Acadlore takes over the publication of JORIT from 2025 Vol. 4, No. 3. The preceding volumes were published under a CC-BY 4.0 license by the previous owner, and displayed here as agreed between Acadlore and the owner.
Open AccessInstitutional Quality and Economic Growth in Tanzania
View Full Article|
Download PDFAbstract:
In this paper, we use the ARDL method to find the Impact of institutional quality on economic growth in Tanzania from 1990 to 2021. The ARDL technique frees variables from residual correlation as all variables are assumed to be endogenous. They distinguish between dependent and explanatory variables in any long-run relationship, identify the co-integrating vectors with multiple co-integrating vectors, and derive the Error Correction Model (ECM) or Error Correction Model (ECM) Vector Error Correction Model (VECM) by integrating short-run adjustments with long-run equilibrium without losing extended-run information. Our results show all adjustment terms in the respective models that have a long-run relationship have correct (negative) signs and are more than one, implying there is convergence in the long run; that is, the models returned to their long-run equilibrium; the rate (or speed) at which this happened ranged between $15 \%$ to $106.6 \%$ annually. Institutional quality has a significant affirmative (0.047) causal long-run effect on economic growth.
1. Introduction
Countries with broad-based institutions are significantly more likely to achieve sustained growth in the long run, Hartmann and Spurk (2021) as supported by empirical evidence, which is covered below in the literature review section (Abubakar, 2020; Yilmakuday, 2022). This calls for improvement of institutions also, as noted in the literature review section below those institutions are instrumental in influencing or enhancing economic development/growth (North, 1991; Iheonu et al., 2017; OECD, 2001; Thorbecke, 2013 cited in Iheaonu et al., 2021; Hassan and Meyer 2021). It should, however, be cautioned that for this to happen, the improvement should be much more in the productive sectors (Yildirim and Gokalp, 2016, cited in Abubakar, 2020). Tanzania made massive efforts to make economic reforms and improve its institutions. Despite lingering structural constraints and deficiencies, these measures have impacted economic growth. Table 1 below shows that Tanzania's economic growth increased from an average of 1.8% between 1991 and 1995, peaking from 2001 to 2005 before levelling between 2006 and 2020.
Period | Rate |
1991-1995 | 1.80 |
1996-2000 | 4.23 |
2001-2005 | 6.96 |
2006-2010 | 6.12 |
2011-2015 | 6.37 |
2016-2020 | 5.38 |
It should be noted that on top of institutional improvement there also was a surge in production in the mining sector. Using data on Tanzania from 1990 to 2021, this paper applies an autoregressive distributive lag (ARDL) to establish whether institutional quality generates synergies in boosting economic growth and whether economic growth (expressed as the annual growth rate of the real GDP per capita).
The rest of the paper is organized as follows; section 1 will review the literature, and section 2 will discuss the methodology and data used. We discuss the results in section 3, and the last section concludes.
2. Literature Review
The endogenous growth approach based on growth models developed by Lucas (1988), Romer (1986, 1990) and Grossman and Helpman (1991) focuses on the long run and on the internal forces of the economy, particularly those that provide opportunities and incentives to create technological knowledge (Durham, 2004; Dada and Abanikanda, 2022). It augments the first approach as it also addresses gross capital formation. Its premise is that the role (This includes directly increasing capital accumulation and indirectly increasing the stock of knowledge and fostering technological growth of a technologically inferior recipient economy) played by foreign direct investment (Borensztein et al., 1998; de Mello, 1999; Durham, 2004; Tang and Tan, 2018; al Faisal and Islam, 2022) will depend on the absorptive capacity of the host economy. The absorptive capacity enables an economy to benefit from foreign direct investment (Borensztein et al., 1998; de Mello, 1999; Durham, 2004); in addition, a country with strong (weak) absorptive capacity, according to Dada and Abanikanda, 2022 will benefit maximally (minimally) from the growth effect of foreign direct investment (Absorptive capacity includes the level of development of infrastructure (research and development, innovation, the levels of domestic investment) (Bekana, 2016; Naanwaab & Diarrassouba, 2016), the level of institutional development (Alfaro et al., 2004; Benassy-Quere et al., 2007; Farla et al., 2016; Fatima 2016) especially concerning financial markets (Alfaro et al. 2004), the level of human capital development (Borensztein et al., 1998; Fahinde et al., 2015; Bbale & Nnyanzi 2016; Pegkas and Tsamadias, 2016), environmental quality and the degree of openness of the economy (Dada & Abanikanada, 2022)).
