16 Development Finance Agenda

Executive SummaryThere is increased interest in assessing the role of De-velopment Finance Institutions (DFIs) in promoting jobs – direct and indirect – and structural transfor-mation – from low productivity to high productivity activities. This article reviews the practice of DFIs so far in reporting job and productivity impacts. Whilst most DFIs report on the number of jobs supported, there has been much less attention to the indirect job impacts of DFIs and hardly anything on productivity impacts. We summarise new evidence we generated on the productivity impacts of DFIs suggesting tentatively that they can enhance jobs and productivity through providing finance. We conclude there is a need to complement studies on ex-ante multiplier effects of DFIs with other studies (e.g. case studies, econometric studies) on the actual impact on jobs and productivity. Introduction

There is increased interest in the role of Develop-ment Finance Institutions (DFIs) in promoting devel-opment, in part because their exposure is increasing rapidly (doubling commitments over 2003-2009) and in part because there is a growing recognition that aid agencies and DFIs need to create more impact with less funding. The past decade has seen an increased interest in providing more accurate impact assessment of the activities of DFIs at the micro and macro levels (e.g. Massa and Te Velde, 2011). Various DFIs are today aiming to make employment generation – direct but also indirect with a focus on the issue of quality job creation – central in their impact evaluation.

Job creation, productivity and structural change are the main development challenges for low income countries at present (UNECA, 2011; Macmillan and Rodrik, 2011). DFIs help in the process of job creation and structural change by addressing market failures through the provision of loans, grants, equity and guarantees which can lead to better and higher qual-ity investment in poor countries. This note summarises two UK DFID funded studies in Massa (2013) and Jouanjean and Te Velde (2013). The effects of DFIs on development cannot be taken for granted. There is increased research on the effects of DFIs generally (Massa, 2011; Te Velde, 2011) and on job creation more specifically (IFC, 2013), but less on the impact on structural transformation. There is a lively debate on appropriate methods to use in such studies. DFIs affect job creation and structural transformation through a number of static and dynamic effects. Static and direct effects. DFIs affect job creation directly by being additional to domestic investment and they can have a direct effect on productivity through changing the composition and hence the economic structure of an economy. Dynamic and indirect effects. In a dynamic sense, DFIs also create jobs through forward and backward linkages (and the induced effects these generate) and can foster technical change in companies, with possible spillover effects for the sector and the whole economy. DFIs set economic, social and environmental performance standards, have representatives on com-pany boards, direct fund managers, provide technical assistance and act as a port of knowledge through

which investee companies can adopt new product and process innovation, which can provide spillover effects to other firms in the sector or economy as a whole. A number of methods try to measure the impact of DFIs on job creation effects and structural transforma-tion. For example, IFC (2013) distinguishes between direct jobs (jobs in entities directly supported), indi-rect jobs (jobs supported through suppliers), induced effects (jobs supported through increased spending power from increased jobs), second-order growth ef-fects (jobs created through productivity effects) and displaced jobs (e.g. jobs displaced by the DFI sup-ported job). Whilst there is some harmonisation to examine the direct jobs, there is not one acceptable way of examining the indirect job effects. Generally, there are many different estimation meth-ods of the job creation effects, each associated with positive and negative aspects (table 1)

a. Direct jobs in DFI supported companies;b. Production function based estimates;c. Input-output and multiplier model based estimates;d. Econometric estimates based on empirical findings; ande. Detailed case study estimates.

DFIs report the number of direct jobs in companies supported by DFIs (Table 2). Direct job estimates both underestimate and overestimate the true effect. They overestimate because these jobs may also be due to investors other than DFIs (we do not have information on the counterfactual). They underestimate because they do not take into account other indirect and in-duced jobs.

