment, and it leads to a signiﬁcant increase in
both monthly and hourly wages. The main diﬀerence from the baseline results is the magnitude
of these eﬀects, which are generally smaller when using the alternative forecast error weights.
These diﬀerences in magnitudes are consistent with our original rationale for basing the forecast
error weights on total sales. If export-related demand shocks have larger eﬀects on more export-
intensive ﬁrms, such diﬀerences will be downplayed when basing the forecast error weights on

                                                 21
total exports instead of total sales, thus reducing the magnitude of the estimated eﬀects of the
shocks.


5.7.2     Controlling for industry-by-region period eﬀects

We have also estimated a more restrictive empirical speciﬁcation where we control for industry-
by-region period eﬀects in order to take into account that labor markets might be segmented
by industry and region. The results from these estimations are reported in Tables A14 and
A15 in the Appendix. They show that all estimates are very similar in terms of both precision
and magnitude. Thus, our main results are highly robust to this more restrictive empirical
speciﬁcation.


5.7.3     Lagged responses

We also explore if and how our baseline results depend on the decision to include a time lag on
the independent variables. In our baseline analysis, the period deﬁnition of all the independent
variables is lagged one year compared to the period deﬁnition of the dependent variable, which
allows us to capture lagged responses to demand shocks. However, the length of the adjustment
period might diﬀer across diﬀerent types of variables. For example, while it is reasonable to
assume that wage adjustments and investment responses might take some time, sales and export
adjustments might happen much faster. In order to investigate this further, we have re-estimated
the eﬀects of demand shocks on the various ﬁrm performance measures using a model without any
time lags. The results are reported in Table A16 in the Appendix. For several of the variables,
such as sales, exports and employment, the estimated eﬀects with lags (Table 2) and without lags
(Table A16) are qualitatively and quantitatively very similar. On the other hand, the eﬀect on
investments ceases to be statistically signiﬁcant in a model without time lags. Furthermore, the
eﬀect on average wages is also smaller in magnitude and less precisely estimated. These results
are consistent with our conjecture that wage and investment responses to a demand shock are
likely to take longer time, on average, than adjustments on other variables.


5.7.4     Excluding the years of the global ﬁnancial crisis

A ﬁnal worry is that our results might be excessively driven by ﬁrms’ exceptional adjustments
to policies during years of large global recessions, in particular the 2007-2008 global ﬁnancial
crisis which occured during the ﬁrst of our four 3-year periods (see Table A1). In order to
address this potential worry, we have re-estimated our main results (as reported in Table 2
and 3) on a smaller sample where we exclude the ﬁrst period (2007-2009). These results are

                                              22
reported in Table A17 and A18 in the Appendix. Despite losing more than one third of our
observations, unexpected demand shocks still have a signiﬁcantly positive eﬀect on sales and
exports, with coeﬃcient magnitudes that are very similar for exports and even larger for sales.
Furthermore, we ﬁnd signiﬁcantly positive eﬀects on monthly and hourly wages that are larger
than our baseline estimates. Overall, these results suggest that our results are not particularly
driven by exceptional adjustments during the global ﬁnancial crisis.


6    Conclusion

In this paper, we examined the eﬀects of unexpected demand shocks on worker compensation.
We proposed a new methodology to identify the unexpected component of demand shocks at
the ﬁrm-level, exploiting errors in the oﬃcial forecasts in the GDP growth of export markets. In
each destination, the unexpected component of demand shocks was measured as the diﬀerence
between the GDP growth actually observed and recent forecasts published in the World Eco-
nomic Outlook of the International Monetary Fund. We then aggregated these shocks at the
ﬁrm-year level, weighting by the initial share of destinations in ﬁrms’ total sales. Since ﬁrms
initially served diﬀerent destinations to a varying degree, they were diﬀerentially exposed to
these unexpected demand shocks across markets.
    In the empirical analysis, we found that unexpected demand shocks have signiﬁcant eﬀects
on sales, employment, investment and average wages. Using employer-employee panel data,
we reported evidence that revenues from unexpected demand shocks are partly transmitted to
workers in the form of higher average wages, but in a highly unequal way, with most of the wage
increases occurring close to the top of the within-ﬁrm wage distribution. These wage increases,
which include an increase in average base wages, occur despite some evidence that the skill
composition of the workforce is negatively aﬀected by unexpected demand shocks.
    These results suggest that wages to some extent reﬂect rent sharing between ﬁrms and their
workers, and our analysis indicates several diﬀerent channels through which such rent sharing
occurs. One channel is collective bargaining power, which is corroborated by our ﬁnding that
positive base wage responses to demand shocks are much stronger in ﬁrms with ﬁrm-level wage
agreements than in ﬁrms that are subject to more centralized bargaining schemes. Another
channel is individual bargaining power created by hiring or training costs, which is consistent
with our ﬁnding that wage responses tend to be stronger for longer-tenured workers. Finally,
for a subset of the ﬁrms, we also found signiﬁcant increases in bonus-related wage components
in response to an unexpected demand shock, which indicates rent-sharing through the use of



                                               23
performance-based payment schemes. However, this ﬁnding is restricted to the subset of ﬁrms
that are run by high-skilled managers. We also found that the unequal distribution of rents, with
the largest wage increases occurring towards the top of the wage distribution, is to a large extent
driven by the same subset of ﬁrms. Thus, another main ﬁnding of our study is that the way
in which revenues from unexpected demand shocks are transmitted to workers is signiﬁcantly
related to heterogeneity in ﬁrm capabilities, as proxied by managerial skill.



References
Abowd, J., F. Kramarz, and D. Margolis (1999): “High Wage Workers and High Wage Firms,”
  Econometrica, 67(2), 251–333.
Aghion, P., U. Akcigit, A. Hyytinen, and O. Toivanen (2018): “On the Returns to Invention
  within Firms: Evidence from Finland,” American Economic Review Papers and Proceedings, 108(2),
  208–212.
Akerlof, G. A., and J. Yellen (1990): “The Fair Wage-Eﬀort Hypothesis and Unemployment,”
  Quarterly Journal of Economics, 105(2), 255–283.
Amiti, M., O. Itskhoki, and J. Konings (2014): “Importers, Exporters, and Exchange Rate Discon-
  nect,” American Economic Review, 104(7), 1942–1978.
An, Z., J. T. Jalles, and P. Loungani (2018): “How Well do Economists Forecast Recessions?,”
  IMF Working Paper 18/39.
Araujo, B., and L. Paz (2014): “The Eﬀects of Exporting on Wages: An Evaluation Using the 1999
  Brazilian Exchange Rate Devaluation,” Journal od Development Economics, 111(C), 1–16.
Bastos, P., and N. P. Monteiro (2011): “Managers and Wage Policies,” Journal of Economics and
  Management Strategy, 20(4), 957–984.
Bastos, P., N. P. Monteiro, and O. R. Straume (2009): “Firm Heterogeneity and Wages in
  Unionised Labour Markets: Theory and Evidence,” Labour Economics, 16(4), 440–450.
Bastos, P., J. Silva, and E. Verhoogen (2018): “Export Destinations and Input Prices,” American
  Economic Review, 108(2), 353–392.
Baumgarten, D. (2013): “Exporters and the Rise in Wage Inequality Evidence from German Linked
  Employer-Employee Data,” Journal of International Economics, 90(1), 201–217.
Bender, S., N. Bloom, D. Card, J. Van-Reenen, and S. Wolter (2018): “Management Practices,
  Workforce Selection, and Productivity,” Journal of Labor Economics, 36(S1), S372–S409.
Bertrand, M. (2004): “From the Invisible Handshake to the Invisible Hand? How Import Competition
  Changes the Employment Relationship,” Journal of Labor Economics, 2(22), 723–765.
Bertrand, M., and S. Mullainathan (2003): “Enjoying the Quiet Life? Corporate Governance and
  Managerial Preferences,” Journal of Political Economy, 111(5), 1043–1075.
Bertrand, M., and A. Schoar (2003): “Managing with Style: The Eﬀect of Managers on Firm
  Policies,” Quarterly Journal of Economics, 118(4), 1169–1208.
Bloom, N., and J. V. Reenen (2007): “Measuring and explaining management practices across ﬁrms
  and countries,” Quarterly Journal of Economics, 122(4), 1351–1408.
Borusyak, K., P. Hull, and X. Jaravel (2022): “Quasi-experimental Shift-Share Research Designs,”
  Review of Economic Studies, 89(1), 87–130.
Brambilla, I., D. Lederman, and G. Porto (2012): “Exports, Export Destinations and Skills,”
  American Economic Review, 102(7), 3406–3488.
Card, D., A. R. Cardoso, J. Heining, and P. Kline (2018): “Firms and Labor Market Inequality:
  Evidence and Some Theory,” Journal of Labor Economics, 36(S1), S13–S70.


                                                24
Card, D., A. R. Cardoso, and P. Kline (2015): “Bargaining, Sorting, and the Gender Wage Gap:
  Quantifying the Impact of Firms on the Relative Pay of Women,” Quarterly Journal of Economics,
  131(2), 633–686.
Card, D., J. Heining, and P. Kline (2013): “Workplace Heterogeneity and the Rise of West German
  Wage Inequality,” Quarterly Journal of Economics, 128(3), 967/1015.
Cardoso, A. R., and P. Portugal (2005): “Contractual Wages and the Wage Cushion under Diﬀerent
  Bargaining Settings,” Journal of Labor Economics, 23(4), 875–902.
Danthine, J.-P., and A. Kurman (2006): “Eﬃciency wages revisited: the internal reference perspec-
  tive,” Economics Letters, 90(2), 278–284.
di Giovanni, J., A. A. Levchenko, and I. Mejean (2018): “The Micro Origins of International
  Business Cycle Comovement,” American Economic Review, 108(1), 82–108.
          (2020): “Foreign Shocks as Granular Fluctuations,” NBER Working Paper 28123.
Egger, H., and U. Kreickemeier (2009): “Firm heterogeneity and the labor market eﬀects of trade
  liberalization,” International Economic Review, 50(1), 187–2016.
Fr´ıas, J. A., D. S. Kaplan, E. Verhoogen, and D. Alfaro-Serrano (2024): “Exports and
  Wage Premia: Evidence from Mexican Employer-Employee Data,” Review of Economics and Statistics,
  forthcoming.
Garin, A., and F. Silve   ´rio (2024): “How Responsive are Wages to Demand within the Firm? Evidence
  from Idiosyncratic Export Demand Shocks,” Review of Economic Studies, forthcoming.
Gopinath, G., E. Boz, C. C. F. Diez, P.-O. Gourinchas, and M. Plagborg-Moller (2020):
  “Dominant Currency Paradigm,” American Economic Review, 110(3), 677–719.
Grigsby, J., E. Hurst, and A. Yildirmaz (2021): “Aggregate Nominal Wage Adjustments: New
  Evidence from Administrative Payroll Data,” American Economic Review, 111(2)(2), 428–471.
Helpman, E., O. Itskhoki, M.-A. Muendler, and S. J. Redding (2017): “Trade and Inequality:
  From Theory to Estimation,” Review of Economic Studies, 84, 1–36.
Hummels, D., R. Jorgensen, J. Munch, and C. Xiang (2014): “The Wage Eﬀects of Oﬀshoring:
  Evidence from Danish Matched Worker-Firm Data,” American Economic Review, 104(6), 1597–1629.
Irarrazabal, A., A. Moxnes, and K. H. Ulltveit-Moe (2013): “Heterogeneous Firms or Heteroge-
  neous Workers? Implications for Exporter Premiums and the Gains from Trade,” Review of Economics
  and Statistics, 95(3), 839–849.
Juhn, C., K. McCue, H. Monti, and B. Pierce (2018): “Firm Performance and the Volatility of
  Worker Earnings,” Journal of Labor Economics, 36(S1), S99–S131.
Kline, P., N. Petkova, H. Williams, and O. Zidar (2019): “Who Proﬁts from Patents? Rent-
  Sharing at Innovative Firms,” Quarterly Journal of Economics, 134(3), 1343–1404.
Lazear, E. (1986): “Salaries and Piece Rates,” Journal of Business, 59(3), 405–431.
          (2000): “Performance Pay and Productivity,” American Economic Review, 90(5), 1346–1361.
Macis, M., and F. Schivardi (2016): “Exports and Wages: Rent Sharing, Workforce Composition, or
  Returns to Skills?,” Journal of Labor Economics, 34(4), 945–978.
Manning, A. (2021): “Monopsony in Labor Markets: A Review,” Industrial and Labor Relations Review,
  74(1), 3–26.
Melitz, M. J. (2003): “The Impact of Trade on Intra-Industry Reallocations and Aggregate Industry
  Productivity,” Econometrica, 71(6), 1695–1725.
Munch, J. R., and J. R. Skaksen (2008): “Human Capital and Wages in Exporting Firms,” Journal
  of International Economics, 75(2), 363–372.
Portugal, P., and H. Vilares (2013): “Sobre of sindicatos, a Sindicalizacao e o Premio Salarial,”
  Banco de Portugal, Boletim Economico de Inverno.
Revenga, A. L. (1992): “Exporting Jobs? The Impact of Import Competition on Employment and
  Wages in U.S. Manufacturing,” Quarterly Journal of Economics, 1(107), 255–284.
Schank, T., C. Schnabel, and J. Wagner (2007): “Do Exporters Really Pay Higher Wages? First
  Evidence from German Linked Employer-Employee Data,” Journal of International Economics, 72(1),
  52–74.


