Sie sind auf Seite 1von 25
Strikes and Political Activism of Trade Unions: Theory and Application to Bangladesh Author(s): Jean-Paul Azam

Strikes and Political Activism of Trade Unions: Theory and Application to Bangladesh Author(s): Jean-Paul Azam and Claire Salmon Reviewed work(s):

Source: Public Choice, Vol. 119, No. 3/4 (Jun., 2004), pp. 311-334 Published by: Springer

Accessed: 26/11/2011 05:17

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact

information about JSTOR, please contact Springer is collaborating with JSTOR to digitize, preserve

Springer is collaborating with JSTOR to digitize, preserve and extend access to Public Choice.

Public Choice 119: 311-334, 2004.

© 2004 KluwerAcademicPublishers. Printedin theNetherlands.


Strikes and political activism of trade unions: Theory and application to Bangladesh *


1ARQADE,IDEI, Universityof Toulouse, InstitutUniversitairede France,F-31000 Toulouse, France; E-mail:; 2 Universityof Savoie, F-74016 Annecy, France

Accepted 3 March2003

Abstract. Thebehaviorof Bangladeshi tradeunionsis characterized by political activismand

momentous strikes, calledhartals there, have played a crucial part in most politicalchanges in this country. We offera theoreticalframeworkfor discussing this fact, andwe test empirically

its main predictionby strikesin Bangladesh.

bringing out the political cycle that characterizesthe occurrenceof

1. Introduction

The economic importance of trade unions in developing countries is not commensuratewith the size of their membership. In Bangladesh, the share of the active population concerned by unionizationwas officially estimated between 3 and 4% in 1992. However, this figure is largely irrelevantfor an evaluationof their power, as the urban population was only about 18%of the total Bangladeshipopulation in 1995, and virtually no tradeunionsexist in the ruralsector.Withinthe urbanite workingpopulation, and especially in the formal sector, the picture is quite different.Almost 100%of the workers and employees of the public sector are unionized, while one out of six of the wage earnersin the private formal sector are unionized. The jute and cotton sectors, which were nationalizedin 1971, in the wake of the struggle for independence, and then privatized to some extent in the 1980s, are the most unionizedsectors.Unions arealso important in the transport sectorand in variousservices. As emphasizedby Pencavel (1995), among others, unions in developing

countries get their power

parties, and in many cases with the government. The Bangladeshi tradeuni- ons areno exception, andare well knownfor their lobbying the government

* This paper owes much to many people met in Bangladeshduring our numerousvisits there, whom we wish to thank collectively. We thankalso FrancescoGoletti for providing some of the data used, without implicating.Helpful comments by an anonymous refereeare gratefullyacknowledged, without implicating.

from their privileged relationship with political


ratherthan acting vis-at-vis the private sector.In the public enterprises, the government fixes the wage schedule, andit fixes also the minimum wages for the private sector.The government intervenesin most industrial disputes, in all the sectors.Inthe privatesector, collective bargaining takes placeofficially at the firm level, butthe powerfulSKOP, the mainfederationof unions, deals directly with the government, and has eventually the determinantinfluence on wage settlements. The influenceof unions goes beyond the standardreachof industrialre- lations, and extends to the political arena.All the political parties, even the smallestones, exertsome controlovera tradeunion.The threemain political parties have their own tradeunion federation, which accounts for 64% of the unionizedworkers.The unionshave played an activerole in most major political eventsof this country, like the massivedemonstrations (hartals) that brought GeneralErshaddownin 1990, orthosewhich pushed thedemocratic- ally elected government of Begum KhaledaZia to resign in 1996. According to the World Bank (2001), an average of 21 full working days were lost annually due to hartalsin the 1980s, andan average of 47 full workingdays per year in the 1990s. This report estimatesthatabout5%of GDP is lost on average in the 1990s. Therefore, the behaviorof the Bangladeshi tradeunions seems quite dis- tant from the standard approach to the theory of trade unions, as surveyed for example by Oswald (1985) or by Booth (1995), which emphasizes the relationships between the union and the firmsthat employ its members.Its analysis is thus especially interesting, as a way to broadenour conception of the scope of tradeunion activity, witha view to take explicitly into accountits political economy dimension.The massive demonstrations organizedby the unions, called the hartals, areaimedat affecting the government muchmore thanthe firms.This type of political activismof the tradeunionsseems fairly widespread in developingcountries, as well as in some developedcountries, like Francefor example. One canremembertherole of Solidarnoscin Poland for bringing communism down. In Britain, the conflict between Margaret Thatcherandthe miners'unionin the late 1970s was a majorpoliticalevent, ratherthana simple industrial dispute. The aim of the presentpaper is to provide a simple theoreticalframework for analyzing this political dimensionof tradeunion activity, and to test its main implication in the case of Bangladesh. Our approach is thus drastic- ally differentfromthe one followed by Devarajan,Ghanem, andThierfelder (1997), who have analyzed the behaviorof tradeunions in Bangladesh and Indonesia, withinthe moreusual frameworkwhereunionsseek to affect the behaviorof the firms,ratherthanto influencethe government.Rama(1997) andRamaandTabellini(1998) also analyzemodels wherethe union affects