The Role Of Development Finance Institutions In Promoting Jobs And Structural Transformation

Marie-Agnes Jouanjean Isabella Massa Dirk Willem te VeldeOverseas Development Institute

The

role

of

deve

lopm

ent

fina

nce

inst

itut

ions

in

prom

otin

g jo

bs a

nd s

truc

tura

l tr

ansf

orm

atio

n

Development Finance Agenda 17

Table 1: Pros and cons of assessment methods for job creation effects of DFIs

Approach Positive aspects Negative aspects Possible data sources

Direct employment in DFI supported projects

Directly measurable Does not measure displacement effects, indirect, induced or second-order growth effectsMight overstate effects directly attrib-utable only to DFIs

Company reports

Macro production function approaches multiplier analysis

Can be used at macro level to see how (DFI) investment leads to output chang-es (could use ICOR, C-D / CES / Leontief / TFP approaches) which could then lead to employment effects. Useful for quick assessments at aggregated level, for manufacturing, but less useful when the quantity of “output” is not main or only factor of interest.

Involves use of assumptions, estima-tions of production functions and employment intensities and are based on predicted rather than empirical effects.

Does not measure second order growth / productivity effects

Requires (sectoral-level) national accounts

Input-output models Useful to examine backward linkages across industries in traditional indus-tries and hence indirect employment , could be linked to different types of skills, tax etc. to get a SAM

Useful to obtain multipliers by sectors relatively easily.

Not useful in case of transformative changes in production structures (e.g. large scale infrastructure investments) or when inputs are price dependent and substitutable, or when behavioural links change (in which case input-output coefficients would change).

Measures expected impacts

Labour force surveys

National accounts

Firm level / national level econometrics Useful to examine the empirical effects of the level and quality of services supply on firm performance amongst a range of factors (and hence the induced effects, including on employment)

Data intensive(needs panel data), needs good identi-fication strategies.

Existing firm level surveys (e.g. WB enterprise survey)National databases

Household level econometrics Useful to examine the importance of DFI supported services in the household budget

Data intensive(panel data)

Household level surveys

Case studies Useful to get detailed impact to verify multiplier effects or aggregated econo-metric effects.

Data intensive, difficult to obtain macro effect and counterfactual

Field work

Source: Jouanjean and Te Velde (2013)

Table 2: Direct jobs supported by DFIs (portfolio / created in 2011), examples

DFIs Direct jobs supported (by portfolio in 2011)

Direct jobs created in 2011

IFC 2,500,000 200.000

CDC 976,000

DEG 800.000 110,000

Proparco 89,000

IFU 4,500

Source: Massa (2013), DEG, IFC, Proparci Data refer to jobs supported by portfolio or data on new jobs created in 2011. Note that these include only direct jobs and do not take into account indirect jobs.

18 Development Finance Agenda

Table 3: Relevance of DFI impacts on employment creation by sector xe

Sector of DFI investment

Direct job eff ects Indirect job eff ects (static and dynamic)

Induced and second order growth eff ects

Manufacturing such as garments

Very important (but depends on type of manufacturing)

Potentially important Less important

Tourism Medium important Very important Less important

Infrastructure Less important Mostly temporary Very important

Agriculture Very important Less important Less important

Source: Massa (2013), Jouanjean and te Velde (2013)

Figure 1: Trade – off between value addition per job and number of jobs per investment

Source: data from Kapstein et al (2012) for Tunisia. Jobs include direct and indirect eff ects based on input-output models. Value addition (US$) per job (vertical axis) and number of jobs per US$ Mn investment (horizontal axis)

A number of studies have examined the indirect job im-pacts and confi rm that the indirect eff ects diff er mark-edly by sector (see Table 3 for a characterisation). One study for Tunisia conducted for the IFC (2013) shows the trade-off between jobs created and value added by those jobs (Figure 1). If such data are repre-sentative for all developing countries it would suggest that DFIs such as DEG, EIB, Swedfund and Finfund that

are relatively more exposed to industry and agribusi-ness have the largest employment generation eff ects, whilst other DFIs (e.g. CDC, EBRD, IFC, Proparco) might have a larger potential for increasing value-addition or growth. Jouanjean and Te Velde (2013) provide new produc-tion based estimates of the direct and indirect employ-ment eff ects of DFIs at national level. In order to test

the eff ects of DFIs on structural change and labour productivity, we derive and estimate a labour demand equation which incorporates the eff ects of DFIs on productivity and structural change. According to the model, labour productivity increases (labour intensity decreases) when capital substitutes for labour decreases (e.g. when the real wage increases) or when labour aug-menting technical change increases e.g. through DFIs