                                                25
Song, J., D. J. Price, F. Guvenen, N. Bloom, and T. von Wachter (2019): “Firming Up
  Inequality,” Quarterly Journal of Economics, 134(1), 1–50.
Verhoogen, E. (2008): “Trade, Quality Upgrading, and Wage Inequality in the Mexican Manufacturing
  Sector,” Quarterly Journal of Economics, 123(2), 489–530.




                                               26
                               Figure 1: Weighted actual and forecast GDP growth, ﬁrm-level data
                 30       20
        percentage points
              10 0
                 -10




                               2006                   2010                    2014                   2018

Notes: Figure depicts means and ranges of weighted actual GDP growth (in gray) and forecast GDP growth (in
black) at the ﬁrm-level, using the estimation sample. Means are displayed in lines, while ranges are shown in
bars.




                                                              27
                               Table 1: Summary statistics, estimation sample, 2007-2018

     Variables                               mean            sd         min        max         mean           sd         min          max
                                                             Levels                                          Changes
     Weighted forecast error               -0.0576      0.7014 -14.3957         5.9409       0.1257     0.9272      -4.8760 14.4744
     Weighted forecast growth               0.4503      1.4551  -2.8063        25.8987      -0.3424     1.7559     -24.6233 8.0253

     log   sales                          14.9961       1.5749        7.9455   22.9203      -0.0742     0.4380      -6.4713 3.3672
     log   exports                        12.6600       2.6719        0.0180   22.1646      -0.0878     1.3947     -14.8877 11.7114
     log   (1 + fixed tangible assets)    10.3196       3.4960        0.0000   20.2001      -0.5649     3.1132     -15.6905 15.1426
     log   (1+ intangible assets)          3.2776       4.4074        0.0000   19.5609      -0.0477     4.2399     -17.1895 16.5090
     log   employment                      3.2396       1.3249        0.0000   10.1227      -0.0136     0.3178      -4.0431 3.1974
     log   value added                    13.5306       1.5294        2.7091   20.9644      -0.0676     0.5215      -9.1936 4.4144
     log   value added per worker         10.2976       0.7017        1.0996   18.4347      -0.0540     0.4551      -9.0113 4.1573
     log   avg worker pay                  9.6312       0.4456        6.8120   12.9764       0.0064     0.2032      -2.4518 3.3622

     log monthly wage                       7.0310      0.4080        6.1200   10.1300       0.0183     0.1920      -2.0650      3.1720
28




     log hourly wage                        1.8790      0.4110        0.9570    5.0310       0.0161     0.1920      -2.0900      3.1690
     log monthly base wage                  6.8110      0.3740        6.1200    9.9280       0.0139     0.1450      -1.5260      2.5270
     log (1 + overtime pay)                 0.7320      1.4420        0.0000    8.1080       0.0235     1.0080      -8.1080      7.4550
     log (1+ other pay)                     5.0070      1.2980        0.0000    9.9510       0.1200     1.1350      -8.3480      8.0050
     log total hours                        8.0160      1.4070        4.9130   14.6700      -0.0049     0.4130      -5.0790      5.8480
     share with a degree                    0.1470      0.1890        0.0000    1.0000       0.0164     0.0943      -1.0000      1.0000

     N (obs.)                                               44398                                             22199
     Notes: Table reports summary statistics on the firm-level data from the estimation sample for 2007-2018, both in levels and in
     changes. Levels refer to variables averaged over 3-year periods, changes refer to variation between 3-year periods.
                                       Table 2: Eﬀects of forecast errors and forecast growth on ﬁrm performance



                                           (1)        (2)           (3)                 (4)             (5)             (6)          (7)        (8)
     Dep. variable:                     log sales log exports log (1+ inv.         log (1+ inv.         log          log value    log value  log avg.
                                                              fixed tangible        intangible      employment         added      added per worker pay
                                                                  assets)             assets)                                      worker

     Weighted forecast error          0.0701*** 0.1189***           0.1224*          0.1224*         0.0226***      0.0574*** 0.0329*** 0.0130***
                                       (0.0098)  (0.0188)           (0.0669)         (0.0645)         (0.0053)       (0.0097)  (0.0088)  (0.0045)
     Weighted forecast growth         0.0320*** 0.0785***            0.0500           0.0585         0.0092***      0.0248*** 0.0138**   0.0048*
                                       (0.0054)  (0.0123)           (0.0454)         (0.0454)         (0.0030)       (0.0058)  (0.0056)  (0.0026)

     Period x region FE                    Y             Y              Y                Y                Y              Y             Y             Y
     Period x industry FE                  Y             Y              Y                Y                Y              Y             Y             Y
     Period x export share                 Y             Y              Y                Y                Y              Y             Y             Y
29




     Firm FE                               Y             Y              Y                Y                Y              Y             Y             Y
     N (obs.)                           22199         22199          22199            22199            22199          22199         22199         22199
     N (firms)                           9306          9306           9306             9306             9306           9306          9306          9306
     Adj. R2                            0.0719        0.0707         0.0157           0.0131           0.0535         0.0345        0.0142        0.0306
     RSS                                 1580         22750          129170           291606            858            2582          2378           465
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level.
     *10% level of significance, **5% level of significance, ***1% level of significance.
          Table 3: Eﬀects of forecast errors and forecast growth on worker compensation and worker composition



                                       (1)            (2)          (3)            (4)           (5)           (6)           (7)           (8)
     Dep. variable:                    log        log hourly       log         log (1 +      log (1+       log total    share with    person FE
                                     monthly         wage        monthly       overtime     other pay)       hours       a degree
                                      wage                      base wage        pay)

     Weighted forecast error         0.0090**      0.0085**      0.0060*      0.0328**        0.0210      0.0314***       0.0028       -0.0423**
                                     (0.0041)      (0.0041)      (0.0035)      (0.0144)      (0.0186)      (0.0066)      (0.0019)       (0.0167)
     Weighted forecast growth         0.0033        0.0030        -0.0001     0.0304***       0.0143      0.0108***       0.0002         -0.0131
                                     (0.0023)      (0.0024)      (0.0019)      (0.0104)      (0.0130)      (0.0041)      (0.0013)       (0.0105)

     Period x region FE                  Y             Y             Y             Y             Y             Y             Y             Y
     Period x industry FE                Y             Y             Y             Y             Y             Y             Y             Y
     Period x export share               Y             Y             Y             Y             Y             Y             Y             Y
     Firm FE                             Y             Y             Y             Y             Y             Y             Y             Y
30




     N (obs.)                         22199         22199         22199         22199          22199        22199         22199         12631
     N (firms)                         9306          9306          9306          9306          9306          9306          9306          6012
     Adj. R2                          0.0210        0.0229        0.0556        0.0222        0.00951       0.0601        0.0231        0.0490
     RSS                               472           471           249          15012          16665         1711         105.3          1546
     Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are clustered at firm-
     level. *10% level of significance, **5% level of significance, ***1% level of significance.
     Table 4: Eﬀects of forecast errors and forecast growth on worker compensation: high vs low earners

                                       (1)        (2)         (3)        (4)        (5)        (6)
       Dep. variable:                                       log monthly wage
       High vs. low earners            high       low        high       low        high       low
       Definition                      5%         95%        15%        85%        25%        75%

       Weighted forecast error       0.0172**   0.0073**   0.0128**   0.0063**   0.0098*     0.0051*
                                     (0.0068)   (0.0030)   (0.0059)   (0.0029)   (0.0055)    (0.0029)
       Weighted forecast growth       0.0062     0.0015     0.0037     0.0014     0.0021      0.0008
                                     (0.0043)   (0.0020)   (0.0036)   (0.0020)   (0.0033)    (0.0021)

       Period x region FE                    Y       Y          Y         Y           Y          Y
       Period x industry FE                  Y       Y          Y         Y           Y          Y
       Period x export share                 Y       Y          Y         Y           Y          Y
       Firm FE                               Y       Y          Y         Y           Y          Y
31




       N (obs.)                            20888  20888      20888     20888       20888      20888
       N (firms)                           8745    8745       8745      8745        8745       8745
       Adj. R2                           0.00868  0.0193     0.0110    0.0214      0.0127     0.0223
       RSS                                 1602    293        1106       272        908        258
       Notes: In each column, the dependent variable is the change between the averages of each 3-year
       period. Standard errors are clustered at firm-level. *10% level of significance, **5% level of
       significance, ***1% level of significance.
     Table 5: Eﬀects of forecast errors and forecast growth on worker compensation, high vs low earners, according to managerial skill
     (managers deﬁned by occupational category)


                                                                          (1)          (2)           (3)          (4)          (5)          (6)
                             Dep. variable:                                                        log monthly wage
                             High vs. low earners                        high         low           high         low          high         low
                             Definition                                  5%           95%           15%          85%          25%          75%
                             A. Firms with high-skilled managers
                             Weighted forecast error             0.0435***           0.0057      0.0359**       0.0067     0.0288**        0.0044
                                                                  (0.0159)          (0.0068)     (0.0139)      (0.0061)    (0.0133)      (0.0061)
                             Weighted forecast growth            0.0191**            0.0011      0.0130*        0.0009      0.0088        -0.0009
                                                                  (0.0083)          (0.0042)     (0.0072)      (0.0043)    (0.0066)      (0.0052)

                             Period x region FE                            Y            Y            Y             Y            Y            Y
                             Period x industry FE                          Y            Y            Y             Y            Y            Y
32




                             Period x export share                         Y            Y            Y             Y            Y            Y
                             Firm FE                                       Y            Y            Y             Y            Y            Y
                             N (obs.)                                    6226         6226         6226          6226         6226         6226
                             N (firms)                                   2739         2739         2739          2739         2739         2739
                             Adj. R2                                    0.0242       0.0366       0.0276        0.0440       0.0293       0.0433
                             RSS                                          447          80           292           71           236          66
                             B. Firms with low-skilled managers
                             Weighted forecast error                    -0.0028      -0.0013      -0.0147       -0.0044      -0.0139      -0.0041
                                                                       (0.0131)     (0.0065)     (0.0111)      (0.0065)     (0.0106)     (0.0065)
                             Weighted forecast growth                    0.0040      -0.0034      -0.0044       -0.0034      -0.0039      -0.0028
                                                                       (0.0094)     (0.0040)     (0.0080)      (0.0039)     (0.0077)     (0.0039)

                             Period x region FE                            Y            Y            Y             Y            Y            Y
                             Period x industry FE                          Y            Y            Y             Y            Y            Y
                             Period x export share                         Y            Y            Y             Y            Y            Y
                             Firm FE                                       Y            Y            Y             Y            Y            Y
                             N (obs.)                                    6478         6478         6478          6478         6478         6478
                             N (firms)                                   3015         3015         3015          3015         3015         3015
                             Adj. R2                                    0.0383       0.0376       0.0314        0.0419       0.0326       0.0466
                             RSS                                          406          72           276           66           224          63
                             Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors
                             are clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of significance.
     Table 6: Eﬀects of forecast errors and forecast growth on worker compensation, high vs low earners, according to managerial skill
     (managers deﬁned as top 5% earners)


                                                                          (1)          (2)           (3)          (4)          (5)          (6)
                             Dep. variable:                                                        log monthly wage
                             High vs. low earners                        high         low           high         low          high         low
                             Definition                                  5%           95%           15%          85%          25%          75%
                             A. Firms with high-skilled managers
                             Weighted forecast error                    0.0191       -0.0006      0.0129        -0.0011       0.0049      -0.0027
                                                                       (0.0128)     (0.0051)     (0.0107)      (0.0047)     (0.0098)     (0.0047)
                             Weighted forecast growth                   0.0084       -0.0016      0.0028        -0.0022      -0.0018      -0.0041
                                                                       (0.0076)     (0.0032)     (0.0061)      (0.0032)     (0.0054)     (0.0035)

                             Period x region FE                            Y            Y            Y             Y            Y            Y
                             Period x industry FE                          Y            Y            Y             Y            Y            Y
33




                             Period x export share                         Y            Y            Y             Y            Y            Y
                             Firm FE                                       Y            Y            Y             Y            Y            Y
                             N (obs.)                                    9299         9299         9299          9299         9299         9299
                             N (firms)                                   3682         3682         3682          3682         3682         3682
                             Adj. R2                                    0.0213       0.0374       0.0225        0.0407       0.0259       0.0415
                             RSS                                          72           123          466           113          377          106
                             B. Firms with low-skilled managers
                             Weighted forecast error                   0.0154*     0.0105***     0.0121*      0.0095**      0.0118*      0.0086**
                                                                       (0.0080)     (0.0037)     (0.0073)     (0.0038)      (0.0069)     (0.0037)
                             Weighted forecast growth                   0.0062       0.0037       0.0050       0.0038        0.0051       0.0039
                                                                       (0.0051)     (0.0026)     (0.0046)     (0.0027)      (0.0043)     (0.0027)

