Efficiency Wages and Labor Markets: A Study on Per-Capita Consumption and Productivity, Study notes of Development Economics

Download Efficiency Wages and Labor Markets: A Study on Per-Capita Consumption and Productivity and more Study notes Development Economics in PDF only on Docsity! “Surplus” Labor, Underemployment and Unemployment What happens to rural output when laborers leave for the industrial sector? “Surplus Labor” models assumed no effect - a costless shift How is this possible? Two routes: 1. Family labor-supply behavior (A. Lewis, A. Sen) 2. Unemployment and rigid wages (Mirrlees, Stiglitz) Distinguish: number of workers (N), time worked per worker (h), effort per time worked (8) Labor-Supply Model 1: Autarchy (Sen, 1966) Each household has N workers and n total members (dependents) Owns a productive asset (land) A Technology: X = F(L, A), where L=Nh Family welfare function: U = U(c, l), where c = X/n FONC: Uc/Ul = (N/n)FL Removal of laborer leaves mp unchanged if mrs invariant: fully compensatory family labor supply Labor-supply Model 2: Labor market (family, hired labor perfect substitutes: separability) c* = B/n + Wh(N/n): per-capita family consumption, where B = profits from cultivation Does the removal of a worker (reduction in N) lead to an increase in the labor supply of other family members? FONC: Ul/Uc = W(N/n) Slutsky equation in elasticity form: 0h,N = 0h,w c(n - N)/N - N(Wh - c*),h,B, where 0h,N = elasticity of per-capita work wrt a change in N 0h,w c = compensated elasticity of per-capita work wrt the wage (own price elasticity) > 0 ,h,B = income elasticity of work < 0 A. Price effect: decreases work by family members because removing a laborer increases the family dependency ratio (n - N)/N more sharing = tax on work hours But, what happens to the unemployed? Do they starve? A. Egalitarian Sharing model: employed give some food to the unemployed then, we = F(N8(we))/n = c*, where n = total population we < w*, per-capita consumption of workers is lower if share so rural output increases when a laborer is removed! B. Alternative sources of income Some workers have their own land Per-capita consumption of landed worker with no dependents C = w* + B/N assuming excess supply of landless Because productivity varies with consumption, employers will pay different wages to different workers Will employers pay more to the higher-consumption landed or less? 1. Monopsony case: one employer Can employ both landed N1 and landless N0 laborers at different time wages, w1 and w0 max F(N18(w1 + B/N) + N08(w0)) - N1w1 - N0w0 FONC: 8’(w0) = 8’(w1 + B/N) 1. Wage rates selected to equalize consumption of both types of workers, so w0 = w* > w1 2. But, if there is unemployment, always the landless 3. Time wages vary inversely with landholdings 4. Land reform: providing land to the landless lowers their wage (but not their consumption)! 2. Competitive case: Given total land and the number of workers, there is a marginal product of the total number of efficiency units In a competitive economy, that sets the wage per efficiency unit, which must be paid to every worker 1. Landless still paid w*, have lowest time wage, and are unemployed 2. Time wages vary positively with wealth: w1 > w* What is the Evidence for Nutrition-Based Efficiency Wages? 1. Test implications: wage diversity Do time wages vary by wealth or number of dependents? Causation issue: high wages lead to more wealth; more dependents(?) 