governmentpolicy. They provide a theoreticalmodel wherethe tradeunions lobby the government with a view to get some tariff protection for the firms, in orderto create a rentthat they can captureby imposing a wage increase. Then, unionized workers may get a redistributionof income in their favor. Similarly, Saint-Paul (2000) analyzes the political economy of labormarket institutionsas the result of redistributive pressures between rich and poor, skilled andunskilledlabor.In his model, the government, or politicians, are


labormarket. An additionalbenefitof our approach is to provide a theory wherea pos-

itive level of strike activity takes place in equilibrium. This easily observable phenomenon is usually absent from the standardmodels of trade unions.

lot of empirical studieshave been done for developed countries

Although a

(see Booth, 1995;Kennan,1985;Card,1990, andreferences therein),analyz- ing the durationof strikes,very little work has been done in the theoretical

literature aiming at explaining the occurrenceof strikes.While the paper on this topic (Ashenfelter and Johnson, 1969), and a small

of followers in the same vein (Farber,1978; Kennan, 1986), were in fact assuming thata strikewas going on, for analyzing its durationas a function of the concessions made by the firm, the most recent literaturehas takena game-theoreticapproach(e.g., Booth and Cressy, 1990; Hayes, 1984; and Tracy,1987), wherethe strikeresultsfrom asymmetric informationbetween the firm and the union. Mumford (1993) presents a recent survey of both the theoreticalandthe applied literatureon strike activity. Our approach also applies an imperfect information framework, in a game betweentheunionand



the representatives of the various conflicting interests present in the

the government.However, insteadof relying on asymmetric informationfor generating theoccurrenceof the socially sub-optimalconflict, ourmodelrests on commitmentfailure.We emphasize the role of the imperfectcredibility of


the workers, in providing the incentivefor the union to go on strike.Unlike

a large part of the literatureon strike activity, which is disconnectedfrom the morestandardliteratureon unionsand wage determination, we deal with these two issues in a unifiedframework.

a triadic one, as discussed

government,regarding its promises to adopt some policies thatbenefit

Moreover, our model must be understoodas

by Basu (2000), as the representative voter is a crucial actor in this game.

We assume thatthe latter applies a retrospectivevoting strategy, a la Barro- Ferejohn(Barro,1973; Ferejohn,1986). Then, the failureof the government to keep the unions quiet affects negatively its chances of being re-elected. This gives the urbanworkerssome leverage on the government that is not availableto the restof the population.Therefore,ouranalysismay be viewed as a contributionto the theory of theconflictbetweentownandcountry which


was emphasizedby Sahand Stiglitz(1992) as a centralthemein development theory. The next section presents the basic model, while Section 3 presents the assumed political setting. Section 4 describesthe equilibrium in case of im- perfect commitment.Section 5 presents the empiricalresults,analyzing the

political cycle effects on the

strikesin Bangladesh, we show the impact of forthcoming elections on their

occurrence.Section6 concludes.