                             Period x region FE                            Y            Y            Y             Y            Y            Y
                             Period x industry FE                          Y            Y            Y             Y            Y            Y
                             Period x export share                         Y            Y            Y             Y            Y            Y
                             Firm FE                                       Y            Y            Y             Y            Y            Y
                             N (obs.)                                   11585        11585        11585         11585        11585        11585
                             N (firms)                                   5063         5063         5063          5063         5063         5063
                             Adj. R2                                    0.0200       0.0250       0.0219        0.0262       0.0262       0.0277
                             RSS                                          845          165          619           154          511          145
                             Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors
                             are clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of significance.
     Table 7: Eﬀects of forecast errors and forecast growth on diﬀerent wage components, high vs low earners, according to managerial
     skill (deﬁned by occupational category)
                                                                (1)           (2)               (3)          (4)            (5)          (6)
                               Dep. variable:                    log base wage                 log overtime pay              log other pay
                               High versus low earners         high          low               high         low            high         low
                               Definition                      15%           85%               15%         85%             15%          85%
                               A. Firms with high-skilled managers
                               Weighted forecast error        0.0087       -0.0026            0.0292        -0.0093       0.0927      0.0820**
                                                            (0.0120)      (0.0051)           (0.0397)      (0.0456)     (0.0638)      (0.0386)
                               Weighted forecast growth      -0.0035      -0.0059*            0.0198         0.0096      -0.0220       0.0257
                                                            (0.0058)      (0.0036)           (0.0270)      (0.0301)     (0.0514)      (0.0224)

                               Period x region FE                Y                 Y             Y            Y             Y             Y
                               Period x industry FE              Y                 Y             Y            Y             Y             Y
                               Period x export share             Y                 Y             Y            Y             Y             Y
                               Firm FE                           Y                 Y             Y            Y             Y             Y
                               N (obs.)                        6226              6226          6226         6226          6226          6226
                               N (firms)                       2739              2739          2739         2739          2739          2739
34




                               Adj. R2                        0.0629            0.0886        0.0590       0.0684        0.0271        0.0229
                               RSS                              148               38           6480         4630          6604          3038
                               B. Firms with low-skilled managers
                               Weighted forecast error       -0.0072            -0.0020    0.1208***        0.0357       -0.0407        0.0263
                                                            (0.0083)           (0.0051)     (0.0389)       (0.0360)     (0.0579)      (0.0450)
                               Weighted forecast growth      -0.0089            -0.0026    0.1016***        0.0362       -0.0006       -0.0101
                                                            (0.0060)           (0.0030)     (0.0297)       (0.0262)     (0.0476)      (0.0282)

                               Period x region FE                    Y             Y             Y            Y             Y             Y
                               Period x industry FE                  Y             Y             Y            Y             Y             Y
                               Period x export share                 Y             Y             Y            Y             Y             Y
                               Firm FE                               Y             Y             Y            Y             Y             Y
                               N (obs.)                            6478          6478          6478         6478          6478          6478
                               N (firms)                           3015          3015          3015         3015          3015          3015
                               Adj. R2                            0.0572        0.0894        0.0419       0.0357        0.0294        0.0298
                               RSS                                  191           35           5114         4063          7719          3410
                               Notes: In each column, the dependent variable is the change between the averages of each 3-year period.
                               Standard errors are clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of
                               significance.
     Table 8: Eﬀects of forecast errors and forecast growth on diﬀerent wage components, high vs low earners, according to managerial
     skill (deﬁned as top 5% earners)
                                                                (1)           (2)               (3)          (4)            (5)          (6)
                               Dep. variable:                    log base wage                 log overtime pay              log other pay
                               High versus low earners         high          low               high         low            high         low
                               Definition                      15%           85%               15%         85%             15%          85%
                               A. Firms with high-skilled managers
                               Weighted forecast error        0.0011       -0.0042         0.1185*** 0.0539*              0.0681      0.0608**
                                                            (0.0084)      (0.0038)          (0.0342)  (0.0296)          (0.0478)      (0.0298)
                               Weighted forecast growth      -0.0040      -0.0046*         0.0719*** 0.0636***           -0.0294       -0.0025
                                                            (0.0044)      (0.0025)          (0.0239)  (0.0212)          (0.0386)      (0.0177)

                               Period x region FE               Y                  Y             Y            Y             Y             Y
                               Period x industry FE             Y                  Y             Y            Y             Y             Y
                               Period x export share            Y                  Y             Y            Y             Y             Y
                               Firm FE                          Y                  Y             Y            Y             Y             Y
                               N (obs.)                       9299               9299          9299         9299          9299          9299
                               N (firms)                      3682               3682          3682         3682          3682          3682
35




                               Adj. R2                       0.0323             0.0808        0.0361       0.0404        0.0215        0.0280
                               RSS                             272                59          10771         7595         11250          5042
                               B. Firms with low-skilled managers
                               Weighted forecast error      0.0109*           0.0097***       0.0301        0.0308       -0.0175       0.0067
                                                            (0.0062)           (0.0028)      (0.0225)      (0.0217)     (0.0368)      (0.0286)
                               Weighted forecast growth      0.0012           0.0049**       0.0299*        0.0154        0.0088       0.0038
                                                            (0.0039)           (0.0021)      (0.0164)      (0.0145)     (0.0262)      (0.0181)

                               Period x region FE                    Y             Y             Y            Y             Y             Y
                               Period x industry FE                  Y             Y             Y            Y             Y             Y
                               Period x export share                 Y             Y             Y            Y             Y             Y
                               Firm FE                               Y             Y             Y            Y             Y             Y
                               N (obs.)                           11585         11585         11585        11585         11585         11585
                               N (firms)                           5063          5063          5063         5063          5063          5063
                               Adj. R2                            0.0377        0.0719        0.0265       0.0217        0.0164        0.0185
                               RSS                                  423           85           8736         6446         15292          8820
                               Notes: In each column, the dependent variable is the change between the averages of each 3-year period.
                               Standard errors are clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of
                               significance.
     Table 9: Eﬀects of forecast errors and forecast growth on worker compensation: longer-tenured vs newly hired workers

                                                 (1)                    (2)                (3)           (4)
            Dep. variable:                  log monthly             log monthly        log (1 +    log (1+ other
                                                wage                 base wage       overtime pay)      pay)
            A. Newly hired workers (less than 4 years)
            Weighted forecast error            0.0091                    0.0058          0.0234              0.0233
                                              (0.0062)                 (0.0063)         (0.0207)           (0.0252)
            Weighted forecast growth           0.0004                   -0.0019         0.0331**            -0.0140
                                              (0.0037)                 (0.0037)         (0.0148)           (0.0174)
            Period x region FE                   Y                         Y                 Y                 Y
            Period x industry FE                 Y                         Y                 Y                 Y
            Period x export share                Y                         Y                 Y                 Y
            Firm FE                              Y                         Y                 Y                 Y
            N (obs.)                          18700                     18700             18700             18700
            N (firms)                          8149                      8149              8149              8149
36




            Adj. R2                           0.0139                    0.0248            0.0277            0.0105
            RSS                                 920                      811              15993             16824
            B. Longer-tenured workers (at least 4 years)
            Weighted forecast error          0.0094**                 0.0075**          0.0430***           0.0099
                                             (0.0046)                 (0.0036)           (0.0159)          (0.0232)
            Weighted forecast growth          0.0036                   -0.0001          0.0331***           0.0110
                                             (0.0030)                 (0.0023)           (0.0113)          (0.0153)
            Period x region FE                           Y                 Y                 Y                 Y
            Period x industry FE                         Y                 Y                 Y                 Y
            Period x export share                        Y                 Y                 Y                 Y
            Firm FE                                      Y                 Y                 Y                 Y
            N (obs.)                                  21418             21418             21418             21418
            N (firms)                                  9019              9019              9019              9019
            Adj. R2                                   0.0331            0.0618            0.0228            0.0088
            RSS                                        582               306              14946             17485
            Notes: In each column, the dependent variable is the change between the averages of each 3-year
            period. Standard errors are clustered at firm-level. *10% level of significance, **5% level of significance,
            ***1% level of significance.
     Table 10: Eﬀects of forecast errors and forecast growth on worker compensation: ﬁrm-level vs more centralised wage agreements

                                                                  (1)             (2)                (3)           (4)
               Dep. variable:                                log monthly      log monthly        log (1 +    log (1+ other
                                                                 wage          base wage       overtime pay)      pay)
               A. Firm-level wage agreements
               Weighted forecast error                        0.3924***         0.2185**            -0.6267         0.6367**
                                                               (0.1029)         (0.0823)           (1.2591)         (0.3091)
               Weighted forecast growth                        0.0722*           0.0429             -0.8687          0.1116
                                                               (0.0410)         (0.0429)           (0.6234)         (0.0994)

               Period x region FE                                  Y                Y                 Y                 Y
               Period x industry FE                                Y                Y                 Y                 Y
               Period x export share                               Y                Y                 Y                 Y
               Firm FE                                             Y                Y                 Y                 Y
               N (obs.)                                           137              137               137               137
               N (firms)                                          55               55                 55               55
37




               Adj. R2                                           0.734            0.521             0.564             0.767
               RSS                                               0.476            0.385             35.030            4.906
               B. More centralized wage agreements
               Weighted forecast error                         0.0087**          0.0060*         0.0337**             0.0202
                                                               (0.0041)          (0.0035)         (0.0144)           (0.0186)
               Weighted forecast growth                         0.0032            -0.0001        0.0311***            0.0139
                                                               (0.0024)          (0.0019)         (0.0104)           (0.0130)

               Period x region FE                                  Y                 Y                 Y                Y
               Period x industry FE                                Y                 Y                 Y                Y
               Period x export share                               Y                 Y                 Y                Y
               Firm FE                                             Y                 Y                 Y                Y
               N (obs.)                                         22062             22062             22062             22062
               N (firms)                                         9256              9256              9256             9256
               Adj. R2                                          0.0209            0.0552            0.0209           0.00951
               RSS                                               469               247              14892             16628
               Notes: In each column, the dependent variable is the change between the averages of each 3-year period.
               Standard errors are clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of
               significance.
      Table 11: Eﬀects of forecast errors and forecast growth on ﬁrm performance: persistent versus transitory demand shocks

                                           (1)        (2)           (3)                 (4)             (5)             (6)          (7)            (8)
     Dep. variable:                     log sales log exports log (1+ inv.         log (1+ inv.         log          log value    log value      log avg.
                                                              fixed tangible        intangible      employment         added      added per       worker
                                                                  assets)             assets)                                      worker          pay
     Weighted forecast error
      - transitory shocks               0.0154    0.1101             -0.0924          0.0618           0.0038        0.0436**   0.0360*    0.0049
                                       (0.0153)  (0.0835)           (0.1346)         (0.1712)         (0.0120)        (0.0193)  (0.0195)  (0.0081)
      - persistent shocks             0.0880*** 0.1136***           0.1829**          0.1232         0.0290***       0.0625*** 0.0322*** 0.0159***
                                       (0.0130)  (0.0226)           (0.0902)         (0.0805)         (0.0067)        (0.0126)  (0.0112)  (0.0061)
     Weighted forecast growth
      - transitory shocks              0.0191*  0.0989**             0.0256           0.0961           0.0039          0.0198       0.0135       0.0021
                                       (0.0098)  (0.0470)           (0.0812)         (0.1034)         (0.0075)        (0.0126)     (0.0129)     (0.0054)
      - persistent shocks             0.0391*** 0.0750***            0.0722           0.0555         0.0118***       0.0269***     0.0136**     0.0060**
                                       (0.0065)  (0.0132)           (0.0533)         (0.0511)         (0.0032)        (0.0067)     (0.0063)     (0.0030)
38




     Period x region FE                    Y             Y              Y                Y                Y               Y            Y             Y
     Period x industry FE                  Y             Y              Y                Y                Y               Y            Y             Y
     Period x export share                 Y             Y              Y                Y                Y               Y            Y             Y
     Firm FE                               Y             Y              Y                Y                Y               Y            Y             Y
     N (obs.)                           22199         22199          22199            22199            22199           22199        22199         22199
     N (firms)                           9306          9306           9306             9306             9306            9306         9306          9306
     Adj. R2                            0.0734        0.0707         0.0160           0.0131           0.0538          0.0344       0.0141        0.0307
     RSS                                 1577         22748          129122           291592            858             2581         2378          465
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level.
     *10% level of significance, **5% level of significance, ***1% level of significance.
     Table 12: Eﬀects of forecast errors and forecast growth on worker compensation and worker composition: persistent versus transitory
     demand shocks

