Introduction

In many countries, jobless growth, which refers to a significant decline in employment growth despite output growth, has been prevalent over the last few decades (Basu and Das 2016). Employment growth is fundamentally linked to output growth since employment demand is derived from output. However, a disconnect may arise when non-labor factors such as capital replace labor in the production process. This can potentially widen the gap between employment growth and output growth, leading to jobless growth.

Scholars have studied jobless growth in developed countries such as the United States (Basu and Foley 2013; Khemraj et al. 2006), and there is a growing body of literature documenting the importance of jobless growth in developing countries. For instance, Mukherjee (2014) notes that developing countries may experience jobless growth in the formal sector during liberalization episodes, such as trade reform policies that increase competition from foreign markets. Although the natural growth process of an economy typically involves labor migration from agriculture to more productive service and industry sectors, displaced workers from high-productivity sectors in developing countries with large inter-sectoral productivity gaps are more likely to end up in low-productivity (mostly informal) sectors (McMillan and Rodrik 2011). While employment and output dynamics vary across sectors, analyzing recent trends can provide important insights. Another factor to consider for developing countries is the risk of “premature deindustrialization,” where the economies of developing countries transform into service-oriented economies without undergoing proper industrialization, potentially affecting economic growth adversely (Rodrik 2016).

Bangladesh’s economy has grown rapidly in recent decades, despite its comparatively small economic base. However, in the ten years leading up to 2013, the country’s employment elasticity remained low at 0.4, lower than most countries in Southeast Asia (ILO, 2013). Given that employment growth has significantly lagged behind economic growth, there have been some concerns that Bangladesh’s recent growth may have been jobless. The government’s growth strategy outlined in the five-year plans has considerably accelerated GDP growth and shifted the production structure from agriculture to industry and services. While the agriculture sector’s share of GDP has gradually declined over the past few decades, the shares of industry and services have been increasing steadily (Bangladesh Planning Commission 2020).Footnote 1 In 2016–2017, agriculture accounted for 40.6% of the employed population, while industry and services accounted for 20.4 and 39%, respectively. This is despite the fact that agriculture’s share to the GDP was only 14.74% in 2016–2017, indicating an apparent structural imbalance between employment and output.

Although many jobs have been created in manufacturing, construction, professional services, transport, trade, and other informal services, Bangladesh still faces some challenges. According to the projections in the 8th Five-Year Plan (July 2020–June 2025), the country’s labor force is expected to grow at a rate of 2.2% per year, which is faster than the rate of 1.6% in 2010–2017. Ensuring the creation of more decent jobs, along with GDP growth, is a significant policy challenge that lies ahead for Bangladesh. It is also important to scrutinize the inclusivity of economic growth to facilitate poverty reduction. As noted by Rahman and Islam (2003) and Alessandrini (2009), employment-inducing and equitable growth facilitates poverty reduction, as the benefits of growth for the poor depend on both the nature and extent of employment generated by growth itself.

In this context, this study aims to address two important questions: (i) To what extent has the recent economic growth in Bangladesh been employment-intensive? and (ii) How has the structure of employment responded to economic growth? We address these questions by utilizing the employment elasticity of output in Bangladesh between 2002 and 2017. However, aggregate employment elasticity is insufficient to analyze the responsiveness of employment growth to output growth. Therefore, we consider sectoral trends, as well as trends in rural and urban areas, and by gender. We generate sectoral and aggregate elasticities using conventional methods of calculating elasticities. Using this method, we calculate both the level and change decomposition of elasticity. The former measures the impact of a sector’s employment elasticity on the level of aggregate elasticity, whereas the latter allows for the analysis of the impact of changes in sectoral elasticities and composition on aggregate elasticity. We also estimate the elasticities in rural and urban areas and by gender by employing a panel data regression model with sectoral fixed effects.

The empirical findings show that aggregate employment elasticity has declined sharply since 2013, suggesting that overall economic growth is increasingly leading to jobless expansion. Although agriculture remains the largest source of employment, its influence is declining, while the construction and manufacturing sectors are gradually becoming more prominent in terms of employment share. However, because of its large employment share, changes in sectoral elasticity in agriculture still dictate the direction of aggregate elasticity, given that Bangladesh’s economy is still at a stage where this is the case. Furthermore, the weak effect of changes in employment share on changes in aggregate elasticity suggests weak evidence of rapid structural change in sectoral employment shares. The regression estimates suggest that employment elasticities have generally declined across different employment categories, although the results are not statistically significant in all cases. Throughout the period, economic growth has exerted a strong positive impact on female employment, whereas male employment elasticity has shown a relatively subdued response.