occurrenceof strikes. Using monthly data on

2. The basic model

In this section, we analyze a variantof the standardutilitarianunion model discussed among others by Oswald (1982), assuming thatthere is a unique representative unionfor the whole urbanlabor force, with no internal migra-

tion. The only additionto this model donehereis the inclusionof the level of strike activity in thetool kitof theunion.Weassumethatthelattercan decide



government. Weassumethatthe lattercan spend its resourcesin two different ways. First, it can spend G units as employment-generatingexpenditures. This may be comprised both of personnelexpenditures, assumed to entail

a non-negativeemploymentmultiplier, and of other types of productive ex- penditures that enhance the productivity of labor in the private sector. For example, expenditures on basic infrastructures etc., have a positive impact on employment, for a given wage rate. It can also increase tariffs,boosting employment in the import-substitutionsector, at the cost of a higher price for imports(Rama, 1997). Second, it can spend the remaining resourceson governmentconsumption which affects its utility positively, with no posit- ive fallout for the workers. Hence, the political activismof tradeunions is here modeled as part of the conflict over the sharing of the resourcesof the government betweenthe formerandthe latter. Assumethattherateof employment m is a decreasing functionof the wage

rate w and an increasing functionof the level of governmentemployment-

generatingexpenditures G. This capturessimply the ability of the government

to affect the level of employmentby either hiring more public agents, or by

providing some productivity-enhancingpublicgoods, or

ing aggregatedemand, in a Keynesianway. Assume also thatall the workers

are identical, andthushavethe same probability of being

determined by:

put up a level of S units of strike activityper worker, with a unit cost of

utils to the representativeworker, with a view to affectthe behaviorof the

indirectly,by boost-

employed, whichis









We perform this analysis within the frameworkof the monopoly union model, wherethelatterdeterminesthe wage rate, whichis thentakenas given

by the firms, when making their hiring decisions. This assumption is more realistic, atleast in the Bangladeshicase, thanthe optimalbargaining solution discussed by McDonaldandSolow (1981). However, we assumethatthetradeunionmustchoose thelevel of the wage ratebeforethe actuallevel of employment-generatingpublic expenditures is known. Hence, the union selects the level of the wage rate w and of strike activity S on the basis of the expected level of governmentemployment-

generatingexpenditures Ge.This expected level of

on explicit promises made by the government, if the latteris credible, or on the analysisby theunionof the government'sbehavior, if it is subject to moral hazard.We will restrictthe analysis to a rational expectationsequilibrium, whereG = Ge.If the government cancommit credibly to any level of G, then

theunion regards Geas exogenous. Let U(w) (U'(w) > 0, U"(w) < 0) be the level of utility reached by a

workerwhen employed, while earning a wage rate w. As usual in this

of models, we assumethatthe worker gets a fixed level

when being unemployed.Then, if the lattertakes Ge as given, the utilitarian unionselects the couple {w, S} thatsolves the following problem:

expendituresmay be based



utility U < U(w)

maxw,sw =


andS >

st. m

mU(w) + (1 -

m)U -



m(w, Ge)




Notice thatthis set up assumesthatall the workersmust participate in the

strike,independently of their employmentstatus, andthusbearthe utility cost

This will



theirworkersat random.The first-ordercondition resulting from

setting the derivativeof W with respect to w equal to zero can be writtenas:

policy beforethe

be thecase if theunionsmustselect their {w,



U(w) -



m(w, Ge)


In (5), as in the other equations of this paper, a subscript is used to rep- resentthe partial derivativeof the functionwith respect to the corresponding variable.Condition(5) is pretty standard(e.g., Oswald, 1985), and Figure 1 illustratesthe resultingequilibrium, in the {w, m} space, in the case where











Figure 1. Union equilibrium with credible government.

G = Ge. A similar diagrammay be found in Oswald (1982). Equilibrium is

found where the union balancesthe benefit of the increased wage with the

increasedrisk of unemployment borne by the workers, at the tangencypoint

A between the union'sindifferencecurveW and the employment ratecurve

m(w, G). The new elementintroducedin this model is the effect of Ge. Carefulexaminationof equations(2)-(5) leads to the main point of this exercise, which can be statedas:

Proposition 1.If the government can credibly commit to a predetermined




employment-generatingpublic expenditures, then a no-strike

The reason for this result is that from (5) and (3) one can derive the preferred levels of w andm as functionsof Geonly:


w* = w*(Ge)




m* = m*(Ge).