                                              (1)            (2)           (3)            (4)           (5)            (6)            (7)           (8)
         Dep. variable:                       log        log hourly        log         log (1 +      log (1+        log total     share with    person FE
                                            monthly         wage         monthly       overtime     other pay)        hours        a degree
                                             wage                       base wage        pay)
         Weighted forecast error
          - transitory shocks               0.0100         0.0107        -0.0010        0.0328         0.0516        -0.0134        0.0068       -0.0555**
                                           (0.0070)       (0.0070)      (0.0060)       (0.0370)       (0.0448)      (0.0154)       (0.0047)       (0.0245)
          - persistent shocks              0.0115**       0.0108**      0.0113**        0.0204         0.0011      0.0466***        0.0037         -0.0374
                                           (0.0054)       (0.0054)      (0.0046)       (0.0169)       (0.0209)      (0.0081)       (0.0026)       (0.0230)
         Weighted forecast growth
          - transitory shocks                -0.0041       -0.0048     -0.0099*** 0.0645***           0.0485*        -0.0013        -0.0046        -0.0114
                                            (0.0043)      (0.0042)      (0.0037)   (0.0223)           (0.0279)      (0.0107)       (0.0030)       (0.0155)
          - persistent shocks               0.0048*         0.0044       0.0025   0.0234**             0.0048      0.0168***         0.0009        -0.0130
39




                                            (0.0028)      (0.0028)      (0.0023)   (0.0111)           (0.0136)      (0.0042)       (0.0014)       (0.0126)
         Period x region FE                      Y             Y            Y          Y                  Y              Y              Y            Y
         Period x industry FE                    Y             Y            Y          Y                  Y              Y              Y            Y
         Period x export share                   Y             Y            Y          Y                  Y              Y              Y            Y
         Firm FE                                 Y             Y            Y          Y                  Y              Y              Y            Y
         N (obs.)                             22199         22199         22199     22199              22199          22199          22199        12631
         N (firms)                             9306          9306         9306       9306               9306           9306           9306         6012
         Adj. R2                              0.0213        0.0233       0.0565     0.0224            0.00960         0.0610         0.0243       0.0489
         RSS                                  472.1         471.2         248.3     15008              16662           1709          105.2         1546
         Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are clustered at firm-level.
         *10% level of significance, **5% level of significance, ***1% level of significance.
        Table 13: Eﬀects of forecast errors and forecast growth on ﬁrm performance: positive versus negative demand shocks

     Dep. variable:                    log sales   log exports    log (1+ inv.    log (1+ inv.         log          log value     log value  log avg.
                                                                 fixed tangible    intangible      employment         added       added per worker pay
                                                                     assets)         assets)                                       worker
     Weighted forecast error
      - positive shocks                 0.0250        0.1333         0.1059          0.0257           0.0061        0.0516**       0.0427*        0.0040
                                       (0.0217)      (0.1216)       (0.1546)        (0.1784)         (0.0142)        (0.0217)      (0.0225)      (0.0082)
      - negative shocks               0.0826***     0.1077***        0.1128         0.1758**        0.0273***       0.0580***     0.0291***     0.0155***
                                       (0.0126)      (0.0301)       (0.0858)        (0.0835)         (0.0067)        (0.0121)      (0.0111)      (0.0058)
     Weighted forecast growth
      - positive shocks               0.0321***     0.1030***        0.0977           -0.0308        0.0090*        0.0278***     0.0174**        0.0046
                                       (0.0076)      (0.0273)       (0.0654)         (0.0830)        (0.0053)        (0.0087)     (0.0087)       (0.0036)
      - negative shocks               0.0379***     0.0709***        0.0409          0.0921*        0.0114***       0.0248***     0.0117*        0.0060*
                                       (0.0068)      (0.0209)       (0.0554)         (0.0552)        (0.0037)        (0.0070)     (0.0069)       (0.0032)

     Period x region FE                    Y             Y             Y                Y                Y               Y             Y             Y
     Period x industry FE                  Y             Y             Y                Y                Y               Y             Y             Y
40




     Period x export share                 Y             Y             Y                Y                Y               Y             Y             Y
     Firm FE                               Y             Y             Y                Y                Y               Y             Y             Y
     N (obs.)                           22199         22199         22199            22199            22199           22199         22199         22199
     N (firms)                           9306          9306          9306             9306             9306            9306          9306          9306
     Adj. R2                            0.0729        0.0708        0.0157           0.0132           0.0537          0.0344        0.0141        0.0307
     RSS                                 1578         22747         129156           291569            858             2581          2378          465
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level.
     *10% level of significance, **5% level of significance, ***1% level of significance.
     Table 14: Eﬀects of forecast errors and forecast growth on worker compensation and worker composition: positive versus negative
     demand shocks

                                           (1)         (2)        (3)         (4)         (5)        (6)         (7)         (8)
         Dep. variable:                    log         log        log      log (1 +    log (1+    log total    share       person
                                         monthly      hourly    monthly    overtime      other      hours      with a        FE
                                          wage        wage       base        pay)        pay)                  degree
                                                                 wage
         Weighted forecast error
          - positive shocks               -0.0013 -0.0014   0.0073   0.0338   -0.0528  0.0138 0.0113** -0.0244
                                         (0.0097) (0.0098) (0.0061) (0.0394) (0.0455) (0.0153) (0.0052) (0.0240)
          - negative shocks              0.0119** 0.0115** 0.0059    0.0233  0.0434* 0.0371*** 0.0007 -0.0530**
                                         (0.0053) (0.0053) (0.0045) (0.0179) (0.0244) (0.0086) (0.0027) (0.0223)
         Weighted forecast growth
          - positive shocks              0.0031   0.0020   -0.0009 0.0614*** 0.0077   0.0081   -0.0006                     -0.0016
                                         (0.0035) (0.0035) (0.0025) (0.0211) (0.0214) (0.0063) (0.0018)                   (0.0129)
          - negative shocks               0.0047   0.0045   -0.0001 0.0230*   0.0255 0.0137*** -0.0007                    -0.0204*
41




                                         (0.0030) (0.0030) (0.0024) (0.0120) (0.0157) (0.0051) (0.0016)                   (0.0123)

         Period x region FE                  Y           Y          Y           Y         Y            Y          Y           Y
         Period x industry FE                Y           Y          Y           Y         Y            Y          Y           Y
         Period x export share               Y           Y          Y           Y         Y            Y          Y           Y
         Firm FE                             Y           Y          Y           Y         Y            Y          Y           Y
         N (obs.)                         22199       22199      22199       22199      22199       22199      22199       12631
         N (firms)                         9306        9306       9306        9306       9306        9306       9306        6012
         Adj. R2                          0.0211      0.0230     0.0555      0.0225    0.00970      0.0601     0.0237      0.0491
         RSS                              472.2       471.4      248.5       15007      16660        1711      105.3        1545
         Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are
         clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of significance.
     A.1     Appendix Tables


                                          Table A1: Summary statistics, estimation sample, by 3-year period

      Period                                          1 [2007-2009]                          2 [2010-2012]                Change = Period 2-Period 1
      Variables                             mean      sd       min        max      mean       sd      min        max       mean     sd       min        max

      Weighted forecast error              -0.3020   1.4650   -14.4000    2.0080    0.0091 0.2960     -4.0890    3.7550    0.3120   1.4220 -4.8760     14.4700
      Weighted forecast growth              1.3210   2.7770    -0.0472   25.9000    0.0802 0.5990     -2.4560    6.6090   -1.2410   2.4300 -24.6200     1.2220

      log   sales                          14.7600   1.5290    9.5800    22.5800   14.6700   1.5990    7.7600   22.7100   -0.0855   0.4570 -6.4190      3.0240
      log   exports                        12.2400   2.6250    1.4820    21.1100   12.2500   2.6930   -0.0954   21.6800    0.0079   1.5010 -11.1900    11.7100
      log   (1 + fixed tangible assets)    10.6700   2.9090    0.0000    19.5900    9.7280   3.6330    0.0000   20.0100   -0.9450   3.0800 -15.4400    14.1300
      log   (1+ intangible assets)          3.0630   4.3820    0.0000    19.2000    2.9920   4.1930    0.0000   19.3600   -0.0709   4.5440 -16.9900    14.6000
      log   employment                      3.1970   1.3110    0.0000     9.8830    3.1540   1.3190    0.0000    9.8710   -0.0428   0.3240 -3.6020      3.1970
      log   value added                    13.2900   1.4850    7.2170    20.6800   13.2000   1.5480    4.9890   20.7100   -0.0926   0.5200 -5.7110      4.1800
42




      log   value added per worker         10.1000   0.6900    5.4250    18.2000   10.0500   0.7340    3.8900   17.6300   -0.0500   0.4660 -4.5160      3.9600
      log   avg worker pay                  9.6130   0.4640    7.2320    12.5900    9.6250   0.4550    6.8120   12.9800    0.0114   0.2190 -2.3320      2.6010

      log monthly wage                      7.0070   0.4310    6.1200     9.3610    7.0120   0.4090   6.2110 10.0300       0.0041   0.2190   -2.0650    2.2160
      log hourly wage                       1.8590   0.4340    0.9570     4.2860    1.8590   0.4130   1.0580 4.9270       -0.0002   0.2190   -2.0900    2.2160
      log monthly base wage                 6.7910   0.3930    6.1200     8.7350    6.8080   0.3790   6.1890 9.9280        0.0167   0.1660   -1.4970    2.5270
      log (1 + overtime pay)                0.7040   1.4570    0.0000     6.9490    0.6460   1.3770   0.0000 7.4550       -0.0581   1.0650   -6.0660    7.4550
      log (1+ other pay)                    4.7670   1.6530    0.0000     9.2720    4.9070   1.2970   0.0000 9.0240        0.1400   1.4410   -8.3480    8.0050
      log total hours                       7.9270   1.3880    4.9130    14.4500    7.9410   1.4050   4.9900 14.3900       0.0134   0.4670   -5.0790    5.8480
      share with a degree                   0.1230   0.1790    0.0000     1.0000    0.1380   0.1870   0.0000 1.0000        0.0149   0.1010   -1.0000    1.0000

      N (obs.)                                           8540                                      8540                                  8540
      Notes: Table reports summary statistics on the firm-level data from the estimation sample for 2007-2009 and 2010-2012, both in levels and in changes.
      Levels refer to variables averaged over 3-year periods,, changes refer to variation between 3-year periods.
                                   Table A1: Summary statistics, estimation sample, by 3-year period (cont.)

     Period                                         2 [2010-2012]                          3 [2013-2015]                Change = Period 3-Period 2
     Variables                           mean       sd       min       max       mean       sd      min        max       mean     sd       min        max

     Weighted forecast error              0.0122   0.2880   -4.0890    2.9260    -0.0342 0.1990     -2.9710    1.8380   -0.0464   0.2710   -4.8010    3.6580
     Weighted forecast growth             0.0856   0.5980   -2.4560    6.6090     0.2950 0.8830     -2.5010    7.8300    0.2090   0.5020   -1.7510    8.0250

     log   sales                         14.8500   1.5450   10.0100   22.7100   14.8300    1.6000   9.3500    22.7900   -0.0140   0.4320 -4.2830      3.3480
     log   exports                       12.5500   2.6210   -0.0954   21.6800   12.6100    2.6650   2.2330    22.0300    0.0677   1.3110 -11.8600    11.6600
     log   (1 + fixed tangible assets)   10.2200   3.2060    0.0000   20.0100    9.8400    3.7510   0.0000    19.4400   -0.3790   3.0750 -14.8500    14.0800
     log   (1+ intangible assets)         3.2670   4.2940    0.0000   19.3600    3.1830    4.2890   0.0000    19.4200   -0.0843   3.8950 -16.6600    15.6600
     log   employment                     3.2560   1.3060    0.0000    9.8710    3.2220    1.3360   0.0000     9.9950   -0.0341   0.3280 -4.0430      2.2580
     log   value added                   13.3700   1.4820    8.7560   20.7100   13.3500    1.5830   2.5690    20.7700   -0.0228   0.5550 -9.1430      4.4620
     log   value added per worker        10.1200   0.6660    6.4220   17.6300   10.1400    0.7440   0.9600    17.1000    0.0109   0.4770 -8.9610      4.2060
     log   avg worker pay                 9.6470   0.4400    7.3630   12.9800    9.6210    0.4440   7.3900    12.7000   -0.0263   0.1950 -1.7050      3.3620
43




     log monthly wage                     7.0250   0.4010    6.2190   10.0300     7.0380   0.4070   6.2160 10.0800       0.0133   0.1720   -1.4040    1.6640
     log hourly wage                      1.8720   0.4040    1.0650    4.9270     1.8850   0.4100   1.0620 4.9500        0.0130   0.1720   -1.4040    1.6600
     log monthly base wage                6.8180   0.3750    6.2190    9.9280     6.8050   0.3730   6.2160 9.3350       -0.0131   0.1270   -1.1400    1.8780
     log (1 + overtime pay)               0.7030   1.4200    0.0000    7.4550     0.6960   1.3820   0.0000 8.1080       -0.0074   0.9650   -5.9080    5.3940
     log (1+ other pay)                   4.9740   1.1960    0.0000    8.5260     5.1390   1.1600   0.0000 9.8650        0.1650   0.9510   -6.1760    6.5540
     log total hours                      8.0550   1.3870    5.0240   14.3900     8.0080   1.4240   5.0240 14.5300      -0.0475   0.4010   -4.8830    2.9790
     share with a degree                  0.1420   0.1860    0.0000    1.0000     0.1590   0.1960   0.0000 1.0000        0.0167   0.0914   -1.0000    1.0000

     N (obs.)                                           7192                                      7192                                  7192
     Notes: Table reports summary statistics on the firm-level data from the estimation sample for 2010-2012 and 2013-2015, both in levels and in changes.
     Levels refer to variables averaged over 3-year periods,, changes refer to variation between 3-year periods.
                                   Table A1: Summary statistics, estimation sample, by 3-year period (cont.)