Another is the institutional quality approach, which incorporates a complex component of institutions that consist of both informal constraints (sanctions, taboos, customs, traditions, and codes of conduct) and formal rules (constitutions, laws, property rights) (North, 1991; Hassan and Meyer, 2021); the rule of the game in a society or more formally the humanly devised constraints that shape human interaction (Iheonu et al., 2017 citing North, 1981; Islam and Shindiani 2021 citing North 1991; Emara and Rebolledo, 2021; Hartmann and Spruk, 2020). They are instrumental in influencing or enhancing economic development/growth (North, 1991; Iheonu et al., 2017; OECD, 2001; Thorbecke, 2013 cited in Iheaonu et al., 2021; Hassan and Meyer 2021) by boosting the confidence of investors (local and foreign), promoting fairness and equitable distribution of resources (North, 1991), providing a suitable environment for growth-enhancing activities like investment, entrepreneurship and innovation stimulating synergies between foreign direct investment and domestic firms, promoting productivity spill over, inducing complementarities between foreign and domestic investment (Jude and Levieuge 2015; Brahim and Rachdi 2014; Hayat (2019), all quoted in Dada and Abanikanda, 2021), enhancing the finance-growth nexus by providing an enabling environment for the juice of economic prosperity to trickle down to the poorer segments of society (Thorbecke, 2013 cited in Iheaonu et al., 2021), and contributing to improved firm performance (Matashu and Musvoto, 2020). A country with strong (weak) quality institutions like OECD/APAC will maximally (minimally) experience economic growth (Azman-Saini, 2010).
Empirical findings primarily show the presence of an institutional quality-economic growth nexus. Abubakar’s (2020) results show that economic growth responds to institutional quality contract-intensive money and the effective governance index; however, the Impact of the effective governance index is insignificant. Agyei and Idan (2022) assess the moderating role of institutions in the trade openness and inclusive growth nexus in 39 Sub- Saharan Africa (SSA) countries for 1996 to 2017 data and find that institutions strengthen the positive relationship between trade openness and inclusive growth in SSA.
Da Veiga et al. (2022) assess which economic, social, and institutional determinants of economic development are essential to the development of African countries for the years 1996 and 2014. They find a positive association amongst institutional, economic, and social determinants of development, which means that countries that exhibit a reliable performance in institutional indicators also have a satisfactory performance in economic and social indicators, and vice-versa.
However, the results are not as precise for 2014 as for 1996. Dada and Abanikanda (2022) investigate the moderating role institutional quality plays in Nigeria's foreign direct investment-led growth hypothesis from 1984 to 2018 using the autoregressive distributed lag estimation technique. The findings reveal that the interactive effect of institutional indicators with foreign direct investment significantly impacts economic growth in most models, implying that institutions serve as a vital absorptive capacity. They conclude that good institutional quality matters for Nigeria's foreign direct investment and growth.
Emara and Rebolledo (2021) investigate the relationship between economic freedom (the size of government, property rights, monetary policy, access to international trade, and regulation of credit labour and businesses) and economic performance in the APAC and OECD countries for the period 1980-2017 and find that economic freedom positively affects economic performance in the selected countries.
Hartmann and Spurk (2020) examine the contribution of de jure and de facto institutional instability to long- run growth and development for a large panel of countries in the period 1820 – 2016. The evidence suggests that greater de jure and de facto institutional instability has a strong negative impact on income and growth, whereas de facto instability is more important than de jure instability.
Hassan and Meyer (2021) explore the moderating effect of institutional quality on the external debt- economic growth nexus in highly indebted developing countries (HIPC) and find that institutional quality (institutional quality index, government stability, government effectiveness, and law and order) mitigate the negative Impact of external debt on economic growth.
Iheanou et al. (2021) assess the Impact of institutional quality (control of corruption, government effectiveness, regulatory quality, and the rule of law as institutional quality indicators) on economic performance in 12 West African countries from 1996 to 2015 and find that all the indicators of institutional quality employed in the study have a positive and significant impact on economic performance in West Africa.
Islam and Shindaini (2021) examine the impact of institutional quality (INQ) on economic growth (EG) linkage in Bangladesh for 1990 – 2019 and find that INQ affects long-run EG positively.