W*(Ge, S) = maxwWI(3) -




it is straightforward to show that aW*/aS

non-negativity constraint (4) mustbe binding in suchan equilibrium. Moreover, under very mild conditions, one can show that:




0. Hence, the

w*(Ge)' > 0 and m*(Ge)' > 0.


However, it is quiteunlikely thatsuch a no-strike equilibrium will prevail, as the announcedlevel of employment-generatingpublic expenditures will not generally be credible, as we show below. Nevertheless, it is useful to characterizefurtherthis equilibrium, as a benchmark against whichto analyze the consequences of the government's lack of credibility. This is done in the next section.

3. The political setting

We basically model the game between the government and the union as a conflictover the sharing of fiscal resourcesbetween employment-generating expendituresG, thatthe union likes, andnon employment-generatingpublic expenditures, which only affect positively the government'sutility function. The latteris assumedlinearwith a unitaryslope, for the sake of simplicity. Whetherit remains in power or not, the government bears the cost of the employment-generatingexpenditures G. If it stays in power, the government enjoys a utility level Om- G. The positive effect of employment on the gov- ernment'sobjective function (0 > 0) simply reflectsthe fact thatthe higher

the level of employment, the higher is the resulting level of economic activity,

and thus, for a given taxing

government looses the benefitof this budget if it does notremainin power.

probability of the governmentstaying in power.

Beside exogenous factors, like the proximity of elections, etc., we assume that q is simply an increasing functionof the level of utility reached by the "politicallyrepresentativeagent"(PRA). The identity of the PRA obviously depends on the political regime. In a democracy, the PRA would be the median voter. In a non democratic regime, the PRA would probably be a representativeagent of the constituency that supports the ruler (e.g., ethnicor religious group,etc.). However, Alesina and Rodrik (1994) have suggested thatthe medianvoter assumption is a good approximation even in the case of non democratic regimes. In a full blown political economy model, the PRA shouldbe determined endogenously, butthiswouldbe a useless complication in the presentsetting. The only relevantinformationthatwe need hereis that the PRA, be it the medianvoteror some devoteeof the ruler,is not likely to be a unionmember,as only a smallfractionof the population is concernedby

capacity, the larger is the available budget. The

Let 0

1 be the





unionizationin developing countries.For example, in Bangladesh, themedian voteris certainly a rural person.However, this does not entail thatthe PRA is a peasant. On the contrary,political power belongs to a large extent to a proprietaryclass, using roughly the same mechanismsas the ones analyzed by Bardhan (1984) for neighboring India.The relevanceof this analysis for Bangladesh comes out clearly fromthe fieldwork presentedby Khan (1989). This can be regarded as exogenous for the type analysisperformed here. WeassumethatthePRA applies the retrospectivevoting rulewhose prop- ertieshavebeen analyzedby Barro (1973) and Ferejohn(1986). In the words of the latter "if the utility received at the end of the incumbent'sterm in office is high enough, he votes to returnthe incumbentto office; otherwise he removes the incumbentand gives the job to someone else." (Ferejohn, 1986: 35). Beside exogenous variables that need not be presentedexplicitly, we assume first that the utility level of the PRA is affected positively by

the employment-generatingpublic expenditures, because there are positive fallout of these expenditures, like greater access to some infrastructure, im- proved public services, etc. Moreover, the hiring of workers in the civil

service or in public enterprisesmay create a constituency in favor

incumbent government, and the positive effect of G on the PRA's utility function can also capture such an effect. This assumption is realistic for

developing countries, where a large part of the governmenthiring policy is aimed at "buying" some political support.Being seen unequivocally to "do something for employment" is almost everywhereregarded as a positive factorfor staying in office. Second, we assumethatthe level of strike activity, with all the disruption that it entails for economic activity in general, as well as the damages to property that may resultfromthedemonstrationsthat they often entail, affects negatively the utility level of the PRA. Let v(S, G)(vs < 0, vG > 0) be this utility function. Then, the probability of the governmentremaining in power

is 0 < q(v(S, G)) <

identity or preferences makes q differentiable, with q' > 0 and q" < 0 in the relevant range.However, in most of the following, we will use the short- handnotation:0 < q(S, G) < 1, with the partial derivativesof this function

definedas follows: qs < 0 and qG > 0, and the second derivativesdefined conformably.