The remainder of the paper is structured as follows. Sect. "Defining the relationship between employment and output" delves into the theoretical link between employment and output. Sect. "Literature review" offers an overview of the existing literature relevant to our study. Sect. "Methodology" elaborates on employment elasticity, its breakdown, and the specifications of our empirical model. Sect. "Data and descriptive statistics" details the data sources and provides a summary of the key variables used in the model. Sect. "Empirical results" presents our primary empirical findings. Sect. "Discussion" is dedicated to discussing the policy implications derived from our results. Sect. "Conclusion" wraps up the study with concluding remarks.

Defining the relationship between employment and output

The existing literature addresses two distinct theoretical aspects that examine the relationship between output and employment: Okun’s law (OL) and the Kaldor–Verdoorn effect (KV). Okun (1962) observed a negative relationship between the unemployment rate and real GDP growth, which is now referred to as Okun’s law. In contrast, the KV effect is Kaldor’s (1967) interpretation of Verdoorn’s (1949) findings, which indicate a positive and statistically significant relationship between industrial output growth and labor productivity growth in advanced industrial countries. Kaldor combined Young’s (1928) understanding of increasing returns as a “macrophenomenon” with Verdoorn’s (1949) empirical findings to propose that output growth itself leads to changes in productivity. Kaldor (1967) empirically formulated this relationship as a regression of productivity growth (\(p_{t}\)) on output growth (\(y_{t}\)) or, alternatively, of employment growth (\(e_{t}\)) with respect to output growth in time t. Algebraically,Footnote 2

$$p_{t} = - \alpha + \beta_{1} y_{t}$$
(1)
$$e_{t} = \alpha + \beta_{2} y_{t}$$
(2)

The empirical stability of Okun’s law is intensely debated among scholars. Policymakers can use Okun’s law as a tool to analyze the cost of the cyclical unemployment rate in terms of aggregate output and design and implement macroeconomic policies accordingly (Hamia 2016). However, Hanusch (2013, p. 1) asserts that “Okun’s law is a crude approach to analyzing the transmission of economic growth into employment, as it does not pay close attention to the multiple structural mechanisms that account for job creation and job destruction.” Despite its limitations, Okun’s law has been employed in various studies to analyze jobless growth and the employment-output scenario in general (see Hamia 2016; Hanusch 2013; Izyumov and Vahaly 2002; Lal et al. 2010).

Basu and Foley (2013) argue that the KV effect is a better-suited tool in many regards than Okun’s law. Firstly, they claim that Okun’s law lacks any theoretical underpinning except the simple idea that labor is needed to produce output (Okun 1962). In contrast, the KV effect is interpreted as a “structural relationship obtaining in capitalist economies and has been used as such within the heterodox macroeconomics tradition” (Basu and Foley 2013, p.1085). Secondly, according to Okun’s law, the relationship between output growth and changes in the aggregate unemployment rate is mediated through changes in the labor force participation rate. Thus, the Okun coefficient can change if the labor force participation rate changes, even though the relationship between output growth and employment remains unaltered. Tejani (2016) reinforces the arguments of Basu and Foley (2013) and advocates for the use of a Kaldorian framework when examining the relationship between output growth and employment growth. In the next section, we provide a more in-depth review of the literature on this relationship.

Literature review

There is a relative scarcity of literature discussing employment and output in the context of Bangladesh. Although some brief discussions have taken place, a comprehensive analysis is still needed. Iqbal and Pabon (2018) argued that there is no evidence of jobless growth in Bangladesh, citing recent statistics from the Labor Force Survey. According to this survey, 1.4 million new jobs were created between 2013 and 2015–2016, and another 1.3 million jobs were added from 2015–2016 to 2016–2017. This resulted in a 1 percent increase in the number of jobs between 2013 and 2015–2016 and a 2.2 percent increase between 2015–2016 and 2016–2017. Additionally, the income elasticity of job creation was around 0.17 between 2013 and 2015–2016, and it increased to 0.30 between 2015–2016 and 2016–2017. However, their analysis employed a relatively short time period and did not consider a detailed long-run employment scenario in all economic sectors and subsectors. In an analysis of the linkages between employment and poverty, Rahman and Islam (2003) discussed the employment intensity of growth and how the employment structure responded to growth in the 1980s and 1990s in Bangladesh. They found that employment responded positively to output growth in various subsectors of the manufacturing industry. However, for most industries, the employment elasticity estimates were less than unity and declined in the 1990s compared to the 1980s. We will build on the findings of Rahman and Islam (2003) for the 2000s and 2010s and expand the analysis to other sectors beyond manufacturing.