2. Test assumptions: A. Increased consumption increases productivity (Strauss) Causation issue B. Increased consumption increases wages Causation issue Observability issue: can employers monitor worker food consumption? The collateral premium: Assume credit is constrained and rationed Consider farmer with initial wealth ù deciding on how much land to buy. The farmer plans to buy the land, invest in it, cultivate and then sell. How much land should be bought to maximize return V? 1. No title nt nt nt nt nt nt nt ntTerminal value of investment = V = F(A , K , Z ) + P A - (1 + c)mP A nt nt nt nt nt ntConstraints: P A + K = ù (long-term) Z = m P A (short-term) nt nt ntOptimal terminal value (given optimal A chosen) = V * = ùdF/dK 2. Titling exists, allowing land to be used as collateral for long-term investment t t t t t t 1 1 t t 2 2 t t 1 t t t tV = F(A , K , Z ) + P A - (1 + i )s P A - (1 + i )s (P A - s P A ) - (1 + c)mP A t t t 1 t t t 2 t t 1 t t t tConstraints: P A + K = ù + s P A Z = s (P A - s P A ) + m P A nt t tOptimal terminal value (given optimal A chosen) = V * = ùdF/dK 1 2 t ntAs long as s , s > 0, V * > V * so everyone would want titled land In market where both types of land are available, to make investors indifferent, Pt > Pnt Can show that output/acre higher on titled land because more K and Z per acre Two reasons farmers more productive on titled land: 1. Because P is higher, can purchase with credit more variable inputs per acre 2. Land can be used as collateral for K and Z (s1, s2 > 0), more investment Evidence on collateral premium (Feder and Feeney, World Bank Economic Review, 1991): Survey of 1300 farms in Thailand where titling program incomplete Collected information on 17 attributes of land (soil quality, location, improvements) Price information: sales, owner assessments, village-elder assessments For given attributes, found titled land sold for 23 to 84% more Potential problem: what determined which lands were titled first by government? Optimal ( output-maximizing) to title best land first - hope that measured characteristics account for everything government knew in allocating titles Evidence on investment intensity (Timoth Besley, Journal of Political Economy, 1997) Survey of 634 households with 1568 individual plots of land in Ghana Amazingly, different plots have different rights: sell, rent, gift, mortgage, pledge, bequest He created an index of rights - sum (0-6 range) Having plot-specific variation enables distinguishing tow hypotheses about land rights: 1. If all that matters is ability to use land as collateral, then whether a household has any collaterizable plots will affect all plots equally: which specific plot has collateral value does not matter: why? 2. If more rights reduce transaction costs of selling, then plot-specific rights matter He looked at whether a plot had been improved since plot was acquired Found: more rights, more likely to have had an investment at plot level for the same farm Problem: again, why do rights vary across plots? Uses instruments to predict How long plot owned, how the plot was acquired, whether household engaged in litigation: are these really independent of unmeasured plot quality? 0.2 Collateral Return random θ(k) increasing & concave with prob q, 0 otherwise. Land as collateral, taken by lender in default at cost φ(Rt+1) with φ0 < 0 (this is the whole story). Social custom forces lender to give borrower u. Borrower wealth is y. So max b,k u(y + bt − kt) + qu(θ(kt))− rtbt) + (1− q)u u0 − rqu0 = 0 −u0 + qu0θ0 = 0 imply rt = θ 0() (surprise). So obviously we’re going to see investment decreasing with r. Since better collateral lets the bank charge a lower r, we’re done. Borrowing is bt = g(rt) with g0 < 0. Lender profits are Π(rt, Rt+1) = q(rt−ρ)g(rt)+(1−q) h p(θ0−1(rt − φ(Rt+1)− x− ρg(rt) i Assume there’s a fixed cost of lending C, and that Π(ρ,Rt+1) < C (fine, as long as p is low). But there is competition, so r∗ = min{Π(r,Rt+1)|Π(r,Rt+1) = C} > ρ. So Π1 > 0 at r∗. And we know Π2 > 0 as well. Hence dr ∗ dRt+1 < 0. Note: this is a connection at the household level. 