Matashu and Musvoto (2020) examine connections between institutional quality (corporate governance, macroeconomic fundamentals, the institutional environment) and economic growth in Sub Saharan African countries and find that aggregated composite; corporate governance, macroeconomic fundamentals, and the institutional environment have a statistically significant solid relationship with economic growth.
Matallah and Benlahcene (2021) investigate the Impact of public service quality on economic growth in 15 MENA countries over the period 1996 - 2018 and find that the government effectiveness index exerts a significant positive impact on economic growth in 15 MENA countries. However, government spending (freedom from the government) exhibits a statistically insignificant positive effect on economic growth in the selected countries.
Sarac and Yaglikara (2022) evaluate the developmental differences between countries in the core, semi- periphery, and periphery countries according to institutional quality differences and find the effect of economic freedom on development in all subgroups of core, semi-periphery, and peripheral countries. In addition, while the effect of democracy on development has an increasing effect in core and semi-periphery country groups, the results in peripheral countries are found statistically insignificant.
Utile et al. (2021) examine the influence of institutional quality on the development of the Nigerian economy in the 21st century using annual time series data covering 2001 to 2019 and find that Institutional Quality (INSQ) exerts a significant negative influence on economic growth.
Yilmakuday (2022) investigates the effects of inflation on per capita income growth for thirty-six developed and developing countries and finds heterogeneity of such effects across countries that are shown to be further connected to the strength of their institutions. While the effects of inflation on growth are adverse and significant in countries with more vital institutions, they are positive and significant in countries with weaker institutions.
3. Methodology and Data
In this paper, we use the ARDL method to find the Impact of institutional quality on economic growth in Tanzania from 1990 to 2021.
Nkoro and Uko (2016) have shown that the ARDL technique frees variables from residual correlation as all variables are assumed to be endogenous. Their model distinguished between dependent and explanatory variables in any long-run relationship, identified the co-integrating vectors with multiple co-integrating vectors, and derived the Error Correction Model (ECM) by integrating short-run adjustments with long-run equilibrium without losing extended-run information. Additionally, by excluding non-stationary variables from the analysis through unit root tests, the technique paves away problems associated with violations of assumptions of constant mean and variances that would, among other things, lead to misleading estimates (Nkoro and Uko, 2016).
ARDL also is appropriate in dealing with variables that are stationarity: at the level $\mathrm{I}(0)$ and at the difference I(1); and by use of the Pesaran-Shin bounds co-integration technique (Pesaran and Shin, 1999, and Pesaran et al., 2001 as cited in Nkoro and Uko, 2016) the technique distinguishes between long run and short models. Utile et al., 2021 citing Solow, 1956; Mankiw et al., 1992; and Lucas, 1988 start with the primary neo-classical production function we adopt for our specification.
where: Y depicts economic growth as a function of capital (K) and labour (L).
Utile et al. (2021) continue to include the adjustments of Romer (1986, 1990 as cited in Nkoro and Uko, (2016), where human capital (H as shown in specification 2 below) is also considered to be the primary determinant of economic growth in endogenous growth specification 1 above takes the form of:
Utile et al. (2021) then introduce institutional quality (INSQ, which we have substituted for as Inq shown in specification 3 below) suggested by Feder (1983), Grossman and Helpman (1991) and Ram (1996) cited in Nkoro and Uko (2016) to come up with:
We proceed from here to show that given that capital formation (K) consists of both domestic gross fixed capital formation (GFCF) and foreign direct investment (FDI), these details are added to have specification 4, where K is substituted with gross capital formation (gfcf) and L and H are merged and re-written as human):
where: $Y_t$ is the annual GDP growth rate per capita, inq stands for institutional quality, gfcf is gross capital formation, fdi is foreign direct investment net stock inflows, and human is human capital, proxied by secondary school enrolment.
These variables are obtained from the following sources:
•Worldwide Governance Indicators: Institutional quality is an index that has been estimated using principal component analysis from six variables (voice and accountability, political stability, government effectiveness, regulatory quality, the rule of law, and control of corruption);
•World Development Indicators:
-The annual growth rate of per capita GDP is a proxy for economic growth obtained from FDI inflows as a percentage of GDP;
-Gross capital formation as a percentage of GDP.
-UNDP Human Development Report: percentage of secondary Schooling enrolment as a proxy for human capital.
Specification 4 is expressed in a long-run form as:
where: $Y_t$ is economic growth, $\beta_1$ through $\beta_4$ are parameters to be estimated, $\beta_0$ is the intercept and $\varepsilon$ is the error term.