Lal et al. (2010) tested the validity of Okun’s law using annual time series data from 1980 to 2006 in several Asian countries, including Bangladesh. The authors concluded that there is a long-run relationship between the output gap and the unemployment gap in Bangladesh, Pakistan, India, Sri Lanka, and China. Hamia (2016), on the other hand, found mixed evidence regarding the validity of Okun’s law in a study involving seventeen developing countries in the Middle East and North Africa region. The author attributed the differences in validity to variations in labor market regulations and economic differences across the countries. By analyzing the trends in eight East Asian countries, Hanusch (2013) concluded that Okun’s law held particularly well for nonagricultural jobs, while the relationship was reversed for agricultural jobs. The structure of the labor market in a country may also account for the instability of Okun’s law.

Basu and Das (2016) examined jobless growth in India and the US by analyzing the employment elasticities of various sectors between 1977 and 2011. They found a similar trend in both countries: a long-term decline in aggregate employment elasticity since the mid-1970s, despite fluctuations over shorter periods. In this study, we employ a similar approach to understand the dynamics of employment and output in twelve sectors of the Bangladeshi economy. We provide further details in the next section.

Tejani (2016) used the Kaldorian framework to analyze Indian data and found a significant decline in the employment elasticity of output growth (i.e., the Kaldor–Verdoorn (KV) coefficients) for both the formal sector and total employment over time. The decline is attributed to endogenous changes in productivity that result in a lag in employment. As a result, jobless growth has become prevalent in India. Similarly, Pieper (2003) found strong support for the Verdoorn law in developing countries, showing a robust positive relationship between the rate of growth of employment and the rate of growth of output in all nine sectors analyzed. The estimated sectoral elasticities with respect to output growth were significantly less than unity.

Onaran (2008) analyzed jobless growth in Central and Eastern European countries (CEECs) by testing the effects of various domestic and international factors on employment. The findings indicate that jobless growth in manufacturing occurs regardless of wage developments in most cases. Despite rapid integration of the CEECs into the European economic arena through foreign direct investment and international trade, job losses in the manufacturing industry persisted. The benefits of higher economic integration were overshadowed by international competitive pressures, resulting in labor-saving growth without generating jobs.

The fourth industrial revolution may result in jobless growth. According to Li et al. (2017), the revolution creates demand for new goods and services, which may lead to the creation of new occupations and productivity enhancements, primarily benefiting investors and innovators. However, labor is increasingly being replaced by capital, and manual labor jobs are being automated, leading to technological unemployment (Peters 2017).

This paper aims to contribute to the existing literature by providing insight into the dynamics of employment and output in Bangladesh based on recent data. This step is necessary in evaluating Bangladesh’s economic growth quality in recent decades. Moreover, through this research, we can understand how the different sectors in Bangladesh differ in their growth and labor absorption capacities from those in other developing and developed countries.

Methodology

The employment elasticity of output growth is a measure of the responsiveness of employment to changes in output. Previous studies have used this measure, including Basu and Das (2016), Basu and Foley (2013), Rahman and Islam (2003), and Tejani (2016). For brevity, we refer to this measure as “employment elasticity” in the rest of the analysis. Since the Labor Force Surveys are only published every three to five years by the Bangladesh Bureau of Statistics, a continuous time series analysis is not possible. Therefore, we use the conventional elasticity equation to calculate sectoral and aggregate elasticities and decompose their levels and changes to assess each sector’s trends separately. To gain a more comprehensive understanding of employment distribution, we examine employment dynamics using a fixed-effects panel regression model to estimate elasticities for rural, urban, male, and female employment. We discuss each of these methods in detail in the following subsections.

Employment elasticity of output and level decomposition

Following the work of Basu and Das (2016), we calculate the employment elasticity, level decomposition, and change decomposition. To begin, we compute the employment elasticity for the aggregate economy using the following equation:

$$\eta = \frac{{ \frac{\Delta E}{E}}}{{\frac{\Delta Y}{Y}}} = \frac{\Delta E}{{\Delta Y}} \times \frac{Y}{E}$$
(3)

Here, \(E\) represents employment, \(Y\) represents real output, \(\Delta E\) represents the change in employment, and \(\Delta Y\) represents change in real output. The methodology aims to analyze trends in various sectors of the aggregate economy, which may exhibit different behaviors in terms of employment elasticity. To achieve this objective, we decompose the level of aggregate employment elasticity into sectoral elasticities, sectoral employment shares, and sectoral proportion of aggregate growth over a specific period. Assuming there are n sectors in the economy, we can calculate:

$$E = \sum E_{i} {\text{ and }}Y = \sum Y_{i} ,$$
(4)

such that the aggregate employment elasticity in Eq. (3) can be defined as:

$$\eta = \mathop \sum \limits_{i = 1}^{n} \eta_{i} g_{i} e_{i} ,$$
(5)

where

$$\eta_{i} = \frac{{ \frac{{\Delta E_{i} }}{{E_{i} }}}}{{\frac{{\Delta Y_{i} }}{{Y_{i} }}}}{\text{ and }}g_{i} = \frac{{ \frac{{\Delta Y_{i} }}{{Y_{i} }}}}{{\frac{\Delta Y}{Y}}}{\text{ and }}e_{i} = \frac{{E_{i} }}{E}.$$
(6)