0.3 Gains from Trade Better rights enable you to capture gains from investment, even if you become less productive: Linear tech: MP of capital is θ ∼ f(θ). Outside value ω ∼ g(ω). (indepen- dent). Cost of trade µ(Rt+1)kt with µ0 < 0. Nash bargaining for price: p∗ = max p (p− (µ+ θ)kt) (ωk − p) implies p∗ = [θ + ω − µ] k/2. So V (k,Rt+1) = kE (max(θ, [θ + ω − µ] /2) ∂V ∂k = Z " Rω−µ(Rt+1){[θ + ω − µ] /2}f(θ)dθ+R ω−µ(Rt+1) θf(θ)dθ # g(ω)dω and this is going to be increasing in R: ∂2V ∂k∂Rt+1 = Z "Z ω−µ(Rt+1)−1 2 µ0(Rt+1)f(θ)dθ # g (w) dω > 0 226 TABLE 3 Wassa: INVESTMENTS IN Tree Pantin (N = 1,074) Rights with Rights without Uninstrumented Instrumented Approval ‘Approval a) (2) (3) (4) Rights with approval 03 (1.93) Rights without approval 02 (1.56) Number of past tree plantings 19 (4.34) Average age =.00 =.01 —.00 (40) (26) (52) Value of durables 00 -.00 -.00 (1.80) (49) (1.98) Livestock value =.00 —.00 00 (2.05) (2.48) (2.77) Formal education of head 01 18 00 (28) (1.68) (04) Women -A1 —.08 04 (78) (2.66) (1.58) Men 01 = 02 03 (2.03) (1.67) (1.32) Rooms =.00 08 ~.03 (45) (3.93) (1.34) Distance from house 01 =.05 01 (44) (1.39) (35) $36 Soil very fertile 07 (1.38) Soil fertile -.07 (1.43) Field area —.02 (12.55) Field purchased Field allocated Field appropriated Field gifted No title deed Number of years owned Trees existing at ime of acquisition Ever litigated on field Test of overidentification restriction (p-value) Village dummy variables Yes F-test on significance of _ instruments (#-value) 2 35 09 (1.52) -10 (1.68) = 02 (12.01) 90 Yes 33 38 Nort.—Absolure values of ¢-statistics are in parentheses, The omitied classification in the mode of acquisition is inheritance and in soil type is poor. b26 TABLE 4 Wassa: INVESTMENTS IN TREE PLanTiNG (with Household Dummy Variables) (N = 1,074) Rights with Rights without Uninstrumented Instrumented Approval ‘Approval (2) (3) @ Rights with approval 07 (1.77) Rights without approval 05 (86) Number of past tree plantings 22 (8.52) Soil very fertile .08 8 -.10 (70) (4.38) (1.29) Soil fertile 05 26 —.07 (49) (2.16) Q.01) Distance from house 00 —.06 —.02 (19) (3.64) (1.78) Field area —.03 00 =.00 (12.78) (1.04) (24) 626 Existing fencing a) (1.12) Existing drainage 56 (14.72) Existing trees — 50 (5.66) Existing access road 59 (1.65) Existing continuous = .02 manuring (36) Existing land —.12 excavation (8.37) Existing irrigation 00 (01) Existing mulching -11 (2.55) Existing shallot = 08 beds 2.14) Field area =11 (73) Distance from house —.03 (1.64) Village dummy variables Yes R 75 Norx.—Absolute values of s-statistics are in parentheses. TABLE 6 AnLoGa: RicHts REGRESSIONS (N = 494) Rights with Rights without Approval ‘Approval Average age 00 —.02 (36) (1.76) Value of durables 00 (1.76) Livestock value 00 (2.34) Formal education of head “14 (1.04) Women 19 (4.93) Men =.05 (1.12) Rooms = 01 (51) Existing fencing 80 (4 Existing drainage 147 (7.55) Existing trees 1.05 (2.12) Existing access road 1.45 (74) Existing continuous manuring 82 B41) Existing land excavation 80 (4.04) Existing irrigation 06 (0.27) Existing mulching 86 3.82) Existing shallot beds 07 (34) Field area -247 (2.97) Distance from house 21 2.34) Land inherited = 97 (2.27) Number of years owned 04 (6.65) No title deed =A (35) Ever litigated on field 124 (66) Village dummy variables Yes F-test on significance of instruments 00 __ (p-value) R 30 84 Nore.—Absolute values of é-statistis are in parentheses The Problem of Eliciting Effort: Comparisons of Contractual Arrangements 1. Time wage: no reward for effort; requires supervisor 2. Piece-rate wage: payment dependent on performance useful when relevant worker output readily measurable requires counter, output measurement why not always used? weeding, harvesting 3. Own cultivation: individual is residual claimant receives all rewards for effort (unless family sharing) 4. Output (profit) sharing with workers: individual has some incentive to provide effort Optimal choice of contractual terms and land to rent by landlord: -NU1{F - (1 - ")F1(dL/d") - [(1 - ")F2 - f2](dT/d")u*} = 0 -NU1 + NU1{(1 - ")F1(dL/d$) - [(1 - ")F2 - f2](dT/d$)u*} = 0 U1[(1 - ")F - Tf2] + U1$ = 0 Which collapses to (1 - ")F1 2/F11 = 0 Optimal solution: " = 1, pure rent contract! Thus, inability to specify labor effort in contract, monitor labor not a justification for the existence of sharecropping So, if there is sharecropping, what is going on? 1. Exploitation - so ban (as in India) 2. Sharecropping