As for ARDL its generalised form is
where: $\mathrm{Y}_{\mathrm{t}}$ is a dependent variable, $\left(\mathrm{X}_{\mathrm{t}}^{\prime}\right)^{\prime}$ is a $k x 1$ vector that is allowed to be purely $I(0)$ or $I(1)$ or co-integrated, $\delta$ is the coefficient of the lagged dependent variable called scalar, $\beta_{\mathrm{j}}$ are $k x 1$ vectors; $p, q$ are optimal lag orders; $\varepsilon i$ is the stochastic error term.
Economic growth, as proxied by the log of growth rate per capita, averaged 0.384, ranging from --0.233 to 0.658 (Table 2). Growth has been slow compared to other developing countries (Dada and Abanikanda, 2022). Institutional quality (inq), not presented in logs, averaged 1.33, ranging from -0.5.282 to 2.580.
Variable | Mean | Std. | Dev. | Min | Max | CV |
Y | 32 | 0.384 | 0.234 | -0.233 | 0.658 | 0.609 |
inq | 32 | 1.365 | 1.633 | -5.282 | 2.580 | 1.228 |
gfcf | 32 | 1.439 | 0.131 | 1.173 | 1.613 | 0091 |
fdi | 32 | 3.557 | 0.575 | 2.589 | 4.210 | 0.162 |
human | 32 | 1.183 | 0.284 | 0.723 | 1.500 | 0.240 |
We use the ratio of standard deviation to the mean, also called coefficient of variation (CV), to check the size of the standard deviation and, therefore, the relative level of variability. As a rule of thumb, a CV>1 shows a higher variability, and a CV<1 indicates a lower variability. From CV, then gfcf (0.091), FDI (0.162), human per capita (0.240) and economic growth (0,609) have lower variability compared with the rest. Only institutional quality (1.228) has a broader variability. So, variability is not a problem for almost all variables.
Given that computations are made using logs, correlations of logs are presented in Table 3. They indicate a moderate correlation with one another (save fdi, which is highly correlated with human capital), as they are below the benchmark of 0.8 , implying an absence of multicollinearity among the variables (Dada and Abanikanda, 2022). Institutional quality (inq) is of interest, with positive though low correlations with all variables, the highest (0.300) being economic growth. Economic growth $(\mathrm{Y})$ also has low correlations with all other variables, the highest (0.317) being human capital.
Variable | Y | inq | gfcf | fdi | Human |
Y | 1.000 | ||||
Inq | 0.300 | 1.000 | |||
Gfcf | 0.161 | 0.299 | 1.000 | ||
Fdi | 0.289 | 0.084 | 0.692 | 1.000 | |
Human | 0.317 | 0.117 | 0.722 | 0.947 | 1.000 |
4. Results and Discussion
Having obtained a functional form and knowing the variables we are using; we proceed to the next step of obtaining optimal lags for the variables we are using. Given that the economic processes id dynamic where a dependent variable takes time to respond to the effect of regressors (Scott Hacker and Hatemi, 2008 cited in Chikalipa and Okafor 2019), there is a need to capture all past information that could entail the estimation framework; failure to do this would result to misspecification. Using lags becomes essential and choosing the optimal lag length is vital. Choosing the optimal number of lags avoids losing degrees of freedom, multicollinearity, serial correlation, and misspecification errors. A rule of thumb is to have between 1 and 2 lags for annual data. We multiply the lags across periods by 4 or 12 , i.e., between 4 and 8 for quarterly data and between 12 and 24 for monthly data.
However, econometric packages (Akaike's Information Criterion (AIC), Hannan and Quinn Information Criterion (HQIC) and Schwarz's Bayesian Information Criterion (SBIC)) are available and are extremely useful in estimating the appropriate (optimal) number of lags. These are employed here to choose the optimal lag length for our series, and the results are presented below. Because of the detailed output of each variable, the details are not shown here; for illustrative purposes, the output for $Y$ is displayed in Table 4. Five criteria have produced an optimal lag of one. Similar tables could show how lag levels arrived at 2, 1, 1 and 2, respectively, inq, gfcf, fdi and human.