Development Finance Agenda 19

Effects on labour intensity

Effect of the treatment (minus con-structed counterfactual) after one year

Effect of the treatment (minus con-structed counterfactual) after two years

Effect of the treatment (minus con-structed counterfactual) after three years

(1) (2) (3)

Treated -0.033 -0.072* -0.132**

(0.221) (0.062) (0.013)

Constant -0.017 -0.044 -0.053

(0.490) (0.232) (0.290)

Observations 244 210 171Notes: see Jouanjean and te Velde (2013) for further details

Table 4 Estimating the effects of DFI intervention on labour intensity using a propensity-score matching difference in differences estimator

Without controls all developing countries

All develop-ing countries

LI countries LMI countries UMI countries All develop-ing countries

LI countries LMI countries UMI countries

One year difference: ((t+1)-t) Three year difference: ((t+3)-t)

“Treated group” -0.032*** -0.057*** -0.027 -0.082** -0.030 -0.120*** -0.075* -0.108* -0.146**

(0.008) (0.001) (0.281) (0.017) (0.195) (0.000) (0.066) (0.059) (0.014)

Log Population 15-65

-0.004 -0.001 -0.003 -0.005 -0.026*** -0.023 -0.028* 0.002

(0.357) (0.961) (0.727) (0.412) (0.006) (0.347) (0.088) (0.881)

Net ODA received per capita (current US$)

-0.000 -0.000 -0.000 0.000 -0.002*** -0.001 -0.001* 0.001

(0.135) (0.812) (0.182) (0.472) (0.000) (0.337) (0.055) (0.337)

Log GDP per Capita (current US$)

-0.026** -0.072** -0.028 -0.059** -0.122*** -0.194*** -0.021 -0.235***

(0.019) (0.020) (0.158) (0.017) (0.000) (0.003) (0.696) (0.001)

Agriculture value added (% of GDP)

-0.002** -0.004** -0.001 -0.004 -0.011*** -0.012*** -0.015*** -0.018**

(0.050) (0.039) (0.225) (0.125) (0.000) (0.008) (0.002) (0.040)

Manufacturing value added (% of GDP)

-0.000 -0.002 0.000 -0.001 -0.001 -0.000 0.002 -0.005

(0.921) (0.377) (0.917) (0.471) (0.572) (0.924) (0.517) (0.216)

Manufactures imports (% of merchandise imports)

-0.001 -0.000 0.002* -0.000 -0.001 -0.000 -0.004* 0.003

(0.158) (0.556) (0.061) (0.061) (0.635) (0.797) (0.096) (0.173)

Constant -0.013 0.372** 0.740** 0.629** 0.649** 1.809*** 2.137*** 1.005 2.183***

(0.193) (0.018) (0.040) (0.035) (0.014) (0.000) (0.006) (0.115) (0.004)

Observations 422 422 114 168 140 296 85 123 88

Table 5: The effects of DFI “treatment” on labour intensity using Weighted Least Square regressions

Notes: ***, ** and * respectively indicates significance at the 1%, 5% and 10% levels.Regressions include year fixed effects. Robust standard errors. See Jouanjean and te Velde (2013) for further details