contract helpful if both unmonitorable labor and no insurance against risk 3. Helps if there is unmonitorable labor and some other input for which there is no market under some conditions Case 2: Unmonitorable labor and risk Expected income of the tenant under share tenancy: "F = LF1 Expected income of the tenant under fixed rent (bears all risk): F - $ = LF1 - $ [1 - Eu1/Eu12] Thus, the tenant pays more on average for a share contract than for a fixed rent contract; benefit is risk reduction (sharing) - risk makes income lower Case 3: Exchange of non-marketed inputs (Eswaran and Kotwal, AER 1984) 1. Added two non-marketed inputs: supervision time (s and S) and management time (m and M), where labor effort L = L(s, T), L1 > 0, L2 > 0 2. Assume: A. Landlords have comparative advantage in management - skill in allocating inputs (m > M) B. Tenants have comparative advantage in supervision - family labor (S > s) Now compare alternative contracts, modes of using outside labor: 1. Wage contract: landlord divides time between supervision and management; tenant supplies no supervision or management (mw>0, sw>0, Mw=Sw=0) 2. Fixed rent contract: landlord supplies no management time or supervision; tenant divides time between both (M fr, Sfr, mfr=sfr=0) 3. Share tenancy contract: landlord supplies some management and tenant supplies some supervision, but less than full amount (ms < >mw, Ss <>Sfr) Thus, share contract may be superior - get some superior management, superior supervision, share two individuals skills rather than one, may overcome disincentive effect of sharing (") Suggests when share contracts are observed " = ½ most likely: as " deviates from ½, disincentive effects for either input: if " increases, less management skill supplied; so fixed rental dominates; if "<½, less supervision supplied, so wage contract dominates Consider two cases distinguished by non-repayment regime: 1. Harsh punishment if does not repay, so tenant never wants to have to reneg: chooses L such that "F(L)2min = B* where 2min = worst-ever outcome then dL/dB* = ("FL2min) -1 > 0 2. Limited liability (bankruptcy): C = max {"2F(L) - B*, C*}, where C* = guaranteed minimum Critical value of 2b when bankruptcy occurs: solve " 2bF(L) - B* = C*, so 2b = (C* + B*)/"F(L) See higher is B*, higher is critical value of 2, so bankruptcy more likely - less effort But, if not bankrupt, higher B* tends to increase effort (higher B* lowers C, so effort increases to compensate The net effect is ambiguous - but B* will affect what the tenant does thus affecting the landlord’s income; the landlord thus cares about borrowing behavior of tenant What is the optimal loan for the landlord to make? Assume " and T are set, and that 2=1 The landlord’s problem: max {(1 - ")F(L) + (1+i - (1+D))B} choosing B, B* but again, if competition: V(B) + EU("2F(L) - B*, L) = u* (reservation utility) Solution is: VC/EUC = 1 + D/((1 - ")F(L)dL/dB* + 1) Market solution is: VC/EUC = 1 + D Thus, if more borrowing induces the tenant to work harder, the landlord provides a larger loan than the tenant would otherwise have obtained: provides more consumption in first period Other instruments of landlord: 1. Production loan (put B in the production function): 3 effects 2. Subsidized lunch for workers 3. Buy product at set price (marketing of the tenant’s product by the landlord) Marketing by the Landlord: Does it Represent Interlinking? The landlord is the exclusive purchaser of the output of the tenant at a pre-specified price P, Assume no uncertainty The tenant’s new problem: max U(C, L) where C = P,aF(L, Z)- Z (unit price for Z) FONC: F, =-U,/U,P,a F, = 1/P,0 See that tenant’s decisions fully determined by the control P,a Thus, the landlord could pay the market price, and still have the same degree of control over the tenant’s effort - not an additional instrument of control Of course, if the government authority fixes « above the equilibrium a, then P, can be set lower than the market price while still honoring the utility