Lag | LL | LR | df | p | FPE | AIC | HQIC | SBIC |
0 | 0.578289 |
|
|
| 0.060342 | 0.030122 | 0.044668 | 0.077701 |
1 | 5.073180 | 8.9898* | 1 | 0.003 | .047022* | -.219513* | -.190422* | -.124355* |
2 | 5.877660 | 1.609 | 1 | 0.205 | 0.047711 | -0.20555 | -0.16191 | -0.06281 |
3 | 5.990220 | 0.22513 | 1 | 0.635 | 0.050891 | -0.14216 | -0.08398 | 0.048156 |
4 | 6.822290 | 1.6641 | 1 | 0.197 | 0.051603 | -0.13016 | -0.05744 | 0.10773 |
Our next step is to check for the stationarity of variables (Mongale et al., 2018) to sieve away non-stationary series, that is, those that have a non-constant mean, a non-constant variance, and a non-constant autocorrelation over time (Yuan et al., 2007 cited in Akinwale and Grobler, 2019; Asteriou and Hall, 2011 cited in Mongale et al. (2018). If we fit regressions that use such series, our results will be spurious, and their outcomes cannot be used for forecasting or prediction (Granger and Newbold, 1974; cited in Akinwale and Grobler, 2019). We are keeping the stationary series. They have non-seasonality, a constant mean, and a constant autocorrelation structure and tend to return to the long-term trend following a shock. Several tests, including the Augmented Dickey-Fuller (ADF), Phillips-Perron, DFGLS, Levin-Lin-Chu and Im-Pesaran-Shin, are used for testing stationarity, also called unit root tests. The ADF test used in our study is an extension of the Dickey-Fuller test, which includes extra lagged terms of the endogenous variable to remove autocorrelation in the error term by adding the lagged difference terms of the regress (Pradhan, 2016 and Makhoba et al., 2019). According to Gujarati and Porter (2009) as cited in llesanmi and Tewari (2017) the ADF test involves estimating the following specification:
where: $\alpha$ is a constant, $\rho$ is an autoregressive coefficient for the series, $y_t$ is the variable in period $t$, $u_t$ is the error term with mean zero and variance one; $t$ the linear time trend and $m$ is the lag order.
Two hypotheses are assessed:
$\mathrm{H}_0: \rho=1$ (contain unit, the data is not stationary).
$\mathrm{H}_{\mathrm{a}}: \rho<1$ (do not contain a unit root, the data is stationary).
The results of the stationarity test are presented in Table 5 below. The criterion is that the test statistic must be greater than the critical value of 5%. All variables are not stationary at the level since their test statistics (column 2) are lower than the critical values at 5% (column 3), also confirmed by non-significant probabilities. We have differenced them to make them stationary at order 1 (column 4 are higher than the critical values at $5 \%$ (column 5 ), also confirmed by significant probabilities. It is noted that fdi at the difference, although significant, is not stationary at $5 \%$. However, it is stationary at 10%, where its statistic of 2.912 is higher than the critical value of 2.626 . At this step then, all variables are proven to be stationary.
Variable | Test Statistics at Level | Critical Value at 5% | Test Statistics at the Difference | Critical Value at 5% |
1 | 2 | 3 | 4 | 5 |
Y | -1.498 | -2.986 | -4.498*** | -2.989 |
inq | -1.677 | -2.986 | -5.306*** | -2.989 |
gfcf | -0.759 | -2.989 | -3.351** | -2.992 |
fdi | -1.750 | -2.989 | -2.912** | -2.992 |
human | -0.805 | -2.989 | -4.078*** | -2.992 |
A bounds co-integration test, our next step, establishes whether a long-run relationship exists between variables. The relationship indicates that time series move together eventually and that the error term resulting from the linear combination of time series quantifies the deviation of the time series from their typical long-run relationship, which can be used to predict their future values. In this case, the test is used to determine the long- run relationship between institutional quality and economic growth in Tanzania (Pesaran and Shin, 1995; Pesaran et al., 2001, cited in Islam and Shindaini, 2021). We first estimate the lag structure for each variable used in estimating the bounds co-integration test, see Appendix A for mathematical formulation. Because of the substantial number of estimates of lag structure we made, we will not display them here, only two for illustration. For instance, when Y is taken as a dependent variable while inq, gfcf, fdi and human as regressors, the lag structure is 1, 0, 1, 1 and 0, respectively, for Y, inq, gfcf, fdi and human.
Similarly, when human is taken as a dependent variable while Y, inq, gfcf and fdi as regressors, the lag structure is 1, 0, 0, 0 and 0, respectively for human, Y, inq, gfcf and fdi. These lags help estimate bounds co- integration tests. We use them to estimate the bounds co-integration test.