of the

1. We assume that a stochasticelement in the PRA's

this game, fromthe government'spoint of view, is found

when assuming thatthe lattercan commit credibly to a pre-announced level

of employment-generatingpublicexpenditure G. Then, the governmentplays

The optimum of

first, and is in the position of a Stackelberg leader. The latter solves following problem:


maxGq(S, G)Om - G,


and w = w*(G) andS =

m = m(w, G)






The first-orderconditionfor this problem is:

0[qGm + q(mww* + mG)] = L.

This condition means that the government should take into accountnot

increasein employment-generatingpublic ex-

penditures on its probability of remaining in power

and on its budget via

the direct employment effect, but also the negative impact on the rate of employment due to the increasein the wage rate.

only the positive impact of an


4. The case of imperfect commitment

However, the prevail, if the

other methods, thatmake its commitment fully credible.This can be stated

as proposition 2:

optimum outcome characterizedabove is not very likely


government does not have access to institutional devices,


Proposition 2. Under imperfectcredibility of the government'scommitment, the optimum outcomeis not a Nash equilibrium.

In order to prove this proposition, we need to clarify the effect of the lack of credibility. When the government is not committed credibly to

a given level of employment-generatingpublic expenditures, this game cannot anymore be solved by backwardinduction as it was done above. Now, the union is playing first, while the government chooses the level of employment-generatingpublic expenditurestaking as given the equilibrium {w, S} pair chosen by the union. Then, the government solves the following problem at the second stage, given {w, S}:

maxGq(S,G)Om - G


m = m(w, G)



This resultsin the following first-ordercondition:

0[qGm + q mG]= 1.



Comparing(16) to (13), and assuming that the second-orderconditions for these two problems are satisfied, we find thatthe government will spend moreon employment-generatingpublicexpenditures in theNash equilibrium without commitmentthan at the optimum. The mechanismat work here is simply that, given the state of the labor market, as characterized by the

optimum {w,

the level of employment-generatingpublic expenditures in orderto increase employment and fiscal resources, and its probability of remaining in office.

It overestimatesthe marginalutility of by not taking into accountthe effect on

rationally takesthelatteras given at stage 2. However, theuniontakesits best-

an increase in public expenditures, employment via the wage rate, as it


pair, the government has an incentive to increase ex post

responsefunction as given at stage 1, anticipatingrationally the government's response to its own choice of the {w, S} pair, andthuschooses the latterwith this response in mind. Moreover, the second-orderconditioncan be writtenas:

0[qGGm+ 2qGmG+ q mGG] < 0.


This can be usedfor analyzing the government's reactionfunction:

G = G*(w, S).


Totallydifferentiating(16), and taking(17) into account,yields the following signs of the derivativesof the government's reactionfunction:

sign aG*/aw = sign {qG mw + q mGw}


sign aG*/aS = sign {qGs m + q SmG}.


These resultswill be discussedandusedbelow for characterizing the equi- libriumof the game. Notice thatthese signs depend on the secondderivatives mGw and qGs, which are playing herea key role.

4.1. Conditions for a positive equilibrium strikelevel

Let us now analyze stage 1 of the game, taking (18) into account. When the government cannot commit credibly, the union is the first mover (see

the Appendix for the detailed timing). Its problem is now to maximize (2), under (3), (4), and (18). We can apply again the Kuhnand Tucker theorem, with complementaryslacknessfor the non negativity constraint(4). This can

be used to characterizean interiorsolution, with S >

first-orderconditionrelativeto the level of strike activity reads:

0. In this case, the



(U(w) - U)mGGs = y.


