Do agricultural commodity prices asymmetrically affect the performance of value-added agriculture? Evidence from Pakistan using a NARDL model

Pakistan’s agriculture sector contributes 22.7% to GDP, employs 37.4% of the labor force, and underpins exports and food security. Despite extensive irrigation and major crops like wheat, cotton, sugarcane, and rice, recent declines in cultivated area and production, along with socio-environmental and political challenges, have hampered development. Commodity price shocks are central to growth dynamics in many developing economies; asymmetric responses to positive and negative price shocks may yield different growth outcomes. The study investigates whether positive and negative shocks in agricultural product prices, alongside credit disbursement and labor force, affect agriculture, forestry and fishing value added (AVA) in Pakistan. Addressing a gap where prior work emphasized linear models and aggregate GDP effects, this study applies a nonlinear approach (NARDL) to answer: to what extent do negative and positive price shocks affect AVA in Pakistan?

Agricultural product prices are highly volatile due to environmental, supply-demand, and market factors, including energy prices and exchange rates. Prior studies offer mixed evidence on commodity price shocks and growth: some find short-run gains from price increases (Deaton and Miller, 1996) but limited long-run per capita income effects (Collier and Gunning, 1999), and harmful effects of price inefficiencies on growth in some regions (Vink, 2012). Volatility raises risks for farmers and can slow growth (Blattman et al., 2007). Methodological limitations—often linear models ignoring potential nonlinear/asymmetric effects—may underlie disagreements on long-run impacts. Recent research emphasizes distinguishing positive and negative shocks, with possible irreversible consequences from negative shocks (Kinda et al., 2018) and asymmetric impacts (Aysan et al., 2009). Motivated by these insights, the study tests asymmetric effects of agricultural price shocks on Pakistan’s AVA and incorporates credit disbursement and labor force. Hypotheses: H1: Positive agriculture price shock increases agricultural economic growth. H2: Negative agriculture price shock increases agricultural economic growth.

Data: Annual time series for Pakistan from 1970–2018. Variables: agriculture, forestry and fishing value added (AVA, current US$, WDI); credit disbursement to agriculture (CD, rupees millions; Pakistani statistical yearbooks and economic surveys); labor force (LF, millions; national sources); agricultural product prices (PI; wholesale price indices of wheat and cotton). All variables transformed to natural logs for estimation. Rationale for commodity choice: wheat and cotton are leading cash crops from different production seasons. Model: Baseline long-run specification AVA = β0 + β1(PI) + β2(CD) + β3(LF) + u. To capture asymmetry, prices are decomposed into partial sums of positive and negative changes (PI⁺ and PI⁻) following Shin et al. (2014), and the NARDL framework is estimated: Long run: AVA_t = Φ0 + Φ1 PI_t⁺ + Φ2 PI_t⁻ + Φ3 CD_t + Φ4 LF_t + ε_t. Short run: An Error Correction Model (ECM) form includes lagged differences of lnAVA, lnPI⁺, lnCD, lnLF and lagged levels for cointegrating relations; optimal lags selected by AIC. Pre-estimation: Although NARDL does not require pre-testing for unit roots when variables are I(0)/I(1), it is invalid with I(2). Thus ADF, PP, and KPSS tests are used to confirm no variable is I(2). Results indicate variables are I(0)/I(1): ADF and PP stationary at first differences; KPSS stationary at levels. Cointegration: Pesaran et al. (2001) bounds testing applied. Diagnostic/stability tests include serial correlation (Breusch-Godfrey), normality (Jarque-Bera), heteroscedasticity (Breusch-Pagan-Godfrey), Durbin-Watson, and stability via CUSUM and CUSUMSQ. Dynamic multiplier graphs (DMG) illustrate asymmetric adjustment paths.