     Period                                         3 [2013-2015]                          4 [2016-2018]                Change = Period 4-Period 3
     Variables                           mean       sd       min       max       mean       sd      min        max       mean     sd       min        max

     Weighted forecast error             -0.0278   0.1930   -2.9710    1.8380     0.0438 0.3900     -4.3230    5.9410    0.0716   0.3410   -3.5240    5.2380
     Weighted forecast growth             0.2930   0.8560   -2.5010    7.8300     0.5240 0.7050     -2.8060    5.3000    0.2310   0.8780   -5.8550    3.6170

     log   sales                         14.9600   1.5520   10.5100   22.7900   15.0400    1.6210    9.4850   22.6300    0.0811 0.4130 -5.4650        3.4850
     log   exports                       12.8500   2.5660    2.2260   22.0300   12.6600    2.8330   -0.0180   21.8600   -0.1800 1.3190 -14.7800      10.8700
     log   (1 + fixed tangible assets)   3.30200   1.3210    0.0000    9.9950    3.3500    1.3530    0.0000   10.1200   -0.0515 3.0070 -13.3400      15.1200
     log   (1+ intangible assets)        10.3200   3.3110    0.0000   19.4400    10.270    3.8890    0.0000   19.6600    0.1070 3.9120 -15.6600      16.4700
     log   employment                     3.4100   4.3440    0.0000   17.9100    3.5160    4.5290    0.0000   18.5900    0.0478 0.2890 -3.3040        2.7730
     log   value added                   13.4900   1.5000    7.8800   20.7700   13.6200    1.5700    7.0810   20.9300    0.1230 0.4730 -4.9180        3.2530
     log   value added per worker        10.2000   0.6630    6.2700   17.1000   10.2700    0.6990    5.0820   18.1100    0.0751 0.4140 -4.6490        3.2410
     log   avg worker pay                 9.6260   0.4320    6.9720   11.9200    9.6630    0.4270    7.2660   12.0800    0.0362 0.1850 -2.4520        2.3920
44




     log monthly wage                     7.0390   0.3990    6.2160   10.0800     7.0820   0.3880   6.2960 10.1300       0.0425   0.1700   -1.5260    3.1720
     log hourly wage                      1.8850   0.4020    1.0620    4.9500     1.9260   0.3910   1.1430 5.0310        0.0409   0.1700   -1.5260    3.1690
     log monthly base wage                6.8040   0.3640    6.1940    9.3350     6.8450   0.3520   6.2960 9.0150        0.0404   0.1260   -1.5260    1.1890
     log (1 + overtime pay)               0.7600   1.4300    0.0000    8.1080     0.9260   1.5840   0.0000 6.7380        0.1660   0.9600   -8.1080    5.9510
     log (1+ other pay)                   5.1570   1.1360    0.0000    9.8650     5.2000   1.0920   0.0000 9.9510        0.0431   0.8160   -6.7460    6.7280
     log total hours                      8.0980   1.4010    5.0240   14.5300     8.1170   1.4280   5.0240 14.6700       0.0182   0.3420   -3.4450    2.6390
     share with a degree                  0.1580   0.1890    0.0000    1.0000     0.1760   0.1980   0.0000 1.0000        0.0181   0.0888   -1.0000    1.0000

     N (obs.)                                           6467                                      6467                                  6467
     Notes: Table reports summary statistics on the firm-level data from the estimation sample for 2013-2015 and 2016-2018, both in levels and in changes.
     Levels refer to variables averaged over 3-year periods,, changes refer to variation between 3-year periods.
     Table A2: Summary statistics on ﬁrm age, location and corporate structure, yearly data

                   Regions - Nuts 2   Variables               Mean     Min   Max    N (obs.)   N (firms)


                   Norte              firm age                 25.0      0   262     46124        4001
                                      # of plants abroad      0.004      0    22     46124        4001
                                      % of foreign capital       7.3     0   100     46124        4001
                                                                                       43%         43%
                   Algarve            firm age                 22.9      0    74       1256         115
                                      # of plants abroad          0      0     0       1256         115
                                      % of foreign capital      6.1      0   100       1256         115
                                                                                        1%          1%
                   Centro             firm age                  25.6     0   166     24988        2166
                                      # of plants abroad     0.0001      0     3     24988        2166
                                      % of foreign capital       8.6     0   100     24988        2166
                                                                                       23%         23%
                   Lisboa             firm age                 26.7      0   235     30324        2771
45




                                      # of plants abroad       0.01      0    28     30324        2771
                                      % of foreign capital     18.3      0   100     30324        2771
                                                                                       28%         30%
                   Alentejo           firm age                 23.4      0   102       4902         456
                                      # of plants abroad        0.0      0     0       4902         456
                                      % of foreign capital     13.2      0   100       4902         456
                                                                                        5%          5%
                   Azores             firm age                 31.5      0   105        354          30
                                      # of plants abroad          0      0     0        354          30
                                      % of foreign capital      4.1      0   100        354          30
                                                                                      0.3%        0.3%
                   Madeira            firm age                 43.9      4    98        480          43
                                      # of plants abroad       0.02      0     4        480          43
                                      % of foreign capital     10.2      0   100        480          43
                                                                                      0.4%        0.5%
                   Total              firm age                 25.6      0   262    108428        9306
                                      # of plants abroad      0.005      0    28    108428        9306
                                      % of foreign capital     10.9      0   100    108428        9306
                                                                                      100%        100%
Table A3: Main export markets, ranked according to export shares in 2006
                                                       Export share
                                      Export
                                                                Estimation
                                       rank     All exports
                                                                sample
 Spain                                     1     0.2818         0.2741
 Germany                                   2     0.1307         0.1384
 France                                    3     0.1307         0.1322
 United Kingdom                            4     0.0676         0.0650
 United States                             5     0.0633         0.0681
 Netherlands                               6     0.0356         0.0344
 Angola                                    7     0.0356         0.0315
 Italy                                     8     0.0346         0.0350
 Belgium                                   9     0.0303         0.0303
 Singapore                                10     0.0215         0.0243
 Sweden                                   11     0.0114         0.0117
 Switzerland                              12     0.0079         0.0075
 Brazil                                   13     0.0075         0.0076
 Finland                                  14     0.0074         0.0078
 Denmark                                  15     0.0069         0.0067
 China                                    16     0.0063         0.0070
 Poland                                   17     0.0060         0.0064
 Turkey                                   18     0.0056         0.0060
 Cape Verde                               19     0.0054         0.0050
 Austria                                  20     0.0054         0.0054
 Ireland                                  21     0.0051         0.0050
 Morocco                                  22     0.0048         0.0046
 Canada                                   23     0.0046         0.0046
 Mexico                                   24     0.0040         0.0042
 Czech Republic                           25     0.0037         0.0038
 Greece                                   26     0.0036         0.0035
 Hungary                                  27     0.0034         0.0036
 Norway                                   28     0.0033         0.0030
 Japan                                    29     0.0032         0.0034
 Russian Federation                       30     0.0029         0.0029
 Malaysia                                 31     0.0026         0.0029
 Hong Kong                                32     0.0024         0.0024
 South Africa                             33     0.0022         0.0023
 Romania                                  34     0.0022         0.0022
 Israel                                   35     0.0021         0.0021
 Mozambique                               36     0.0021         0.0020
 Australia                                37     0.0020         0.0021
 Algeria                                  38     0.0019         0.0019
 Chile                                    39     0.0016         0.0018
 Tunisia                                  40     0.0016         0.0016
 Notes: Table reports the share of exports to each of the top 40
 destinations in 2006, both in the full customs data and in the estimation
 sample.




                                       46
                Table A4: Eﬀects of forecast errors and forecast growth on average export price and quantity



                                           (1)        (2)              (3)               (4)
     Dep. variable:                     log sales log exports       log avg.          log avg.
                                                                     export         export price
                                                                    quantity

     Weighted forecast error          0.0701*** 0.1189***           0.1220***          -0.0033
                                       (0.0098)  (0.0188)            (0.0224)         (0.0118)
     Weighted forecast growth         0.0320*** 0.0785***           0.0799***          -0.0017
                                       (0.0054)  (0.0123)            (0.0133)         (0.0081)

     Period x region FE                    Y             Y               Y                Y
     Period x industry FE                  Y             Y               Y                Y
     Period x export share                 Y             Y               Y                Y
47




     Firm FE                               Y             Y               Y                Y
     N (obs.)                           22199         22199           22194            22194
     N (firms)                           9306          9306            9305             9305
     Adj. R2                            0.0719        0.0707          0.0524           0.0307
     RSS                                 1580         22750           30458             7803
     Notes: In each column, the dependent variable is the change between the average of each 3-
     year period. Standard errors are clustered at the firm-level. *10% level of significance, **5%
     level of significance, ***1% level of significance.
                                   Table A5: Eﬀects of actual GPD growth on ﬁrm performance

                                       (1)        (2)           (3)                (4)             (5)             (6)          (7)           (8)
     Dep. variable:                 log sales log exports log (1+ inv.        log (1+ inv.         log          log value    log value     log avg.
                                                          fixed tangible       intangible      employment         added      added per      worker
                                                              assets)            assets)                                      worker         pay


     Actual GDP growth            0.0274*** 0.0737***            0.0412           0.0507        0.0076***      0.0208***     0.0115**       0.0038
                                   (0.0051)  (0.0128)           (0.0449)         (0.0448)        (0.0029)       (0.0057)     (0.0057)      (0.0025)

     Period x region FE                Y             Y             Y                Y                Y              Y             Y            Y
     Period x industry FE              Y             Y             Y                Y                Y              Y             Y            Y
     Period x export share             Y             Y             Y                Y                Y              Y             Y            Y
     Firm FE                           Y             Y             Y                Y                Y              Y             Y            Y
     N (obs.)                       22199         22199         22199            22199            22199          22199         22199        22199
     N (firms)                       9306          9306          9306             9306             9306           9306          9306         9306
48




     Adj. R2                        0.0642        0.0702        0.0154           0.0130           0.0517         0.0309        0.0129       0.0294
     RSS                             1593         22765         129217           291643            860            2591          2381         465
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-
     level. *10% level of significance, **5% level of significance, ***1% level of significance.
                 Table A6: Eﬀects of actual GDP growth on worker compensation and worker composition
                                          (1)        (2)        (3)         (4)         (5)       (6)         (7)        (8)
     Dep. variable:                       log        log        log      log (1 +    log (1+   log total    share      person
                                        monthly     hourly    monthly    overtime      other     hours      with a       FE
                                         wage       wage       base        pay)        pay)                 degree
                                                               wage

     Actual GDP growth                   0.0027   0.0024   -0.0009 0.0302*** 0.0140 0.0083** -0.0001 -0.0140
                                        (0.0023) (0.0023) (0.0018) (0.0103) (0.0136) (0.0040) (0.0013) (0.0106)

     Period x region FE                    Y           Y          Y          Y          Y          Y           Y          Y
     Period x industry FE                  Y           Y          Y          Y          Y          Y           Y          Y
     Period x export share                 Y           Y          Y          Y          Y          Y           Y          Y
     Firm FE                               Y           Y          Y          Y          Y          Y           Y          Y
     N (obs.)                           22199       22199      22199      22199       22199     22199       22199      12631
     N (firms)                           9306        9306       9306       9306        9306      9306        9306       6012
     Adj. R2                            0.0205      0.0224     0.0544     0.0223     0.00954    0.0580      0.0225     0.0481
49




     RSS                                472.6       471.7      248.9      15012       16665      1715       105.4       1547
     Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are
     clustered at firm-level. *10% level of significance, **5% level of significance, ***1% level of significance.
                                           Table A7: Summary statistics, ﬁrms in the management survey data, 2016