In Eq. (6), \(\eta_{i}\) represents the employment elasticity in sector i, \(g_{i}\) denotes the growth of the ith sector as the ratio of the rate of growth of aggregate output, and finally, \(e_{i}\) represents the share of total employment contributed by the ith sector.

Equation (5) shows that the aggregate employment elasticity is the sum of the product of sectoral employment elasticity, employment share, and growth across all economic sectors. Thus, the contribution of any specific sector to the aggregate employment elasticity is determined by three components: employment elasticity, employment share, and relative sectoral growth.

Change decomposition

In addition to the level decomposition, we also decompose the change in aggregate employment elasticity between two periods to investigate the factors driving the change in elasticity over time. This decomposition exercise also provides insights into the structural changes in the economy over the years. Assuming that \(t_{0}\) refers to an initial period and \(t_{1}\) refers to a final period, the change in the aggregate employment elasticity between the two periods can be expressed as:

$$\eta^{{t_{1} }} - \eta^{{t_{0} }} = \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{1} }} g_{i}^{{t_{1} }} \eta_{i}^{{t_{1} }} - \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{0} }} g_{i}^{{t_{0} }} \eta_{i}^{{t_{0} }} .$$
(7)

Adding and subtracting \(\mathop \sum \limits_{i = 1}^{n} e_{i}^{t1} g_{i}^{t1} \eta_{i}^{t0}\) and \(\mathop \sum \limits_{i = 1}^{n} e_{i}^{t1} g_{i}^{t0} \eta_{i}^{t0}\) on right-hand side of Eq. (7) yields:

$$\eta^{{t_{1} }} - \eta^{{t_{0} }} = \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{1} }} g_{i}^{{t_{1} }} \left( {\eta_{i}^{{t_{1} }} - \eta_{i}^{{t_{0} }} } \right) + \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{1} }} \eta_{i}^{{t_{0} }} \left( {g_{i}^{{t_{1} }} - g_{i}^{{t_{0} }} } \right) + \mathop \sum \limits_{i = 1}^{n} \eta_{i}^{{t_{0} }} g_{i}^{{t_{0} }} \left( {e_{i}^{{t_{1} }} - e_{i}^{{t_{0} }} } \right).$$
(8)

Separating the three components, we get:

$$\eta^{{t_{1} }} - \eta^{{t_{0} }} = \Delta {\text{SEL}} + \Delta {\text{SGW}} + \Delta {\text{SEM}},$$
(9)

where \(\Delta {\text{SEL}}\) captures changes in sectoral elasticities:

$$\Delta {\text{SEL}} = \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{1} }} g_{i}^{{t_{1} }} \left( {\eta_{i}^{{t_{1} }} - \eta_{i}^{{t_{0} }} } \right).$$
(10)

ΔSGW captures changes in relative sectoral growth rates:

$$\Delta {\text{SGW}} = \mathop \sum \limits_{i = 1}^{n} e_{i}^{{t_{1} }} \eta_{i}^{{t_{0} }} \left( {g_{i}^{{t_{1} }} - g_{i}^{{t_{0} }} } \right).$$
(11)

ΔSEM captures changes in sectoral employment shares:

$$\Delta {\text{SEM}} = \mathop \sum \limits_{i = 1}^{n} \eta_{i}^{{t_{0} }} g_{i}^{{t_{0} }} \left( {e_{i}^{{t_{1} }} - e_{i}^{{t_{0} }} } \right).$$
(12)

Equation (9) illustrates that the change in aggregate employment elasticity can be decomposed into three components: changes in sectoral employment elasticities (ΔSEL), changes in relative sectoral growth (ΔSGW), and changes in sectoral employment shares (ΔSEM).

The first component, ΔSEL, is a weighted sum of changes in employment elasticity within each sector between the final and initial period, using employment shares and relative growth rates in the final period as weights. It reflects the contribution of elasticity changes within each sector to the change in aggregate employment elasticity.

The second component, ΔSGW, is a weighted sum of changes in the relative growth rates of each sector, using employment shares in the final period and employment elasticity in the initial period as weights. This component indicates the contribution of changes in output’s relative sectoral growth rates to the change in aggregate employment elasticity.