constraint But, hours are not effort! Hours are observable - is this really a test? How can you assess whether workers are supplying effort, for given hours, without actually observing them? At great expense! (and some cleverness) Idea: 1. Effort (in agriculture) = energy expenditure 2. Exploit balance function: change in body mass reflects energy expenditure and intake Hit = (0 + (HHit-1 + (ZZit + (ccit + Gj8jiLjit + :i + >it, where Hit = body mass for i at time t 8jit = effort under payment scheme j by i in t cit = calorie consumption by i in t Ljit = time worked payment scheme j by i in t Says effect of work time in activity j on body mass at the end of the period of work depends on effort level 8jit If different payment schemes elicit different effort levels, then time worked under different payment schemes should differentially affect loss of body mass (for given consumption levels, illness episodes). What is ranking among own cultivation, share tenancy, piece- rate agricultural work, time-wage work? Why not hire all laborers using piece rate payments? Expense: need data on individual consumption, body mass, work time of workers under different payment schemes in multiple periods Survey of 448 households in four rounds of three months in Bukidnon, Philippines Estimation issues: 1. health, or shocks to health, may affect choice of activity 2. consumption may be affected by health, or shocks to health Method: 1. Compare same individual’s change in body mass under different payment regimes 2. Use instruments: prices, wealth, education Results Table 1 Descriptive Statistics Men Women (N=458) (N=494) Mean Std, Dev. Mean Std. dev. Fraction of people working during study period in: piece-rate 423 2495 +249 «428 time-wage -58l «494 «194 +396 on-farm +530 +500 «367 - 482 non-agric. -366 +482 +251 434 Fraction of total days worked in: piece-rate 052 156 014 065 time-wage 125 ~260 .020 091 on-farm +128 +260 +042 +151 non-agric. 157 -331 .126 +313 Calories 2609.6 990.2 2160.0 861.5 Body-mass index (kg/meter?) 20.17 1.963 20.64 2.815 Iliness (fraction days) .033 144 045 «175 Height (cm) 160.7 6.163 150.1 5.491 Education (years completed) 5.903 5.104 6.581 6.760 Land owned (hectares) 2.253 3.830 1.953 3.458 — security of tenure encourages tenants to invest more. Pretty standard model of sharecropping. m-outside option Y ∈ {YH, YL} and prob(Y = YH) = e c(e) = 12ce 2 with c > 1. e is not observed. Y is observed and contractible. Limited liability: w > 0 is tenant wealth, −w is minimum payment. Everyone risk neutral. Big technical simplification - everything is independent of history, except that tenancy status matters. The equilibrium will be: h and φ for payment to tenant and prob of keeping him if Y = YH = 1, and l and ψ for when Y = YL = 0. • First Best max e e− ce 2 2 implies 1c = e(< 1) • no eviction max e,h,l π = e− [eh+ (1− e)l] subject to h ≥ −(1 +w); l ≥ −w (LLC) eh+ (1− e)l− 1 2 ce2 ≥ m (PC) e = argmax e eh+ (1− e)l− 1 2 ce2 (ICC) Obviously, h > l (or from (ICC) e = 0. So h > −(1 + w). First best e < 1, so let’s ignore corner and use first order approach for ICC: e = h− l c ∈ (0, 1) Put this into π: π = h− l c − (h− l) 2 c − l — Let’s first consider values of m such that the PC constraint doesn’t bind, but LLC does. So l = −w. Substitute this into π and then choose h− l to max π: h− l = 1 2 ⇒ e = 1 2c What’s up? standard tradeoff: provide incentives vs. extract surplus. Can’t make l low enough because of LLC. Want to raise h to generate • Eviction We see that with low enough outside option, tenants get rents due to LLC. Let V̄ be expected lifetime u starting next period for a tenant. M ≡ m1−δ for non-tenant. consider case where LLC binds: l = −w V0 = maxe eh− (1− e)w − 1 2 ce2 +δ [φe+ (1− e)ψ] h V̄ −M i + δM The ICC now has h+w c + δ(V̄ −M) c = e so we just made achieving e cheaper by δ(V̄−M)c . That’s obviously the eviction effect on efficiency. Equilibrium has V̄ = V0, so simplifying we have V̄ −M = eh− (1− e)w − 1 2ce 2 −m 1− δe and plugging in the