                                                                     Firms with Z-score below the median                    Firms with Z-score above the median
                                                                     mean       sd         min      max                     mean       sd        min       max
     log sales                                                       16.6200       1.4080       12.6600     21.5200        17.3600        1.4470       13.5600     22.4700
     Foreign capital                                                  0.2050       0.4040        0.0000      1.0000         0.3320        0.4710        0.0000      1.0000
     Public capital                                                   0.0423       0.2010        0.0000      1.0000         0.0174        0.1310        0.0000      1.0000
     Export status                                                    0.6980       0.4600        0.0000      1.0000         0.7080        0.4550        0.0000      1.0000
     Exports to sales ratio                                           0.2940       0.3690        0.0000      1.0000         0.2910        0.3650        0.0000      1.0000
     % of female employees                                            0.3700       0.2690        0.0000      0.9960         0.3930        0.2360        0.0000      0.9860
     Firm age                                                        25.5900      27.2800        0.0000         516        26.7100       20.0000        0.0000         126
     % of employees with a degree                                     0.1290       0.1600        0.0000      1.0000         0.1990        0.2020        0.0000      0.9100
     Mean employee ability                                            0.1070       0.2530       -0.6400      1.2520         0.2080        0.2780       -0.4710      1.5570
     Managers defined as top 5% earners
       - Mean managerial ability                                      0.8560       0.4400       -0.2680      4.8400          1.0460       0.4180       -0.1780      4.5260
       - % of managers with a degree                                  0.4520       0.3150        0.0000      1.0000          0.5440       0.2850        0.0000      1.0000
50




     log employment                                                   5.1190       1.0650        1.0990      9.1120          5.4650       1.0920        1.9460     10.0600

     % of employees for which FE were computed ˆ3                     0.4630       0.2400        0.0004      1.0000          0.5320       0.2220        0.0002      1.0000
     Standardized management Z-score                                 -0.7920       0.5760       -3.3850     -0.0539          0.7910       0.6450       -0.0526      3.0360
     N (firms)                                                                            804                                                    805

     Managers defined by occupation
       - Mean managerial ability                                      1.0140       0.4660       -0.5190      2.8980          1.1380       0.4060       -0.5850     2.7700  
       - % of managers with a degree                                  0.6040       0.3510        0.0000      1.0000          0.6520       0.3110        0.0000      1.0000
     N (firms)                                                                            668                                                    736
     Notes: Employee ability is the mean level of individual fixed effect measured over the period 2010-2015. Managerial ability is the mean employee ability for managers in
     2010-2015.
                Table A8: Summary statistics, ﬁrms with high versus low skill managers (managers deﬁned by occupational category)

                                               Firms with low-skilled managers          Firms with high-skilled managers
                                             mean       sd         min      max       mean       sd         min      max
     log sales                               15.2392    1.3766     7.9455   21.5164    15.9567   1.6213     10.3172   22.9203
     log exports                             12.9804    2.5652     0.0954   19.5555    13.5086   2.8389      0.0180   22.1646
     log( 1+ inv. fixed tangible assets)     10.6522    3.2207     0.0000   19.4539    11.5118   3.1866      0.0000   20.2001
     log( 1+ inv. fixed intangible assets)    3.5559    4.3950     0.0000   18.6151     4.8535   4.9971      0.0000   19.5609
     log employment                           3.5067    1.1391     0.0000    9.4395     4.0070   1.3435      0.5108   10.1227
     log value added                         13.7798    1.2767     6.4979   20.2622    14.4962   1.5512      9.3873   20.9644
     log value added per worker              10.2770    0.6418     5.3993   14.4470    10.4961   0.7127      5.8780   13.5628
     log avg. worker pay                      9.6330    0.3999     6.8120   12.7015     9.8288   0.4432      8.2351   11.9219

     log monthly wage                         7.0557    0.3564     6.2554   10.0248     7.2477   0.4170      6.2681   10.1296
     log hourly wage                          1.9011    0.3593     1.1021    4.8688     2.0944   0.4232      1.1148    5.0310
     log monthly base wage                    6.8320    0.3295     6.2235    8.3699     7.0037   0.3898      6.2419    9.3352
     log(1+overtime pay)                      0.7378    1.4194     0.0000    6.6701     1.2556   1.6925      0.0000    6.7379
51




     log(1+ other pay)                        5.1262    1.0779     0.0000    9.9509     5.4021   1.0331      0.0000    9.8650
     log (total hours)                        8.3564    1.1483     5.3967   14.3386     8.8685   1.3667      5.4410   14.6728
     share with a degree                      0.1204    0.1369     0.0000    1.0000     0.2536   0.2031      0.0000    1.0000
     N (obs.)                                               6478                                     6226
                           Table A9: Summary statistics, ﬁrms with high versus low skill managers (deﬁned as top 5% earners)

                                                 Firms with low-skilled managers           Firms with high-skilled managers
                                               mean       sd         min      max        mean       sd         min      max
     log sales                                 14.4641    1.2188     7.9455   21.7779    15.8413    1.5780     10.3172   22.9203
     log exports                               12.1594    2.5059     0.0180   18.5216    13.4258    2.7964      1.7021   22.1646
     log( 1+ inv. fixed tangible assets)        9.4902    3.5715     0.0000   17.8895    11.2998    3.3141      0.0000   20.2001
     log( 1+ inv. fixed intangible assets)      2.4499    3.7693     0.0000   14.3087     4.6124    4.9016      0.0000   19.5609
     log employment                             2.8634    0.9711     0.0000    7.0809     3.9677    1.3144      0.0000   10.1227
     log value added                           13.0230    1.1438     2.7091   18.1752    14.3942    1.5083      7.9996   20.9644
     log value added per worker                10.1659    0.6637     1.0996   13.4763    10.4322    0.7207      5.8780   14.7964
     log avg. worker pay                        9.5459    0.3912     7.2659   11.9219     9.7800    0.4437      6.8120   12.7015

     log monthly wage                           6.9459    0.3410     6.2406    9.5171     7.1940    0.4126      6.2681   10.1296
     log hourly wage                            1.7913    0.3421     1.0873    4.4107     2.0406    0.4182      1.1148    5.0310
     log monthly base wage                      6.7299    0.3104     6.2158    9.3352     6.9534    0.3830      6.2158    9.0152
     log (1+overtime pay)                       0.4867    1.2264     0.0000    7.0398     1.1605    1.6626      0.0000    6.3360
52




     log (1+ other pay)                         4.9664    1.1485     0.0000    8.9536     5.3330    1.0432      0.0000    9.9509
     log (total hours)                          7.6359    1.0209     5.3471   11.8676     8.8101    1.3519      5.4410   14.6728
     share with a degree                        0.0996    0.1405     0.0000    1.0000     0.2259    0.2022      0.0000    1.0000

     N (obs.)                                                11585                                      9299
        Table A10: Eﬀects of forecast errors and forecast growth on ﬁrm performance, according to managerial skill (managers deﬁned by
        occupational category)
                                                     (1)        (2)           (3)                 (4)             (5)              (6)          (7)        (8)
     Dep. variable:                               log sales log exports log (1+ inv.         log (1+ inv.         log           log value    log value  log avg.
                                                                        fixed tangible        intangible      employment          added      added per worker pay
                                                                            assets)             assets)                                       worker
     A. Firms with high-skilled managers
     Weighted forecast error             0.0525*** 0.1222***                   0.0946           0.0681          0.0226*        0.0621***     0.0340**       -0.0027
                                          (0.0183)  (0.0444)                  (0.1344)         (0.2065)         (0.0132)        (0.0202)     (0.0149)      (0.0103)
     Weighted forecast growth            0.0247** 0.0893***                    0.0547           0.0654          0.0163**       0.0322***      0.0140        -0.0056
                                          (0.0102)  (0.0215)                  (0.0990)         (0.1364)         (0.0068)        (0.0117)     (0.0099)      (0.0048)
     Period x region FE                              Y             Y              Y                Y                 Y              Y             Y            Y
     Period x industry FE                            Y             Y              Y                Y                 Y              Y             Y            Y
     Period x export share                           Y             Y              Y                Y                 Y              Y             Y            Y
     Firm FEs                                        Y             Y              Y                Y                 Y              Y             Y            Y
     N (obs.)                                      6226          6226           6226             6226              6226           6226          6226         6226
53




     N (firms)                                     2739          2739           2739             2739              2739           2739          2739         2739
     Adj. R2                                       0.137         0.114         0.0416           0.0427            0.0989         0.0551        0.0360       0.0340
     RSS                                            337          6584          21270            94029              180             582           506          88
     B. Firms with low-skilled managers
     Weighted forecast error                    0.0842*** 0.1236***           0.2620**          -0.0282        0.0386***       0.0692***     0.0286*         0.0008
                                                 (0.0176)  (0.0259)           (0.1028)         (0.1443)         (0.0097)        (0.0187)     (0.0173)      (0.0101)
     Weighted forecast growth                   0.0380*** 0.0841***            0.0415           -0.0532        0.0132**        0.0318***     0.0174**       -0.0020
                                                 (0.0096)  (0.0183)           (0.0703)         (0.1007)         (0.0062)        (0.0098)     (0.0080)      (0.0047)
     Period x region FE                              Y             Y              Y                Y                 Y              Y             Y            Y
     Period x industry FE                            Y             Y              Y                Y                 Y              Y             Y            Y
     Period x export share                           Y             Y              Y                Y                 Y              Y             Y            Y
     Firm Fes                                        Y             Y              Y                Y                 Y              Y             Y            Y
     N (obs.)                                      6478          6478           6478             6478              6478           6478          6478         6478
     N (firms)                                     3015          3015           3015             3015              3015           3015          3015         3015
     Adj. R2                                      0.0834         0.104         0.0541           0.0333            0.0657         0.0631        0.0428       0.0529
     RSS                                           299           4978          24619            83059              154             453           394          87
     Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are clustered at firm-level. *10% level of
     significance, **5% level of significance, ***1% level of significance.
        Table A11: Eﬀects of forecast errors and forecast growth on ﬁrm performance, according to managerial skill (managers deﬁned as top
        5% earners)
                                                     (1)        (2)           (3)                 (4)             (5)              (6)          (7)        (8)
     Dep. variable:                               log sales log exports log (1+ inv.         log (1+ inv.         log           log value    log value  log avg.
                                                                        fixed tangible        intangible      employment          added      added per worker pay
                                                                            assets)             assets)                                       worker
     A. Firms with high-skilled managers
     Weighted forecast error             0.0431*** 0.1114***                  0.1502*           0.0821          0.0180**       0.0541*** 0.0311***          -0.0010
                                          (0.0134)  (0.0308)                  (0.0901)         (0.1372)         (0.0090)        (0.0146)  (0.0120)         (0.0088)
     Weighted forecast growth            0.0186** 0.0840***                   0.1046*           0.0485          0.0127**       0.0250***   0.0095           -0.0022
                                          (0.0081)  (0.0181)                  (0.0631)         (0.0957)         (0.0050)        (0.0092)  (0.0080)         (0.0038)

     Period x region FE                              Y             Y              Y                Y                 Y              Y             Y            Y
     Period x industry FE                            Y             Y              Y                Y                 Y              Y             Y            Y
     Period x export share                           Y             Y              Y                Y                 Y              Y             Y            Y
     Firm FE                                         Y             Y              Y                Y                 Y              Y             Y            Y
54




     N (obs.)                                      9299          9299           9299             9299              9299           9299          9299         9299
     N (firms)                                     3682          3682           3682             3682              3682           3682          3682         3682
     Adj. R2                                       0.131        0.0746         0.0369           0.0245            0.0955         0.0697        0.0377       0.0259
     RSS                                            593         10155          36528            138899             324            1002           888          169
     B. Firms with low-skilled managers
     Weighted forecast error                    0.0746*** 0.1326***           0.1357*          0.1380**        0.0280***       0.0568*** 0.0259**           0.0023
                                                 (0.0166)  (0.0277)           (0.0777)         (0.0688)         (0.0067)        (0.0136)  (0.0123)         (0.0048)
     Weighted forecast growth                   0.0581*** 0.1236***           0.0907*          0.0946**        0.0166***       0.0420*** 0.0234***          0.0019
                                                 (0.0089)  (0.0172)           (0.0497)         (0.0432)         (0.0038)        (0.0079)  (0.0070)         (0.0028)
     Period x region FE                              Y             Y              Y                Y                 Y              Y             Y            Y
     Period x industry FE                            Y             Y              Y                Y                 Y              Y             Y            Y
     Period x export share                           Y             Y              Y                Y                 Y              Y             Y            Y
     Firm FE                                         Y             Y              Y                Y                 Y              Y             Y            Y
     N (obs.)                                     11585         11585          11585             11585            11585          11585         11585        11585
     N (firms)                                     5063          5063           5063             5063              5063           5063          5063         5063
     Adj. R2                                      0.0837        0.0426         0.0282           0.00731           0.0496         0.0602        0.0249       0.0467
     RSS                                           1886         19845          113872           175840             899            2699          2122          390
     Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are clustered at firm-level. *10% level of
     significance, **5% level of significance, ***1% level of significance.
     Table A12: Eﬀects of forecast errors and forecast growth on ﬁrm performance using shares of total exports in the forecast error weights
                                           (1)        (2)           (3)                 (4)             (5)             (6)          (7)        (8)
     Dep. variable:                     log sales log exports log (1+ inv.         log (1+ inv.         log          log value    log value  log avg.
                                                              fixed tangible        intangible      employment         added      added per worker pay
                                                                  assets)             assets)                                      worker