This implies that Gs > 0. In view of (20), this can be stated formally as:

Proposition 3. Whenthe government lacks credibility,

in equilibrium is qGS >

which implies in particular qGs

explicitly the PRA'sutilityfunction, in termsof elasticities, as:

(i) a necessary conditionfor S > 0


> 0. This conditioncan be rewritten using




q"v(-) VGG





(ii) a sufficientconditionfor S > 0 in equilibrium is mGGs(U(w) - U) > y

when evaluatedat S = 0.

When expressed in terms of elasticities of the PRA's utility function, the necessary condition for S > 0 in equilibrium does not seem too restrictive, especially in the case where q" < 0 and vGs > 0 in the relevant range. The right-hand side of this conditionis the elasticity of the employment rateto the level of employment-generatingpublic expenditures. The left-handside is the sumof the two second-ordereffects thatan increase

in these expenditures has on the re-election probability:first, it reduces its marginalsensitivity to changes in the PRA's utility level, if the formeris concave, and second, it reduces the latter's sensitivity to the occurrence

of strikes. Then, this condition may

will be positive in equilibrium, under imperfect commitment, when the

government is quite efficient at shielding its re-election probability by


be interpreted as saying that strikes


a high level of

This is whatthe

strikeswill elicit a strongresponse

unionis looking for. As a checkon the plausibility of the necessary conditionstatedat proposi- tion 3, which may seem a bit complex, let us workout a simpleparameterized example. Assume that, over the relevant range, the function governing q

obeys the following specification, where 7, co and a are strictly positive parameters:

in favorof



thatthe neces-

sary condition presented at proposition 3 for S > 0 may be writtenin this

case as:

Thenit can be checked by simplecalculations,using (16),





if 1 >

too demanding if r > 1.If q < 1, it can only hold if q < r. If 1 <


a > 0, and

The intuition behind this result is simply that the union will find it worthwhileto exert pressure on the government to expand its employment- generatingexpenditures,by increasing the level of strike activity,only if it is worthwhilefor the government to respond in such a way to such an increase.

This occursif the expansion in employment-generatingexpenditures can ac-

tually offset thefall in the expectedmarginalbudgetarybenefit, due to thefall

in the probability of remaining in powerresulting fromthe strike.

qr0m. As both0 andm are probabilities, these conditionsdo not seem



inequality abovemust simply be reversed, andthusbecomes irrelevant, as

> 1 holds necessarily in this case.

4.2. Impact on the wage rate

Inorderto complete thecharacterizationof the strike-on equilibrium, we now analyze the level of the wage raterelativeto the one that prevails when the government does not lack credibility. This can be discussed on the basis of the otherfirst-ordercondition.






mGG* wG*v




Compared to (5), this expression shows thatthe wage rate can eitherbe

higher or lowerthanin theno-strike equilibrium,depending on the sign of the

From (19),

we know thatthe latter dependscrucially on the sign

of the m(-) function. Moreover, the effect will depend on the elasticity of U'(w). If the latteris larger than 1, the termon the left-handside of (22) is a decreasing functionof w, so thatthe wage rate will be higher than in the

no-strike equilibrium if G*, > 0. From (19), this occursif:

second termon the right-handside, i.e. mainly, on the sign of

of the cross G*w. derivative

qGmw+q mGw> 0.



implies in

particular that an

increase in



employment-generatingexpenditures reduces the marginal impact of the wage rate on the rate of employment, measured positively (i.e. makes it less elastic). This set of conditions seems ratherrealistic for developing

countries, where workerscan be expected to be stronglyrisk-averse, while

employment-generatingpublic expenditures can plausibly be assumedto re- duce the elasticity of the demandfor labor,especially if they imply a lot of

hiring in

the civil serviceandin various parastatals.