Levels | I (0) -0.1 | I (1)-0.1 | I (0)-0.05 | I (1)-0.05 | I (0)-0.025 | I (1)-0.025 | I (0)-0.01 | I (1)-0.01 |
Critical vl | 2.24 | 3.35 | 2.86 | 4.01 | 3.25 | 4.49 | 3.74 | 5.06 |
$\mathrm{H}_0$: no levels of relationship.
$\mathrm{H}_0$ : accept if F < critical value for I(0) regressors.
$\mathrm{H}_0$ : reject if F > critical value for I(1) regressors.
Variables | F statistic | I (0)-0.05 | I (1)-0.05 | Co-integration |
Y | 7.127 | 2.86 | 4.01 | Yes |
inq | 5.938 | 2.86 | 4.01 | Yes |
gfcf | 2.277 | 2.86 | 4.01 | No |
fdi | 8.707 | 2.86 | 4.01 | Yes |
Human | 13.817 | 2.86 | 4.01 | Yes |
Critical values at 5% of 2.86 at a lower bound and 4.01 at a higher bound displayed in Table 7 are to be read in conjunction with test criteria in Table 5. For illustration, in the model where Y is a dependent variable, the F statistic of 7.127 is higher than the critical value of 4.01 of the higher (I(1)) bound at 5%.
Therefore, the null hypothesis of F critical value for I(1) regressors cannot be rejected, implying cointegration in the model where Y is a dependent variable. This applies to models where inq, fdi and human are dependent variables. As for the model where gfcf, since the F statistic of 2.277 is lower than the critical value of 2.86 of the lower (I(0)) bound at 5%, the null hypothesis of F < critical value for I(0) regressors cannot be rejected, implying there is no co-integration in the model where gfcf is a dependent variable. If the estimated value of the Fstatistic exceeds the critical value (Narayan, 2005 cited in Islam and Shindaini, 2021), also see Table 4, a long-run association is validated; in this case, for all models where Y, inq, fdi and human are dependent variables.
ECM/ ARDL | Variable | Y | Inq | Fdi | Human | gfcf |
1 | 2 | 3 | 4 | 5 | 6 | 7 |
ADJ | L1. | -0.987 (0.186) *** | -0.992 (0.213) *** | -0.150 (0.057) ** | -1.066 (0.147) *** | |
LR | Y | 3.191 (1.601) * | -0.899 (0.54) | -0.016 (0.071) | ||
inq | 0.047 (-0.023) ** | 0.128 (0.078) | 0.005 (0.010) | |||
gfcf | -1.069 (0.471) | 9.104 (3.311) ** | -2.627 (1.426) * | 0.10 (0.168) | ||
fdi | 0.154 (0.252) | -1.901 (2.405) | 0.485 (0.042) *** | |||
human | '0.213 (0.495) | 0.571 (4.636) | 2.596 (0.474) | |||
SR | inq | -0.012 (0.006) * | ||||
gfcf D1. | 2.028 (0.772) | 0.635 (0.187) *** | ||||
fdi D1. | -1.945 (0.596) | 8.378 (4.862) * | ||||
ARDL | gfcf L1. | 0.793 (0.091) | ||||
fdist | 0.008 (0.065) | |||||
human | 0.077 (0.122) | |||||
Y | 0.027 (0.045) *** | |||||
Inq | 0.008 (0.006) | |||||
_cons | 1.201 (0.599) | -8.520 (5.740) | 0.733 (0.127) *** | -0.734 (0.222) *** | 0.173 (0.126) | |
Durbin-Watson | 1.8405 | 1.8029 | 2.2742 | 1.9472 | 1.6359 | |
Breusch-Godfrey Ch2 p | 1.000 0.7516 | 0.742 0.3890 | 1.182 0.2770 | 0.018 0.8923 | 0.948 0.3301 | |
White Ch2 p | 30.00 0.4140 | 29.82 0.3224 | 31.00 0.4154 | 17.82 0.5992 | 2035 0.4363 | |
Normality Ch2 p | 1.70 0.4283 | 8.52 0.0141 | 0.45 0.7973 | 5.81 0.0548 | 0.47 0.7901 | |
The error correction term (ECT) counts for one-period lag values of the error terms; its coefficient $[\lambda]$ indicates the speed of correction to long-run stability from any short-run imbalance and assimilates the short-run constants with long-run ones without sacrificing any long-run facts. A long-run association is validated if $[\lambda]$ is negative, statistically significant, and smaller than one, whilst the significant values of regressors confirm the shortrun coefficients' (Islam and Shindaini, 2021).