     Weighted forecast error          0.0120*** 0.0556***          0.0540***         -0.0409*        0.0041***      0.0104*** 0.0148***         0.0021*
                                       (0.0023)  (0.0078)           (0.0186)         (0.0225)         (0.0015)       (0.0025)  (0.0027)         (0.0011)
     Weighted forecast growth         0.0069*** 0.0410***          0.0294**           -0.0272          0.0010       0.0062*** 0.0074***         0.0016**
                                       (0.0015)  (0.0054)           (0.0136)         (0.0166)         (0.0010)       (0.0018)  (0.0019)         (0.0008)

     Period x region FE                    Y             Y              Y                Y                Y              Y             Y             Y
     Period x industry FE                  Y             Y              Y                Y                Y              Y             Y             Y
     Period x export share                 Y             Y              Y                Y                Y              Y             Y             Y
     Firm FE                               Y             Y              Y                Y                Y              Y             Y             Y
     N (obs.)                           22199         22199          22199            22199            22199          22199         22199         22199
     N (firms)                           9306          9306           9306             9306             9306           9306          9306          9306
55




     Adj. R2                            0.0633        0.0734         0.0161           0.0131           0.0522         0.0140        0.0325        0.0295
     RSS                                 1595         22684          129128           291603            859            2378          2587           465
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level.
     *10% level of significance, **5% level of significance, ***1% level of significance.
     Table A13: Eﬀects of forecast errors and forecast growth on worker compensation using shares of total exports in the forecast error
     weights

                                         (1)            (2)          (3)            (4)           (5)           (6)           (7)           (8)
       Dep. variable:                    log        log hourly       log         log (1 +      log (1+       log total    share with    person FE
                                       monthly         wage        monthly       overtime     other pay)       hours       a degree
                                        wage                      base wage        pay)

       Weighted forecast error         0.0028**      0.0028**      0.0017**       0.0076        0.0100       0.0045**        0.0006       -0.0075
                                       (0.0011)      (0.0011)      (0.0008)      (0.0049)      (0.0064)      (0.0019)      (0.0005)      (0.0046)
       Weighted forecast growth        0.0017**      0.0016**       0.0005        0.0045        0.0064        0.0013        -0.0001        0.0005
                                       (0.0007)      (0.0007)      (0.0006)      (0.0035)      (0.0045)      (0.0013)      (0.0004)      (0.0033)

       Period x region FE                  Y             Y             Y             Y             Y             Y             Y             Y
       Period x industry FE                Y             Y             Y             Y             Y             Y             Y             Y
       Period x export share               Y             Y             Y             Y             Y             Y             Y             Y
       Firm FE                             Y             Y             Y             Y             Y             Y             Y             Y
56




       N (obs.)                         22199         22199         22199         22199          22199        22199         22199         12631
       N (firms)                         9306          9306          9306          9306          9306          9306          9306          6012
       Adj. R2                          0.0209        0.0228        0.0549        0.0219        0.00961       0.0583        0.0230        0.0493
       RSS                               472           472           249          15018          16663         1715          105           1545
       Notes: In each column, the dependent variable is the change between the averages of each 3-year period. Standard errors are clustered at firm-
       level. *10% level of significance, **5% level of significance, ***1% level of significance.
                 Table A14: Eﬀects of forecast errors and forecast growth on ﬁrm performance, period-industry-region eﬀects

                                             (1)        (2)           (3)                 (4)              (5)             (6)           (7)            (8)
     Dep. variable:                       log sales log exports log (1+ inv.         log (1+ inv.          log          log value     log value      log avg.
                                                                fixed tangible        intangible       employment         added       added per     worker pay
                                                                    assets)             assets)                                        worker

     Weighted forecast error            0.0698*** 0.1184***           0.1175*          0.1386**         0.0219***      0.0560***      0.0322***     0.0120***
                                         (0.0100)  (0.0192)           (0.0675)         (0.0654)          (0.0054)       (0.0098)       (0.0089)      (0.0046)
     Weighted forecast growth           0.0316*** 0.0772***            0.0555          0.0773*          0.0089***      0.0238***      0.0130**       0.0046*
                                         (0.0055)  (0.0126)           (0.0459)         (0.0460)          (0.0030)       (0.0058)       (0.0056)      (0.0026)

     Period x industry x region FE           Y             Y              Y                 Y                Y              Y              Y             Y
     Period x export share                   Y             Y              Y                 Y                Y              Y              Y             Y
     Firm FE                                 Y             Y              Y                 Y                Y              Y              Y             Y

     N (obs.)                             21957         21957          21957            21957             21957          21957          21957         21957
57




     N (firms)                             9216          9216           9216             9216              9216           9216           9216          9216
     Adj. R2                              0.0871        0.0824         0.0194           0.0240            0.0747         0.0470         0.0311        0.0447
     RSS                                   1507         21822          125415           281259             817            2485           2272          446
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level. *10%
     level of significance, **5% level of significance, ***1% level of significance.
     Table A15: Eﬀects of forecast errors and forecast growth on worker compensation and worker composition, period-industry-region
     eﬀects

                                                         (1)         (2)          (3)            (4)           (5)           (6)          (7)        (8)
               Dep. variable:                            log         log          log         log (1 +      log (1+       log total   share with person FE
                                                       monthly      hourly      monthly       overtime     other pay)       hours      a degree
                                                        wage        wage       base wage        pay)

               Weighted forecast error                 0.0092** 0.0087**        0.0062*      0.0330**         0.0240     0.0312***       0.0027     -0.0437**
                                                       (0.0042) (0.0042)        (0.0036)      (0.0144)       (0.0187)     (0.0066)      (0.0020)     (0.0171)
               Weighted forecast growth                 0.0034   0.0031          0.0001      0.0303***        0.0155     0.0110***       0.0000       -0.0128
                                                       (0.0024) (0.0024)        (0.0019)      (0.0104)       (0.0131)     (0.0041)      (0.0013)     (0.0108)

               Period x industry x region FE               Y           Y            Y             Y             Y             Y            Y             Y
               Period x export share                       Y           Y            Y             Y             Y             Y            Y             Y
               Firm FE                                     Y           Y            Y             Y             Y             Y            Y             Y
58




               N (obs.)                                 21904       21904        21904         21904         21904         21904        21904         12399
               N (firms)                                 9196        9196         9196          9196          9196          9196         9196          5919
               Adj. R2                                  0.0369      0.0387       0.0704        0.0383        0.0240        0.0794       0.0183        0.0701
               RSS                                       452         451          235          14292         16099          1588         101           1457
               Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-
               level. *10% level of significance, **5% level of significance, ***1% level of significance.
          Table A16: Eﬀects of forecast errors and forecast growth on ﬁrm performance, no lagged variables
                                             (1)        (2)           (3)                 (4)              (5)             (6)           (7)            (8)
     Dep. variable:                       log sales log exports log (1+ inv.         log (1+ inv.          log          log value     log value      log avg.
                                                                fixed tangible        intangible       employment         added       added per     worker pay
                                                                    assets)             assets)                                        worker

     Weighted forecast error            0.0698*** 0.1184***           0.1175*          0.1386**         0.0219***      0.0560***      0.0322***     0.0120***
                                         (0.0100)  (0.0192)           (0.0675)         (0.0654)          (0.0054)       (0.0098)       (0.0089)      (0.0046)
     Weighted forecast growth           0.0316*** 0.0772***            0.0555          0.0773*          0.0089***      0.0238***      0.0130**       0.0046*
                                         (0.0055)  (0.0126)           (0.0459)         (0.0460)          (0.0030)       (0.0058)       (0.0056)      (0.0026)

     Period x industry x region FE           Y             Y              Y                 Y                Y              Y              Y             Y
     Period x export share                   Y             Y              Y                 Y                Y              Y              Y             Y
     Firm FE                                 Y             Y              Y                 Y                Y              Y              Y             Y

     N (obs.)                             21957         21957          21957            21957             21957          21957          21957         21957
     N (firms)                             9216          9216           9216             9216              9216           9216           9216          9216
59




     Adj. R2                              0.0871        0.0824         0.0194           0.0240            0.0747         0.0470         0.0311        0.0447
     RSS                                   1507         21822          125415           281259             817            2485           2272          446
     Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level. *10%
     level of significance, **5% level of significance, ***1% level of significance.
     Table A17: Eﬀects of forecast errors and forecast growth on ﬁrm performance, excluding the 2007-2009 period
                                                       (1)         (2)          (3)            (4)           (5)           (6)          (7)        (8)
             Dep. variable:                            log         log          log         log (1 +      log (1+       log total   share with person FE
                                                     monthly      hourly      monthly       overtime     other pay)       hours      a degree
                                                      wage        wage       base wage        pay)

             Weighted forecast error                 0.0092** 0.0087**        0.0062*      0.0330**         0.0240     0.0312***       0.0027     -0.0437**
                                                     (0.0042) (0.0042)        (0.0036)      (0.0144)       (0.0187)     (0.0066)      (0.0020)     (0.0171)
             Weighted forecast growth                 0.0034   0.0031          0.0001      0.0303***        0.0155     0.0110***       0.0000       -0.0128
                                                     (0.0024) (0.0024)        (0.0019)      (0.0104)       (0.0131)     (0.0041)      (0.0013)     (0.0108)

             Period x industry x region FE               Y           Y            Y             Y             Y             Y            Y             Y
             Period x export share                       Y           Y            Y             Y             Y             Y            Y             Y
             Firm FE                                     Y           Y            Y             Y             Y             Y            Y             Y

             N (obs.)                                 21904       21904        21904         21904         21904         21904        21904         12399
60




             N (firms)                                 9196        9196         9196          9196          9196          9196         9196          5919
             Adj. R2                                  0.0369      0.0387       0.0704        0.0383        0.0240        0.0794       0.0183        0.0701
             RSS                                       452         451          235          14292         16099          1588         101           1457
             Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-
             level. *10% level of significance, **5% level of significance, ***1% level of significance.
     Table A18: Eﬀects of forecast errors and forecast growth on worker compensation and worker composition, excluding the 2007-2009
     period
                                                       (1)           (2)             (3)              (4)              (5)             (6)           (7)        (8)
                Dep. variable:                      log sales    log exports    log (1+ inv.     log (1+ inv.          log          log value     log value  log avg.
                                                                               fixed tangible     intangible       employment         added       added per worker pay
                                                                                   assets)          assets)                                        worker

                Weighted forecast error            0.0831*** 0.0962***             0.0873           0.0766          0.0207***       0.0765*** 0.0557***         0.0084*
                                                    (0.0091)  (0.0156)            (0.0607)         (0.0612)          (0.0053)        (0.0099)  (0.0096)         (0.0045)
                Weighted forecast growth           0.0399*** 0.0554***             0.0350           0.0381           0.0050*        0.0306*** 0.0239***          -0.0006
                                                    (0.0057)  (0.0107)            (0.0402)         (0.0434)          (0.0029)        (0.0057)  (0.0055)         (0.0026)

                Period x region FE                      Y             Y              Y                 Y                 Y               Y            Y             Y
                Period x industry FE                    Y             Y              Y                 Y                 Y               Y            Y             Y
                Period x export share                   Y             Y              Y                 Y                 Y               Y            Y             Y
                Firm FE                                 Y             Y              Y                 Y                 Y               Y            Y             Y
                N (obs.)                             22186         22186          22186             22186             22186           22186        22186         22186
61




                N (firms)                             9302          9302           9302              9302              9302            9302         9302          9302
                Adj. R2                              0.0832        0.0638         0.0216            0.0120            0.0771          0.0535       0.0249        0.0935
                RSS                                   1558         22374          113374            285624             846             2714         2472          501
                Notes: In each column, the dependent variable is the change between the average of each 3-year period. Standard errors are clustered at the firm-level. *10%
                level of significance, **5% level of significance, ***1% level of significance.
A.2       Appendix Figures


    Figure A1: Correlations between weighted forecast errors and initial ﬁrm characteristics




Notes: Figure depicts correlations between weighted forecast errors and ﬁrm attributes in 2006.




                                                      62
                  Figure A2: Actual and forecasted GDP growth in top destinations

                     Spain                                Germany                            France
      4




                                            4




                                                                               4
      2




                                            2




                                                                               2
      0




                                            0




                                                                               0
    %




                                            %




                                                                               %
      -2




                                            -2




                                                                               -2
      -4




                                            -4




                                                                               -4
      -6




                                            -6




                                                                               -6
           2006   2010        2014   2018        2006    2010   2014    2018        2006   2010   2014   2018


                         UK                             United States                      Netherlands
      4




                                            4




                                                                               4
      2




                                            2




                                                                               2
      0




                                            0




                                                                               0
    %




                                            %




                                                                               %
      -2




                                            -2




                                                                               -2
      -4




                                            -4




                                                                               -4
      -6




                                            -6




                                                                               -6

           2006   2010        2014   2018        2006    2010   2014    2018        2006   2010   2014   2018



Notes: Figure depicts actual GDP growth (gray bars) and forecast GDP growth (black bars) for the six most
important export destinations in 2006.