Figure 2 providesan intuitiveinterpretationof this equilibrium. Point A

is the no-strikeequilibrium,as describedin Figure 1. Point E is the new



m(w, G*(w,S))


LE AI W m(w, G) m m* 1
m(w, G)

Figure 2. Equilibrium with commitmentfailure.

equilibrium obtainedin this section when the government cannot commit

credibly. It is located on the m(-)

employment-generatingpublic expenditures, at a point of tangency between

an indifferencecurveof the union anda

rate of employment


represented as positive in Figure2, while it is

thatthe positive effects on employment of theincreasein publicexpenditures, entailedboth by the increasein w and in S, offsets the negative effect of w,


curve corresponding to a higher level of

curve describing the response of the

in w, taking into account the endogenous

to a


employment-generatingpublicexpenditures. The impact on m*is

in fact ambiguous. Thisassumes


However, we could also get

the same sign of the comparative statics of

the wage ratebetweenthe no-strikeandthe strike-on equilibriaby reversing,

roughlyspeaking, Defining e =

staticsexercisewith respect to the wage rateas:

these two sufficientconditions.

-wU"/U', we can summarizetheresultsof this comparative

Proposition 4. The wage rate is higher in the

strike-on equilibrium than in


Hence, beside the case sketched above, we could also predict that the

equilibrium than in the no- degree of risk aversion (e <

wage rate would be higher in the strike-on strike equilibrium if (i) the workershad a low -


did not affect significantly the slope of the employment-rate function.

2w/(U(w) -

U)), and (ii) employment-generatingpublic expenditures

5. Empirical test on Bangladeshi strike data

We now turn our attentionto the empirical test of the main proposition resulting from this theoretical framework,i.e., that the incidence of strike activity should go up when the degree of credibility of the governmentgoes down. In Salmon (1998), a similartest is performed with the wage ratesof the unionized sectors. The main empirical issue is to find some proxies for capturing the credibility of the government's commitmentnot to give in to the pressure of the unions.Our priorassumption is thatthe credibility of the government is weak just before the elections, while it is stronger after the

result is known. The government is tempted to have a more expansionary policy when election time is coming, as this may increase its re-election probability, while its costs in termsof inflationor budgetary restrictionswill

only come

up in the future, and might be borneby the future government,

should the election turnadversely. In the termsof the theoreticalmodel of the previous sections, the proximity of election time may be viewed as a downwardshift of the q(-) function, for any value of v(-), as this reduces exogenously thecost of bringing the government down. Hence, we expect the probability of a strike occurring to be affected positivelyby the approach of an election. Moreover,during the first part of the period covered by our data, therewas a militarydictatorship, underGeneralErshad.It can be argued that

this type of government is potentially morecrediblefor resisting thedemands of the urbancrowd than the more shaky democraticones that followed, as

it is less influenced by marchesand demonstrations, even if their probable

impact is negative on the median voter. After all, this government seized power by a coup, without much democratic scruple.Nevertheless, General Ershadwas eventuallybrought down by popularunrest, with massive and repeated marchesand demonstrations, with an unusualnumberof marchers. These were bloodless events, but involving a momentousmobilization.It is thus worthwhileto test whetherhis government was more or less credible, in its commitmentto resist popularpressure thanthe democratically elected ones thatfollowed. The data on strikeshave been collected by FrancescoGoletti in the re- gional press of Bangladesh between January1988 and December 1992, for the industrialand the transportsectors. We estimate several equations us-


ing these data, which are presented at Table2. They are based on a probit analysis, aiming at identifying the main determinantsof the probability of at least one strike occurring somewherein the country on any given month.

The theoretical analysis presentedabove, and especially propositions 1 and 3, predicts thatstrikesshould mainly be observedwhen the government has

constructeda dummy variable

a low credibility. For this purpose, we have

taking the value I when at least one strike occurredin any districtin the

country sometimes during a month. We have performed these estimations

for the whole sample, and then by separating the industrialsector from the transport sector.Both the necessary andthe sufficientconditions presented at proposition 3 involve a potentiallysector-specific derivative mG, measuring the impact of publicexpenditures on the probability of beingemployed. There

is no reason why we should expect it to be the samein boththe industrialand

the transport sectors. The lagged values of the numberof strikes in either sector is included in the equations, in orderto take care of the inertiathat characterizesthese series, as can be expected at the monthlyfrequency. Then come the variablesrelated to the political cycle. Three elections took place at the national level and five elections took place at the local