Our results show all adjustment terms in the respective models that have a long-run relationship have correct (negative) signs and are more minor than one implying there is convergence in the long run; that is, the models returned to their long-run equilibrium; the rate (or speed) at which this happened ranged between 15% to 106.6% annually. Institutional quality has a significant affirmative (0.047) causal long-run effect on economic growth. Since the coefficient is positive and significant at the 5% level, it authenticates the distributive lag property of the ARDL approach that, in the long run, institutional quality exerts a significant affirmative impression on economic growth.
This finding is supported by extensive coverage in the literature review section: Abubakar (2020), Agyey and Idan (2022), Da Viega (2022), Dada and Abanikanda (2022), Emara and Rebolledo (2021) Hartmann and Spurk (2020), Iheanou et al. (2021), Islam and Shindaini (2021), Matallah and Benlahcene (2021) and Utile et. al. (2021).
Economic growth and gross fixed capital formation have significant (respectively at 10% and 5%) affirmative (3.191 and 9.104) causal long-term effects on institutional quality. As seen above, there is much empirical evidence regarding unidirectional causality from institutional quality to economic growth. However, a few studies, like Kabede and Takyii (2017) find economic growth to be an essential determinant of institutional quality in Sub-Saharan Africa; Dandume (2013) and Babasanya et al. (2021) find bidirectional causality between institutional quality to economic growth in Nigeria. Gross fixed capital formation is estimated to have a significant (at a 10% level) negative (-2.627) causal effect on FDI in the long run. Kurniawati (2017) studied 69 countries from 1990 to 2015. The rate of capital formation also potentially influences FDI and economic growth.
Developing economies with a slight initial degree of capital stock inherit more effective marginal rates of return (productivity) and growth rates if sufficient capital stock is injected based on the neoclassical growth model. In the empirical analysis, Barro (1991), Levine & Renault (1992), Kormendi and Meguire (1985) exhibit that the rate of physical capital formation influences the rate of a country’s economic growth. FDI’s significant (at 1% level) and positive (0.485) causal long-run effect on Human capital are in line with literature where foreign direct investment plays the role of directly increasing capital accumulation and indirectly increasing the stock of knowledge and fostering technological growth of a technologically inferior recipient economy (Borensztein et al., 1998; de Mello 1999; Durham 2004; Tang and Tan, 2018; al Faisal and Islam, 2022).
In the short run, FDI has a marginally significant (at a 10% level) effect (4.862) on institutional quality. Institutional quality has a marginally significant (at a 10% level) negative Impact (-0.012). These findings disagree with the findings by Hyun 2006 in his study on 62 developing countries covering the period 1984-2003, where he found no evidence in favour of short-run causality between FDI and institutional quality. Detailed coverage of the literature review by Belfiq et al. (2021) citing by Daude and Stein (2007), also raised the aspect of FDI becoming endogenous where foreign investors in host countries demand better institutions.
Gross fixed capital formation has a highly significant (at 1% level) positive effect (0.635) on FDI. This indirectly agrees with Ullah et al. (2014) in the study for Pakistan in 1976 to 2010 period, where they find causality from domestic investment to economic growth and from economic growth to FDI. Economic growth has a positive (at 1% level) effect (0.027) on gross fixed capital formation. We approach this from the premise that domestic savings are invested in line with the findings of Sinha and Sinha (1998), Saltz (1999), Agrawal (2001), Anoruo and Ahmad (2001), and Narayan and Narayan (2006), all cited in Abu Al Foul (2010).
In their literature review, Didelija, 2021 extensively showed causality between economic growth and savings. Some of the citations include Abu Al-Foul (2010), who finds a two-way link between savings and economic growth in Morocco; van Wyk and Kapingura (2021), who find a positive causality from economic to savings in South Africa; Tang (2008) finding a bi-directional link between savings and economic growth in Malaysia. So, by that savings will be invested, economic growth has leverage on gross domestic investment.
Our next step is to establish whether our results for every model (Y, inq, fdi and human) can be taken seriously.
Autocorrelation. We have established that our model has no autocorrelation, as the coefficients are significant at 5%. For Model 1, where Y is a dependent variable, the Durbin-Watson statistic is a clear 1.840452, supported by the Breusch-Godfrey test (p-value of 0.7156).