                                                        63
              Figure A2: Actual and forecasted GDP growth in top destinations (cont.)

                      Italy                              Belgium                             Sweden
      5




                                           5




                                                                               5
      0




                                           0




                                                                               0
    %




                                         %




                                                                               %
      -5




                                           -5




                                                                               -5
      -10




                                           -10




                                                                               -10
            2006   2010   2014    2018           2006   2010     2014   2018         2006   2010   2014   2018


                   Switzerland                               Finland                        Denmark
      5




                                           5




                                                                               5
      0




                                           0




                                                                               0
    %




                                         %




                                                                               %
      -5




                                           -5




                                                                               -5
      -10




                                           -10




                                                                               -10

            2006   2010   2014    2018           2006   2010     2014   2018         2006   2010   2014   2018



Notes: Figure depicts actual GDP growth (gray bars) and forecast GDP growth (black bars) for the seventh to
twelfth most important export destinations in 2006.




                                                        64
              Figure A2: Actual and forecasted GDP growth in top destinations (cont.)

                    Angola                              Singapore                             Brazil
      30




                                           30




                                                                               30
      20




                                           20




                                                                               20
     %




                                          %




                                                                                %
    10




                                         10




                                                                               10
      0




                                           0




                                                                               0
      -10




                                           -10




                                                                               -10
            2006   2010   2014    2018           2006   2010     2014   2018         2006   2010   2014   2018


                     China                                   Poland                          Turkey
      30




                                           30




                                                                               30
      20




                                           20




                                                                               20
     %




                                          %




                                                                                %
    10




                                         10




                                                                               10
      0




                                           0




                                                                               0
      -10




                                           -10




                                                                               -10

            2006   2010   2014    2018           2006   2010     2014   2018         2006   2010   2014   2018



Notes: Figure depicts actual GDP growth (gray bars) and forecast GDP growth (black bars) for the thirteenth
to eighteenth most important export destinations in 2006.




                                                        65
               Figure A3: Z-score of ﬁrms with high and low-skilled managers, 2016



                        Panel A: Managers identiﬁed by occupational category
                   .4
                   .3
                   .2
                   .1
                   0




                        -4                    -2                     0     2                4
                        kernel = epanechnikov, bandwidth = 0.2429


                              Panel B: Managers identiﬁed as top 5% earners
                   .4
                   .3
                   .2
                   .1
                   0




                        -4                    -2                     0     2                4
                        kernel = epanechnikov, bandwidth = 0.2244




Notes: Figure depicts the distribution of standardised values of Z-scores of ﬁrms with high share of managers with
a degree (in gray) versus ﬁrms with a low share of managers with a degree (in black). High versus low shares are
deﬁned as above versus below the median. In Panel A, managers are identiﬁed by occupational category, while
in Panel B managers are identiﬁed as the top 5% earners.



                                                                    66
A.3     Data sources and description

In this Appendix, we provide further details about the data sets used in the empirical analysis.
We combine and examine several sources of panel data from Portugal spanning the period 2006-
2018.
   Employer-employee data: Quadros de Pessoal (QP) [Personnel Records] is a compulsory
census run by the Ministry of Employment covering the population of ﬁrms with wage earners
in manufacturing and services. Each ﬁrm is required by law to provide information on an
annual basis about its characteristics and those of each individual that comprises its workforce.
Firm-level information includes annual sales, number of employees, industry code, geographical
location, date of constitution and percentage of capital that is foreign-owned. The industry code
is deﬁned at the 5-digit level of the National Classiﬁcation of Economic Activities (CAE). The
set of worker characteristics includes wages (monthly base wage, overtime pay, and other regular
and irregular components of pay), gender, age, schooling, date of starting, detailed occupation
and hours worked. A worker may also be matched to the ﬁrm. An important feature of these
data is that particular care is placed on the reliability of the information. Indeed, the data
are used by the Ministry of Employment for checking the employer’s compliance with labour
law. Moreover, Portuguese law makes it compulsory for ﬁrms to make this information available
to every worker in a public place of the establishment. Extensive checks have been performed
to guarantee the accuracy of worker and ﬁrm data. After these checks, we kept for analysis
full-time wage earners working at least 100 hours a week, aged between 20 and 60 years old.
   Firm census: Using common unique ﬁrm identiﬁers, we supplement the ﬁrm-year data from
QP with information from the Sistema de Contas Integradas das Empresas (SCIE) [Enterprise
Integrated Accounts System], a yearly census of ﬁrms run by National Statistics Institute (INE).
Since 2006, the main source of the census is administrative data from Simpliﬁed Information
on Enterprises, which consists of ﬁscal, accounting and statistical information provided by ﬁrms
through a single form transmitted electronically through www.portaldasfinancas.gov.pt. By
ﬁling this form, ﬁrms fulﬁl four diﬀerent legal obligations: the annual statement of ﬁscal and
accounting information, the accounts registry, the provision of statistical information to INE,
the provision of information of annual accounts data for statistical purposes to the Bank of
Portugal, and the provision of statistical information to the General Directorate of Economic
Activity of the Ministry of the Economy in the context of the legal regime for access and
provision of activities of commerce, services and hospitality. The main objective of SCIE is to
characterize the economic and ﬁnancial behaviour of ﬁrms, through a set of relevant variables for



                                               67
the corporate sector, as well as ﬁnancial ratios, which are commonly used in the ﬁnancial analysis
of ﬁrms. This data set includes information on total sales, investment, total employment, wage
bill, industry, location, among several other variables.
   International trade statistics: We merge the above data sets with yearly data on ﬁrms’
                              ısticas do Com´
export transactions from Estat´             ercio Internacional (ECI) [Foreign Trade Statistics]
from INE. This is the country’s oﬃcial information source on imports and exports. It comprises
the export ﬂows of virtually all exporting ﬁrms, and provides detailed information on the product
exported, the destination market, and the value and quantity exported. These data are collected
through two diﬀerent systems. Information on trade with countries outside the EU (external
trade) is obtained from the customs clearance system, which covers the universe of external trade
transactions. The data on the transactions with other EU member States (internal trade) are
collected through a separate survey called the Intrastat. In this case, the information providers
are companies engaged in internal trade and registered in the VAT system whose value of annual
shipments exceeds a given statistical threshold. This (legally binding) cut-oﬀ is deﬁned by each
member state so that as many of the smallest exporters as possible are exempted from submitting
statistical declarations, while the quality standard of the statistics remains adequate. Exported
products are classiﬁed according to the eight-digit level of the Combined Nomenclature (CN).
This is the most detailed product classiﬁcation system for foreign trade statistics in the EU.
Export values in these data are free-on-board, thus excluding any duties or shipping charges.
                                                             erito `
   Management practices survey: We further use data from Inqu´          aticas de Gest˜
                                                                   as Pr´             ao
(IPG) [Management Practices Survey] for 2016. IPG is a non-periodical survey conducted by
INE, which collects information on the perceptions of top managers about the management
practices of their ﬁrms. The 2016 survey was the ﬁrst and only of its kind collected in Portugal.
It seeks to evaluate the importance of management practices for ﬁrm productivity, as well
as other key indicators that make it possible to evaluate diﬀerences in productivity between
Portuguese ﬁrms. IPG employed a stratiﬁed sample of ﬁrms operating in Portugal covering the
whole non-ﬁnancial private sector in 2016, excluding micro ﬁrms (with less than ﬁve employees).
The sample is representative by sector (20 sectors corresponding of aggregations of the 2-digit
level of the CAE), ﬁrm size and age, as well as belonging (or not) to a conglomerate. The IPG
survey was an electronic survey targeted at managers, who are typically senior enough to have
a good understanding of management practices. These protocols helped yield a 86.7% response
rate. The survey includes questions that make it possible to evaluate management practices
in three main areas: (1) Strategy, monitoring and information; (2) Human Resources; and
(3) Management and social responsibility systems. We selected questions on 18 management


                                                68
practices that are closely related to those adopted in Bloom and Reenen (2007). First, we
classiﬁed the 18 practices into 4 categories: operational (2 practices), targets (4 practices),
monitoring (10 practices) and incentives (2 practices). Following their approach, our measure of
management quality was constructed by z-scoring (normalizing to mean 0, standard deviation
1) the 18 individual questions in IPG, averaging them, and then z-scoring the average. This
process yields a management practice score with mean 0 and standard deviation 1.
   Actual and forecasted GDP growth: We further use yearly information on actual and
recently forecasted GDP growth from the World Economic Outlook (WEO) of the Interna-
tional Monetary Fund (IMF). WEO is usually published twice a year (in April and Septem-
ber/October). It presents IMF staﬀ economists’ analyses of global economic developments during
the near and medium term. The WEO database is created during the biannual WEO exercise,
which begins in January and June of each year and results in the April and September/October
WEO publication. Selected series from the publication, including actual and forecasted GDP
growth are available in a database format at https://www.imf.org/en/Publications/SPROLLs/
world-economic-outlook-databases. Every April and October, the WEO provides year-
ahead and current-year GDP growth forecasts. We refer to the year for which the forecast is
being made as the target year. Forecasts made in the Fall WEO before the target year are
called year-ahead forecasts and those made during the Spring target year are called current-year
forecasts. During our sample period, forecast data are available for 195 countries. After merging
these data with ECI we were left with 174 destinations, which account for 99.7% of all exports
in 2006. Table A2 reports the export shares to the main destinations in 2006, both in the full
ECI data and in the estimation sample.


A.4        Variable deﬁnitions in the econometric analysis

This section describes the variables used in the econometric analysis and the corresponding
sources:

 Sales: Total value of sales (in Portugal and abroad) during the reference year. Source: SCIE.

 Exports : Export revenue of a ﬁrm during the reference year. Source: ECI.

 Investment in ﬁxed tangible assets : Investment in ﬁxed tangible assets during the reference
      year. Source: SCIE.

 Investment in intangible assets : Investment in intangible assets during the reference year.
      Source: SCIE.

                                               69
Employment : Number of employees during the reference year. Source: SCIE.

Value added : Value added created by a ﬁrm during the reference year evaluated at market
    prices. Source: SCIE.

Share of foreign-owned capital : Share of capital that is foreign-owned in current year. Source:
    QP.

Share of state-owned capital : Share of capital that is state-owned in current year. Source: QP.

Firm age : Number of years passed since a ﬁrm was ﬁrst registered in Portugal. Source: QP.

Exporter : Indicator variable =1 if ﬁrms records some export revenue in reference year. Source:
    ECI.

Export to sales ratio : Ratio between exports and total sales in reference year. Sources: ECI
    and SCIE.

Number of destinations served : Number of diﬀerent export destinations served by a ﬁrm during
    the reference year. Source: ECI.

Weighted forecast error : Weighted diﬀerence between observed GDP growth in destinations
    and growth forecast for that destination in the Spring edition of the WEO of the IMF
    (weighs: share of exports to that destination in 2006). Sources: ECI and IMF.

Weighted forecast growth : Weighted growth forecast for destinations in the Spring edition of
    the WEO of the IMF (weighs: share of exports to that destination in 2006). Sources: ECI
    and IMF.

Avg. labor costs : Ratio between the wage bill (including wages, social security contributions,
    beneﬁts, etc.) and the number of paid employees. It corresponds to the average gross
    earnings per paid worker. Source: SCIE.

Avg. monthly wage : Average monthly wage. Source: QP.

Avg. hourly wage : Average hourly wage. Source: QP.

Avg. monthly base wage : Average monthly base wage. Source: QP.

Avg. overtime pay : Average overtime pay. Source: QP.

Avg. other pay : Average other components of pay. Source: QP.


                                              70
 Total hours : Total hours worked by employees at the ﬁrm. Source: QP.

 Share with a degree : Share of workers with higher education. Source: QP.

 Avg. person ﬁxed eﬀects : Average estimated person eﬀect using AKM models. Source: QP.

 Mean employee ability : Average estimated person eﬀect of non-managerial employees using
     AKM models. Source: QP.

 Mean manager ability : Average estimated person eﬀects of top managers using AKM models.
     Source: QP.

 Share of female employees : Share of female employees in reference year. Source: QP.

 Share of high-skill managers : Share of workers with higher education. Managers deﬁned as
     workers in top ﬁfth percentile of wage distribution. Source: QP.

 Share of employees for which ﬁxed eﬀects were computed (raised to the power of 3) : Share of
     workers in connected ﬁrms. Source: QP.

All monetary variables are in euros and have been deﬂated to constant 2018 prices using the
Portuguese GDP deﬂator or the CPI Index (for wages) from INE.




                                              71