new ICC we have V̄ −M = 1 2 ce2 − w −m > 0 (*) if there are effective threats. Recall that landlord profits are e−eh−(1− e)l. Substitute in the LLC l = −w and the ICC h = ce−w− δ(V̄ −M) to get the landlord’s problem: max e (1− ce+ δ(V̄ −M))e+w to find e = 1 + δ(V̄ −M) 2c (**) And we have a picture increase in m shifts ** down. OB could reduce efficiency. Two main empirical approaches: 1. Diffs in diffs for West Bengal and Bangladesh 2. looking across districts in WB DD: This is really the whole story. Rest is just checking robustness. TABLE 2 DIFFERENCE-IN- DIFFERENCE Movets oF Loc or Rice YIELD PER HEc F, (1969-93) DIFFERENCE a (1969-78) 1969-93 Excluding 1981-82 (1) (2) (3) West Bengal 004 (=1) (17) West Bengal x we —.01 (1979-83)* (.38) West Bengal x 058* (1984-88) (2.00) West Bengal x 05* (1988-93) (1.78) District fixed effects statistic wee 44.55 42.61 Year fixed ef fects statistic 4.268** 31.81 %** Re 12 81 Sample size 256. 659 Nove. —estatisties are in parentheses. A few specification checks: Questions: Errecr or RecistratTion TABLE 6 on THE Loc or Rick Yep iv West Bencar, 1979-87 (N= 126) Model Model Model Model Model Model la Ib 2a 2b 3a Sb Sharecropper 46° 46% AU registration (2.71) (2.73) 4 2.54) Log(real wages) i (1.07) Log (price of —11 rice) (-.98) Log (rainfall) Log (public irrigation) (2.19) Log (roads) 08 (47) HYV share of Gi 49 rice area (2.14) ) (1.34) Fstatistic: South x year yes yes Left Front x year yes yes Sharecropping x year yes yes District fixed effects 29.34" 5.98%" Year fixed effects 20.208 11.20 R 89 90 Nore. —¢statistics are in parentheses. What About Land Ownership - Why Are There No Land Sales? All tenancy models take landownership as given. Any empirical findings or models that suggests that large (small) farms are inefficient have to explain why farmers do not sell their land - e.g., if big farms are less profitable, a large farmer could break up his holdings and sell them - the value of the sum of the parts > total India: 1. In 1970-71 only 1.5% of farm households sold land 2. Between 1971 and 1982, only 2% of households experiencing a reduction in landholdings sold their land (most from household splits) 3. Not only no sales, but male family members stay with family land: A. In 1970-71, 62% of farm households “intergenerationally extended”: related adults of two generations (father-sons) coresiding and farming together on the family land B. ICRISAT survey: 70% But, not “horizontally extended”: only 7.5% of farm households had adult siblings co-residing Hypothesis: Extended household structure, dearth of land sales rooted in: 1. Technological stagnation (traditional agriculture) 2. Riskiness of agricultural production associated with weather variability 3. Unmeasurable variation in land plots - land heterogeneity Manifestations of optimal implicit contract between generations that maximizes gains from specific experience - experience with varieties of weather conditions specific to owned plots of land Theory: simple overlapping generations model A. Biology: everyone lives 3 periods: 1. As a child laborer on the family farm owning A acres of land 2. As an adult laborer for two periods earning a wage W B. Land market: no land owned in first two periods but can buy land out of savings in period 2 C. Technology: an adult’s (child’s) productivity on a land plot worked one period increases by D ($D) on that plot (return to specific experience) Mover’s consumption path (person who leaves the family farm): Period 2: C2 = W - pA, where p = market price of land Period 3: C3 = W + (D + r +p)A, where r = return on land Return on selling land rr = (r + D)/p 0 5 10 15 20 25 30 35 Elder Alone Younger Alone Old+Younger Percent Increase in Bad-Weather Farm Profits, by Household Structure: NCAER-ARIS Farmers, 1968-69 – 1970-71 0 2 4 6 8 10 12 14 16 18 5 Years 25 Years 40 Years Percent Fall in Real Farm Profits from a 1 mm per-day Decline in Daily Rainfall, by Farmer’s Years of Experience on Unirrigated Land: ICRISAT Farmers, 1975-83