The third and final component, ΔSEM, is a weighted sum of changes in the employment share of each sector, using employment elasticity and relative growth in the initial period as weights. This component represents the contribution of changes in each sector’s employment share to the change in aggregate employment elasticity and helps to analyze the extent of structural changes occurring in the economy.

Regression model

We conducted an econometric analysis following Tejani’s (2016) methodology by using least square regressions of employment on output to determine the employment elasticity with respect to output growth with sectoral fixed effects. The fixed-effects model controls for the unobserved characteristics of each sector. Moreover, the methodology for evaluating employment elasticities—or the indirect effects of economic growth on employment, as outlined by Kapsos (2005) and Anderson (2015)—involves specifying an employment equation where output is the sole explanatory factor. We estimated separate coefficients for total employment, rural employment, urban employment, female employment, and male employment. As the output data is not disaggregated by these categories, a regression model is an appropriate tool for this analysis. Additionally, the regression model allows for causal analysis between the variables. To interpret the coefficients as elasticities, we transformed the variables into logs. We also used robust standard errors to deal with heteroscedasticity. The regression model is defined as follows:

$$\log \left( {e_{it} } \right) = \alpha + \beta \log y_{it} + a_{i} + u_{it}$$
(13)

where \(e\) denotes employment, \(y\) denotes output, \(a_{i}\) represent unobserved characteristics, \(u\) is the error term, \(i\) and \(t\) denote sector and time, respectively, and the coefficient \(\beta\) is interpreted as employment elasticity of output.

Data and descriptive statistics

We analyzed the dynamics of employment and output growth by collecting data from various rounds of Labor Force Surveys (LFS) and National Accounts Books published by the Bangladesh Bureau of Statistics from 2002 to 2016–2017. According to the latest Labor Force Survey, Bangladesh Bureau of Statistics defines employment as—“All persons older than 15 years who, during a specified period (7 days prior to the survey) was involved at least for 1 h, in any form of work for wage or salary, profit or family gain and including the production of goods for own consumption.” The Labor Force Surveys conducted in 2002 and 2005 categorized employment by industries disaggregated according to ISIC Rev 3.1, while from 2010 onward, ISIC Rev-4/BSIC 2009 at 1 digit-Section was adopted. For consistency in estimations, we selected twelve economic sectors in accordance with both standards. We obtained employment data from multiple rounds of the Labor Force Survey conducted by the Bangladesh Bureau of Statistics, including the years 2002–2003, 2005–2006, 2010, 2013, and 2016–2017.Footnote 3 The employment data include both formal and informal employment. We used GDP at constant prices at the base year of 2005–2006 for each sector as a measure of output or value added. We summed up the GDP and employment of the twelve sectors considered to obtain the aggregate measures.

Descriptive statistics

The figures presented in Table 1 demonstrate a clear trend: Employment variability is higher than its mean value. This phenomenon may be attributed to various factors, including differences in skill levels, education, and work experience within the workforce. Moreover, the large informal sector in the economy may contribute to unstable employment conditions, variable incomes, and a lack of job security. This is especially pronounced in the rural sector, where employment variability is nearly six times greater than that observed in urban areas.

Table 1 Summary statistics
Full size table

Empirical results

In the following subsections, we present the study’s findings, which include a breakdown of sector-specific employment elasticities, as well as employment elasticities by gender and geographic area. These analyses provide insight into the phenomenon of jobless growth in Bangladesh.

Trends and level decomposition of employment elasticities

Figure 1 illustrates the changes in aggregate employment elasticity (Eq. 3) during the subperiods 2002–2005, 2005–2010, 2010–2013, and 2013–2017, corresponding to the publication of the Labor Force Survey. The peak value of aggregate elasticity was 0.481 in 2010. After a modest drop to 0.419 in 2013, it saw a sharp decline to 0.167 by 2017. The elasticity decline is explained by the significant decrease in employment growth from over 13 percent in 2010 to a mere 3.7 percent in 2017. This was despite nominal output growth exceeding 20% during these periods, which supports the evidence of “jobless growth.” In the next subsection, we analyze the sectoral breakdown of aggregate employment elasticity.

Fig. 1
figure 1

Source Authors’ illustration using data from LFS and BBS National Accounts (various years)

Aggregate employment elasticities from 2005 to 2016.

Full size image

To gain insight into the extent of labor absorption in Bangladesh, Table 10 in Appendix presents employment elasticities from selected comparable countries over recent decades. It is important to note that there is not a universally accepted benchmark for comparing countries’ historical elasticities. As shown in Table 10, the elasticity figures from recent periods closely resemble those of Bangladesh, indicating similar labor absorption trends among these nations.