20 Development Finance Agenda

It is the latter effect we are most interested in e.g. how do DFIs affect labour productivity through increasing labour augmenting technical change? Based on this model, the paper uses various estima-tion techniques to identify the effect of the support by DFIs on labour productivity. According to the model, the labour intensity of production is explained by the real wage and labour-augmenting technical change captured by a country specific trend and the impact of DFI. The objective in this analysis is to capture the sig-nal effect of support by DFIs in addition than the effect of the amount invested. It should be noted that with the chosen approach we estimate the overall impact of the intervention of DFIs on the economy. This overall impact can stem from the direct but also from indirect and spill overs such as those mentioned previously.However, there might be various challenges that hin-der a proper identification of the true effect of DFIs on labour productivity when using such specification. One problem in evaluating the impact of DFIs is that the selection of countries in which DFIs invest is not random. This introduces difficulties in estimating the effect of DFI support. DFIs are supposed to support countries in which the investment environment is too risky for the private sector to invest by itself. It is possible that (i) labour productivity is lower in countries where DFIs are active than in countries in which they do not invest (but also vice versa, for conflict affected countries); (ii) growth in labour productivity in countries in which DFIs invest could be initially lower than growth in labour produc-tivity in countries in which they do not invest due to unobserved reasons; and (iii) the expected increase in growth in labour productivity in countries in which DFIs invest is higher than without DFIs investment but remains lower than growth in labour productivity in countries in which they do not invest. The paper adopts various methods controlling for this selection bias based on a “treatment effect” ap-proach to the analysis of the impact of DFI support on labour productivity. The underlying approach is to recreate an experimental setting allowing comparing changes in an outcome for DFIs intervention benefi-ciaries and a control group identified as similar accord-ing to observable characteristics. The Average Treat-ment Effect (ATE) estimates the difference in labour productivity between the treated and control groups The paper also implements methodologies allow-ing for the calculation of a long-run effect of DFIs in-tervention with the evaluation of the impact of a DFI investment over three years on productivity growth as well as the impact of three years of DFIs intervention in a row on productivity. The paper uses a number of different data sets. First, it constructs and uses data on DFIs using information provided by EBRD, EIB, CDC, DEG, IFC, and Proparco on

their investments in the whole economy for all coun-tries, see Massa (2011) and Te Velde (2011). It uses a comprehensive database covering investments from all those institutions for the period 2004 to 2009. The analysis of the probability of investment by a DFI in a particular country highlights that countries with a lower GDP per capita are less likely to receive DFI investment. The same holds for countries for which agriculture represents a larger share of GDP. On the other hand, the greater the importance of manufac-tureing to GDP as well as the greater manufacturing in proportion of imports, the higher the probability of DFI investment in the country. This reflects that invest-ment by DFIs is more likely to take place in better-off countries with a dynamic manufacturing sector. Such results could suggest that DFIs investments are more likely to influence average labour productivity and structural change in lower income countries, once we control for the selection bias of DFI investment. However, those results are based on few observations, and should therefore be treated with caution. After three (two) years, results show that on aver-age, compared to the control group, the increase in labour productivity is around 13 % (7%) for countries with DFI support compared to the control group. The estimates in table 4 use a propensity-score matching difference-in-differences (PSM-DID) estimator For each propensity score, it tests the immediate effect of DFI interventions - effect in the treatment year on the inverse of labour productivity - and longer term effect of DFIs interventions - persistence after 2 and 3 years of the treatment on the inverse of labour productivity. A negative coefficient means a lower labour intensity and hence a higher labour productivity. It uses Kernel esimation for the matching technique. Table 5 presents results using Weighted Least Squares, see Jouanjean and Te Velde (2013) for further details. The results are consistent with the average treatment effects in table 1, with a significant and expected sign on the coefficient for both the immedi-ate and longer-term effect of a DFI intervention, and with a larger impact over the longer term. Examining the effects for different country groups in more detail, the paper finds that the short run effects are strongly significant for lower middle income (LMI) countries, whilst the long run effects three years after a DFI in-vestment suggest a significant impact on labour in-tensity for all groups with the effect increasing from LI (7.5%), to MLI (10.8%) and UMI (14.6%) countries. The econometric study on the productivity effects of DFIs provides only initial results which need to be extended in a number of ways in the future. Future studies could do estimations (i) on the productivity and employment effects at sector level; (ii) using a variety of measures of DFI exposure, and the variety of financial instruments they employ; (iii) using other

estimations procedures and instruments; (iv) using methods that can help to understand which factors are conducive to greater effects; and (v) using differ-ent measures and data on employment. There are two issues for discussion. The first is on how DFIs and their shareholders should report on em-ployment impacts and whether and how they should also report impacts on (labour) productivity and struc-tural transformation. The second is on how DFIs can improve their impact on jobs and labour productivity. The latter issue is very important, but we did not elab-orate on it further in this study. Here we argue that adequate reporting is required to fully understand the role of support by DFIs (and similar organisations). This could then lead to a discussion on what is required to improve the impact.