type, and provides the numberof elections aretakeninto account by

four preceding months.We

our election variablein the equationspresented at Tables 2 and 3, without affectingdrastically the results.Our samplestops in December 1992, butwe havetakenintoaccountthemonths preceding the January 1993elections.The


level of local governments, or at a lower level, namely the 'Upzila Parishad'. We have also includeda dummy variablefor the 'hartals' (demonstrations) called at the nationallevel by the unionsandthe politicalparties. We use as control variablesthe change in the quantity of money, with a view to controlfor the impact of thebusiness cycle, with an expectedpositive sign, as well as the change in theinformalsectorreal wage, in orderto capture the changes in the relative wage in the formalandthe informal sector, while avoiding to include the real wage in the formal sector, which is likely to be

endogenous. This should capture thecost

We thus expect morestrikeswhen the informalsector wage goes up. We also

take into account natural disasters, like cyclones and

floods, unfortunately

very frequent in this country. We expect them to have a negativeimpact on the occurrenceof strikes, as publicexpenditures arethen quite naturally tied up for relief operations,and are unlikely to be divertedfor employment- generatingpublic actions, whateverthe unions activity.Moreover,at such

during our period of observation.Table 1 lists them, distinguishedby

voters concerned by each of them. The a dummy variable indicating the threeor

have experimented with differentdefinitionsof

elections are takingplace eitherat the 'UnionParishad' level, i.e., atthe

of loosing ajob in theformalsector.


Table1. List of elections in our sampleperiod


10 February 1988

3 March1988


January 1989

14-25 March1990


September 1991



February 1991



February 1992

January 1993


Union Parishad



Upzila Parishads



Union Parishad


Source: Bangladesh ElectionCommission.

Numberof voters

47 millions

49,8 millions



Local body election National

Local body election



Local body election






49 millions

Local body election



body election

times, the impact of the strikes on the PRA would not be in favor of the unions.

The resultsare presented in Table2. The firstthreecolumnsdo not


ate the industrialand transport sectors.The nextthreecolumns only concern the industrial sector, while the resultsfor the transport sectorare presented in

the last threecolumns.The fit is significantly betterfor the industrialsector

thanfor the transport sector.Withineach group of three columns, the results differ by the definitionused for the election variable.In column (1), the in-

cluded variableis a dummy which takes the

before any election listed in Table1. In column (2)

elections are distinguished, and are still indicated by a dummy taking the value 1 overthe four preceding months.In column (3), the nationalandlocal elections are again separated, but now indicated by a dummy for the three preceding months only. The results suggest that only the nationalelections matterat this level of analysis, whether they are indicated by the three or four preceding months.The resultsfor the industrialsector are presented in the next three columns, which show qualitatively similarresultsthanin the preceding threecolumns. All the estimated coefficients, which measurethe marginal effects of the explanatoryvariables, are larger in size, but of the same sign and level of significance,except for one. Now, we find that the distinctionbetween local and nationalelections does not matter. Hence, the impact of the electoral periods is positive and significant for the industrial sector, as predicted, The resultsfor the transport sectorare noticeably differ- ent. In particular, theelections areeither insignificant, orhavethe wrongsign when measuredas the fourmonths preceding the nationalelections.This is a

value 1 during the four months

the nationalandthe local


puzzlingresult, which might be a reflectionof thefactthatwhentheothersare

on strike,they need more transport services to go and participate in marches

anddemonstrations. They would probably be accusedof breaking the

if they were stopping their activity at these times. Although the militarydictatorshipperiod witnesseda lowerlevel of strike

activity than the subsequent democratic one, this does not seem to be due

to a higher level

dummy variableis never significant in any of the columns. The impact of

nationaldemonstrationsis positive and significant for the industrial sector,

and the sum of the two, significantly so. This is

transportation services are required when large demonstrationsare organized

in Dhaka, the capitalcity, providing the transporter with an incentiveto re-

mainactiveat thattime.

disastersconformswith the theory in the industrial