- Unit root and stationarity: ADF and PP indicate all variables are I(1); KPSS indicates stationarity at levels; no I(2) variables, meeting NARDL conditions. - Cointegration: Bounds F-statistic = 8.649 (> 5% upper bound 3.49), indicating a long-run cointegrating relationship among AVA, PI⁺, PI⁻, CD, and LF. - Short-run NARDL results: Both positive and negative price shocks significantly increase AVA, with negative shocks having a larger short-run effect. Reported elasticities: PI⁻ shock +0.507%, PI⁺ shock +0.168%. Credit disbursement (CD) positively affects AVA (+0.379% for a 1% rise in CD). Labor force (LF) positively affects AVA (+1.291% for a 1% rise in LF). Model fit and diagnostics are satisfactory: Adjusted R² = 0.999; no serial correlation or heteroscedasticity; residuals largely normal; D-W ≈ 1.96. - Long-run NARDL results: Both PI⁺ and PI⁻ significantly and positively affect AVA, with PI⁺ having a larger magnitude: PI⁺ coefficient ≈ +0.418%; PI⁻ coefficient ≈ +0.109%. CD positively affects AVA (+0.055%), and LF positively affects AVA (+1.138%). - Stability and dynamics: CUSUM and CUSUMSQ indicate parameter stability within 5% bounds. Dynamic multiplier graphs show positive asymmetry, with PI⁺ shocks exerting a larger long-run impact on AVA than PI⁻ shocks. Overall: H1 and H2 are supported in both short and long run; positive price shocks have a stronger long-run effect than negative shocks.

The study addresses whether agricultural commodity price shocks asymmetrically affect Pakistan’s agricultural value added. Findings indicate both positive (PI⁺) and negative (PI⁻) shocks raise AVA, confirming asymmetric effects with stronger long-run influence from positive shocks. Mechanisms differ: price increases can improve farmer revenues and incentivize production, while price decreases can stimulate consumer demand and potentially expand output to meet higher consumption. However, rising prices can exacerbate inflation and disproportionately burden low-income households, highlighting a policy trade-off between farmer incentives and consumer welfare. Credit disbursement supports AVA in both horizons, though the long-run elasticity is modest, reflecting implementation challenges, governance issues, and the need to direct credit toward productivity-enhancing investments (e.g., technological innovation). The labor force contributes strongly to AVA, underscoring the importance of skills and human capital; yet, constraints like low literacy hinder full productivity gains. Stability tests (CUSUM/CUSUMSQ) and dynamic multipliers corroborate model robustness and the presence of positive asymmetry. Contextual factors—including global crises (late-1990s Asian financial crisis, 2008 and 2011 food/oil crises), floods (2010, 2014), energy and input costs, exchange rates, and policy interventions—interact with price shocks to shape outcomes, indicating the price-growth link is complex and mediated by structural conditions.

The paper contributes by applying a NARDL framework to decompose agricultural price shocks and estimate asymmetric impacts on Pakistan’s agriculture, forestry and fishing value added (1970–2018). Empirically, both positive and negative price shocks increase AVA in the short and long run, with positive shocks exerting a larger long-run effect. Credit disbursement and labor force also significantly and positively influence AVA over both horizons. Policy implications include the need for vigilant price control and credible monetary policy to manage inflationary pressures while preserving incentives for producers. Credit programs should be carefully calibrated and targeted to technological upgrading to reverse declining agricultural shares and reduce reliance on traditional methods. Enhancing education and awareness among agricultural labor can improve resource use and productivity. Future research directions proposed include extending analysis to other agriculture-based countries (e.g., African nations), incorporating additional commodities (maize, rice, sugarcane), examining technological, environmental, and energy factors, considering structural breaks (e.g., 2008/2012 crises) within NARDL, and employing alternative time-series and panel methods (ARMA, Johansen, SARIMA, VAR, ARDL, GMM, PMG, AMG).

The study focuses on Pakistan using 1970–2018 data and models price indices for wheat and cotton; structural breaks are not explicitly modeled. The authors note that while both PI⁺ and PI⁻ raise AVA, evidence of asymmetry can be scarce and the price-growth relationship remains influenced by factors such as weather, market forces, legislation, and infrastructure, suggesting scope for broader commodity coverage, country contexts, and methods in future work.