Figure 2 illustrates the employment elasticities by sector, while Table 2 reports the elasticities for each sector (Eq. 5) for the four subperiods. Several remarks are worth mentioning. Firstly, there has been a noticeable decline in employment elasticity in the agriculture, utilities, manufacturing, and financial service sectors during the most recent period. Given that both agriculture and industry contribute to 60% of total employment, the moderately negative elasticities are particularly worrisome. Secondly, the high employment elasticity in construction is consistent with the rapid expansion of construction activities, thanks to robust growth in foreign remittance income. Employment growth of male workers in the construction sector increased from 3 to 17 percent since 2000 (Sen et al. 2021). Thirdly, after declining for the previous two periods, both the hotel and transport sectors show higher elasticities of employment with respect to output growth. Rapid urbanization and rising income are likely the two important factors behind the increased employment elasticities in these two service sectors. Fourthly, the mining and power sectors are interdependent, with inputs (coal, gas) from the former being used to produce output (electricity) in the latter sector. The very high elasticities in these sectors during 2005–2013 partly reflect the introduction of independent power producers (IPPs) to ramp up power generation capacity. Currently, nearly 50 percent of the generation capacity is served by IPPs. Finally, the comparably persistent high elasticity in the public administration and defense sector provides an early indication of the huge and rising cost of public-sector pensions in the years ahead.

Fig. 2
figure 2

Source Authors’ illustration using data from LFS and BBS National Accounts (various years)

Employment elasticities by sector from 2005 to 2016.

Full size image
Table 2 Employment elasticities of different sectors in the periods 2002–2016/2017
Full size table

Table 3 presents the relative output growth by sector over the four subperiods. The agriculture, forestry, and fisheries sector has consistently grown at a rate below unity, and in the two most recent periods, it has dropped below 0.5, indicating that the sector has been growing slower than the aggregate economy. The other sectors, in general, have grown at a rate above or close to the growth of the aggregate economy. These observations imply that the agricultural sector in Bangladesh has been stagnating since the 2000s decade. From Table 4, it can be observed that even though the employment share of agriculture is larger than all other sectors by far, it has been slowly declining over the periods. In most of the other sectors, the employment share has remained virtually at the same level. Only the manufacturing and construction sectors have witnessed a gradual increase in the employment share. Thus, it can be inferred that there has been a movement of labor from agriculture to manufacturing and construction sectors since the 2000s decade.

Table 3 Relative sectoral growth in the periods 2002–2017
Full size table
Table 4 Relative share of employment for each sector in the periods 2002–2017
Full size table

Table 5 presents the contribution to aggregate employment elasticity for each sector. The manufacturing, wholesale and retail trade, and transport, storage, and communication sectors were the main contributors to the aggregate elasticity in the 2002–2005 period. However, in the 2005–2010 period, agriculture was the prime contributor, accounting for 46.43% of the aggregate elasticity. But, agriculture’s contribution turned negative and very large (-47.79% of the aggregate elasticity) in the most recent period, indicating that employment elasticity in agriculture has the opposite sign to that of the aggregate elasticity. In the 2010–2013 period, the manufacturing sector’s contribution was quite high (66.21% of the aggregate elasticity), but it turned negative in the following period. Subsequently, in the 2013–2017 period, the transport, storage, and communication sector was the largest contributor to aggregate employment elasticity.

Table 5 Aggregate elasticity contributions by sector: 2002–2017 (in percentage)
Full size table

The trend in elasticities suggests that, since 2013, economic growth across the entire economy has increasingly led to jobless expansion (refer to Fig. 2).Footnote 4 Concerningly, employment elasticity was negative in six sectors during the 2013–2017 period, indicating that employment had declined while output grew. These observations suggest the presence of labor-saving technological changes and productivity growth in the agriculture and industry sectors. Moreover, the growth of agriculture has significantly lagged behind the aggregate growth, and its employment share is gradually decreasing. Conversely, the growth rate of other sectors has consistently been equal to or higher than the aggregate growth rate. Thus, while agriculture remains a primary source of employment, its importance is diminishing, and the construction and manufacturing sectors are gradually becoming more significant in terms of employment share.

Change decomposition of employment elasticities

Table 6 displays the contribution of the specific component to the change in overall elasticity. The results demonstrate that the change in aggregate elasticity was positive only during the 2005–2010 period. In the subsequent two periods, it turned negative. The magnitude of the negative elasticity is larger in the 2013–2017 period, which accounts for the sharp decline in elasticity previously discussed. The total row entry for the change in sectoral elasticity has dominated the corresponding totals for sectoral growth and sectoral employment. This implies that aggregate employment elasticity has changed primarily due to the change of elasticity within individual sectors. To put it another way, recent changes in the relationship between economic growth and employment have been driven more by changes in each sector’s specific employment rate rather than by other factors, such as changes in government policies or technological advancements.