Conclusions And Policy SuggestionsThere are various options for DFIs and shareholders to (continue to) report on employment and productivity impacts:

a) DFIs can report their direct employment “impacts” – e.g. number of employees in DFI supported com-panies. DFIs are harmonising the variables. But we argue this is still insufficient for designing appropriate policy, given the nature of the indirect effects.

b) DFIs can (continue to) extend reporting on employ-ment impact by including direct employment impacts through:

(i) Systematic monitoring of each DFI investment and their suppliers;(ii) Selected in-depth case studies of linkages; and/or(iii) Ex-ante assessments using rules of thumb (e.g. sector X in country Y is associated with jobs multiplier Z).

c) DFIs can (continue to) extend reporting on induced employment and labour productivity through:

(i) Selected in-depth case studies (ii) Detailed econometric modelling

There are varying capacities and incentives in DFIs to report on employment and productivity impacts. DFIs can continue to monitor the direct employment impacts, but further discussion and co-ordination of capacities (e.g. via a knowledge platform as started by the IFC) would be required on reporting on indirect impacts to avoid duplication and benefit from the public goods aspects of knowledge generation. Whilst it would be valuable to collect a set of ex-ante multi-plier effects based e.g. on input-output models, and

Development Finance Agenda 21

to which DFIs can contribute peer-reviewed studies on a continuing basis, studies on the expected eff ects of job creation will need to be augmented by studies (e.g. case studies, econometric studies) on the actual impact and such studies also need to include the im-pact on productivity.

ReferencesIFC (2013), IFC Jobs study, IFC Jouanjean, M.A and D.W. te Velde (2013), The role of development fi nance institutions in promot-ing jobs and structural transformation, A quantitative assessment¸ODI Working Paper. Kapstein, E.B., R. Kim, and H. Eggeling (2012), Modeling the Socio-Economic Impact of Potential IFC Investments in Tunisia - An Assessment of Employ-ment and Value Added, study for the IFC. Kim, R. , H. Eggeling, E.B. Kapstein (2011), Quan-tifying the IFC’s Contributions to Job Creation in Ghana and Jordan: A Scoping Study, Powerpoint 3 August 2011, Steward Redqueen Löwenstein, W. (2011), Evaluation of selected DEG Energy Sector Projects in Asia, DEG, IEE. Massa, I. (2013), A brief review of the role of de-velopment fi nance institutions in promoting jobs and productivity change, ODI report. Massa, I. (2011) The Impact of Multilateral De-velopment Finance Institutions on Economic Growth.’ Paper prepared for the UK DFID. McMillan, M. and D. Rodrik (2011). “Globalization, Structural Change and Productivity Growth.” NBER Working Paper Series, Vol. w17143. http://ssrn.com/abstract=1866102 UNECA (2011), Industrial Policies for the Structural Transformation of African Economies: Options and Best Practices (Addis Ababa: United Nations Commis-sion for Africa). Velde, D.W. te (2011), “The role of development fi -nance institutions in tackling global challenges”, Paper prepared for UK DFID.

Footnotes1. We defi ne these as international development fi -nance institutions providing fi nance (loans, equity etc.) to the private sector, e.g. IFC (International Fi-nance Corporation), CDC (UK Development Finance in-stitution), FMO (Dutch Development Finance Institu-tion, DEG (German Investment Cooperation), Swefund (Swedish Development Finance Institution), Norfund (Norwegian Development Finance Institution) and parts of EIB (European Investment Bank) etc. We ex-clude IDA and national development banks.

2. e.g. speeches by UK Secretary of State Justin Green-ing in March 2013

Conscious conferencing

SPIER WINE FARMThe Spier Conference Centre includes the auditorium that seats 430 people and three breakaway rooms which can be used separately, or combined to seat 250 people. Conference rooms are adjacent to the Eerste River and have views of the Helderberg Mountains. In winter, delegates can relax around fireplaces and in the summer, chill on the pool terrace. The Spier Hotel has 155 rooms to accommodate delegates, and restaurants, wine and leisure activities on site to cater for different needs.

We offer ‘conscious conferencing’, which means that you can help us in our support of local industries and fair trade, and in our commitment to a sustainable planet. For your tailor-made package, please email conference@spier.co.za or call 021 809 1100.