Table 6 Change in elasticity for each sector in the periods 2002–2017
Full size table

The other two components have a relatively weak influence in comparison. The weak effect of sectoral growth can be attributed to the fact that there has not been much change in the relative position of sectors with respect to growth over the periods, and the weak effect of sectoral employment is due to the very slow change in sectoral shares of employment (see Table 7). It could be that some sectors are growing at relatively the same rates across time and the shares of employment in each sector have changed very slowly over time.

Table 7 Change in sectoral growth for each sector in the periods 2002–2017
Full size table

Comparing the contribution of each sector to the change in employment elasticities, agriculture had the most pronounced effect in 2005–2010, whereas construction and manufacturing had the largest effect in 2010–2013 and 2013–2017, respectively (see Table 8). Furthermore, the employment elasticity of agriculture has changed in the direction of aggregate employment elasticity in all the three periods under consideration.

Table 8 Change in sectoral employment for each sector in the periods 2002–2017
Full size table

Expression (7) shows that in computing ΔSEL, the change in sectoral elasticity between two consecutive periods is multiplied by employment share and relative growth of the sector in the latter period. Due to the high employment share of agriculture, its elasticity change had a more pronounced impact on overall elasticity change relative to other sectors. As previously discussed, even though agriculture’s share in total employment is gradually declining, it still dominates the other sectors by a large margin in this measure.

In sum, Bangladesh is still at a stage where sectoral elasticity change in agriculture dictates the direction of aggregate elasticity because of its large employment share. However, if the current declining trend in its employment share continues, the country will reach a stage in the coming decades where it will no longer decide the direction of aggregate elasticity change. India has reached such a stage by the mid-2000s by following a similar trajectory (Basu and Das 2016). The stagnancy and weak effect of change in employment share on the change in aggregate elasticity suggest weak evidence of rapid structural change in terms of sectoral employment shares.

Employment elasticity by category of employment

This section estimates Kaldor’s structural equation for Bangladesh, as discussed in Sect. "Defining the relationship between employment and output," by constructing a panel dataset by sector due to limited observations. The regression model specified in Eq. (13), as discussed in Sect. "Regression model," is used to obtain the estimations. The results are presented in Table 9.

Table 9 Fixed-effects regression analysis by category of employment
Full size table

The regression results provide interesting insights. The coefficient for total employment is statistically significant at the 5% level in both periods and has decreased, suggesting that employment in the sectors has become less responsive to output growth. There is also a decline in elasticity for both rural and male employment. However, for the period from 2010 to 2017, the coefficients for rural employment are statistically insignificant. On the other hand, there is a sharp increase in urban employment elasticity, and the coefficient is highly significant for the 2010–2017 period. Additionally, while there is a notable and statistically significant decline in elasticity for female employment, the coefficients are either above or nearly equal to unity. This suggests that female employment has been more responsive to output growth compared to other categories.

For the entire period from 2002 to 2017, there are significant observations. The coefficient for female employment stands at 1.316, which is substantially higher than that of any other category, suggesting that female employment is particularly responsive to GDP growth. In contrast, the coefficient for male employment is a more moderate 0.405, indicating that male employment growth has been less responsive to output growth. Ostensibly, there is evidence that employment elasticities are generally declining across different employment categories, although the results are not statistically significant in all cases.

The high employment elasticity of female employment, coupled with declining elasticity in rural areas and increasing elasticity in urban areas, leads to a few important takeaways. Bangladesh has experienced significant economic growth through its exports of ready-made garments. This rapid expansion in the garment sector is primarily driven by an abundant and cost-effective female workforce (Jahan 2012). This trend aligns with the observation that since 2002, GDP growth has been especially favorable to female employment. Additionally, from 2002 to 2013, female employment in the agricultural sector has seen a significant surge, even though the growth in agricultural output during this time was relatively muted. This trend suggests rural female employment has been more reactive to output growth than in urban areas, leading some, like Sen et al. (2021), to term this the “feminization of agriculture.”

Conversely, male employment elasticity over this period is markedly lower than that of females. One reason could be the sectors with high male participation, such as transportation and wholesale/retail trade, have seen robust output growth throughout this time, but male employment growth lagged. However, in sectors primarily based in urban areas, male employment has notably increased. For example, male employment in the construction sector has more than doubled throughout this period. This trend is corroborated by the observation that male employment elasticity is higher in urban regions compared to rural areas.

Notwithstanding, the fall in female employment elasticities between the periods can be attributed to the decline in small-scale RMG enterprises since 2012, which are typically intensive in low-skilled female-oriented employment. This development had a substantial negative impact on the job creation of urban female youth.

In rural areas, agriculture remains the primary source of employment, even with the ongoing diversification into other sectors such as manufacturing, construction, and education. A decline in elasticity highlights a reduction in agricultural employment elasticity and points to the adoption of labor-saving technologies in farming and other agricultural activities. Additionally, employment elasticity for both urban males and females saw an increase from 2010 to 2016 compared to the 2002–2010 period. However, the figures for the 2002–2010 duration are not statistically significant. Urban female employment responsiveness exceeds that of their male counterparts. Both rural male and female employment elasticity experienced a statistically significant decrease between these periods. Notably, female employment elasticity in urban regions outstripped that of rural areas during the 2010–2016 period. This pattern could be indicative of a surge in job opportunities for women in the RMG sector, predominantly found in urban settings. Detailed findings are presented in Table 11 in Appendix.

We should exercise caution when interpreting the regression results, as caution is warranted due to the possibility of endogeneity of output growth in Kaldor’s model (Rowthorn 1975; Skott 1999). If output growth is endogenous, ordinary least squares regression may be inconsistent. Unfortunately, we cannot explicitly test for endogeneity due to a lack of data for a suitable instrumental variable for GDP. Basu and Foley (2013) used aggregate consumption expenditure and government expenditure as instruments for GDP, while Tejani (2016) used gross investment by sector as an instrument for GDP. Unfortunately, industry disaggregated-level data for such variables are not available for Bangladesh, and therefore, we are unable to utilize these instruments for our analysis.

Discussion

Understanding how to foster increased job growth is fundamental for effective policymaking. Creating new jobs is not only pivotal for reducing absolute poverty (Islam 2004) but also essential to harness the country’s demographic dividend. One objective of this study is to offer rigorous evidence on the sensitivity of employment growth to output growth, both at the aggregate and sectoral levels, ensuring that the findings support the goal of efficient job creation.

This study’s findings highlight the gradual rise in prominence of the manufacturing and construction sectors in terms of employment share. Given this trend, it is imperative to amplify focus on these sectors, aiming for enhanced productivity and superior employment quality to cater to the expanding labor force. Notably, in recent times, sectors such as construction, hotels and restaurants, transportation, public administration, and defense have showcased higher employment intensity.

Thus, through rigorous monitoring, and with corrective policy interventions and employment-driven strategies, there should be a concerted effort to channel labor from conventional domains like agriculture to these emerging sectors. Beyond the ready-made garment (RMG) sector, manufacturing niches such as processed food, leather and footwear, light engineering, and pharmaceuticals shine brightly on the growth horizon. Intriguingly, these sectors echo the labor-centric attributes of RMG and have substantial potential in external markets, making them well-suited to create jobs, elevate exports, and stimulate value-added growth.

Tackling the escalating unemployment rates among the educated youth is central to optimizing employment elasticity. With the onset of the Fourth Industrial Revolution, there is a heightened emphasis on superior education standards and progressive skill training initiatives. As such, efforts must be channeled to fortify the capabilities of our educated youth, priming them to adeptly navigate the ever-fluctuating demands of the service sectors.

Conclusion

Bangladesh faces a significant policy challenge of generating employment for its expanding labor force despite experiencing robust economic growth. To investigate whether this growth has been jobless, we analyzed employment and GDP data from 2002 to 2017. Our analysis provides insight into the evolution of employment and output structure in various sectors and categories of the economy, adopting a sectoral perspective as different sectors may behave differently regarding labor absorption. We also examined employment and output dynamics in rural and urban areas and by gender.

By decomposing the aggregate employment elasticity into three components and examining changes in sectoral elasticities, relative growth rates, and sectoral employment shares, we found a sharp decline in aggregate employment elasticity since 2013. The elasticity of total employment also saw a statistically significant drop from 2002–2010 to 2010–2017, suggesting that employment has become less responsive to GDP growth over time.

We identified that the agricultural sector’s elasticity changes still determine the direction of aggregate elasticity due to its large employment share, although the construction and manufacturing sectors are gradually becoming more prominent in terms of employment share and relative output growth. However, there is no evidence of rapid structural change in terms of the impact of sectoral employment shares on the change in aggregate elasticity.

Our fixed-effects panel regression analysis revealed that employment elasticities are generally declining across different employment categories (i.e., rural, urban, male, and female), although the results are not statistically significant in all cases. Throughout the entire period, economic growth has been highly favorable to female employment, while male employment has been much less responsive.