Heteroskedasticity. Our models are all homoscedastic. For Model 1, where $Y$ is a dependent variable.
Normality. All our models, save the one with institutional quality and a dependent variable, passed the normality test where errors are typically distributed as the coefficients are significant at 5\% for each equation or the overall model. With institutional quality as a dependent variable, the probability of the $\mathrm{Ch}^2$ statistic (0.0141) is insignificant.
Stability. We checked and found that all models are stable. For illustration, Figure 1 shows the results for the model where Y is the independent variable.
5. Conclusion
In this paper, we use the ARDL method to find the Impact of institutional quality on economic growth in Tanzania from 1990 to 2021 by analysing the Impact of institutions on economic growth and examining if the eventual Impact differs depending on the development degree. The results support the central hypothesis that institutional quality positively impacts economic growth. In addition, all adjustment terms in the respective models that have a long-run relationship have correct (negative) signs and are smaller than one implying there is convergence in the long run; that is, the models returned to their long-run equilibrium; the rate (or speed) at which this happened ranged between 15% to 106.6% annually. We contend that Institutional quality has a significant affirmative (0.047) causal long-run effect on economic growth. This finding is supported by extensive coverage in the literature review section (Abubakar, 2020; Agyey and Idan, 2022; Da Viega, 2022; Dada and Abanikanda, 2022; Emara and Rebolledo, 2021; Hartmann and Spurk, 2020; Iheanou et al., 2021; Islam and Shindaini, 2021; Matallah and Benlahcene, 2021; and Utile et, al., 2021).
Economic growth and gross fixed capital formation have long-term causal effects on institutional quality. However, it must be considered that all the empirical research investigating the relationship between institutions and economic growth still must face at least two kinds of problems upstream. The first difficulty is related to the determination of good institutional quality indicators: the impressive number of indicators elaborated by multilateral organizations, risk-rating agencies, academic institutions, and nongovernmental organizations present ambiguous results stemming from endogenous variables or collinearity between them, and they often lack a theoretical framework linking the indicator to previously defined institutional quality criteria. In addition, most of them refer to the socio-political sphere neglecting the administrative one due to the impossibility of declining variables about the various legal and juridical systems in a homogenous cross-country way.
The definition of growth itself then represents the second problem. Economic growth is currently associated with GDP per capita, a valuable tool to approximate growth trends easily comparable among countries and for these reasons, for a long time, worldwide accepted as an indicator for well-being and development too. Unfortunately, finding new tools is complex, and the debate is still ongoing. One thing is for sure: "we cannot face the challenges of the future with tools from the past.
Appendix A:
Bounds co-integration test formulation
$\begin{aligned} \Delta Y_t= & a_{10}+\beta_{11} \Delta Y_{t-i}+\beta_{21} \Delta \text { inq }_{t-i}+\beta_{31} \Delta \text { gfcf }_{t-i}+\beta_{41} \Delta \text { fdi }_{t-i}+\beta_{51} \Delta \text { human }_{t-i}+\sum{ }_{i=1}^p a_{1 j} \Delta Y_{t-i}+\sum{ }_{i=1}^{q 1} a_{2 j} \Delta \text { inq }_{t-i} \\ & +\sum{ }^{q 2}{ }_{i=1} a_{3 j} \Delta \text { ggfcf }_{t-i}+\sum^{q 3}{ }_{i=1} a_{4 j} \Delta \text { fdi }_{t-i}+\sum{ }^{q 4}{ }_{i=1} a_{5 j} \Delta \text { human }_{t-i}+\varepsilon_{1 t}\end{aligned}$ (8)
$\begin{aligned} \Delta \text { inq }_t= & a_{20}+\beta_{12} \Delta Y_{t-i}+\beta_{22} \Delta \text { inq }_{t-i}+\beta_{32} \Delta \text { gfcf }_{t-i}+\beta_{42} \Delta \text { fdi }_{t-i}+\beta_{52} \Delta \text { human }_{t-i}+\sum p_{i=1} a_{1 j} \Delta \text { inq }_{t-i}+\sum{ }^{q 1}{ }_{i=1} a_{2 j} \Delta Y_{t-l} \\ & +\sum{ }^{q 2}{ }_{i=1} a_{3 j} \Delta \text { ggfc }_{t-i}+\sum{ }^{q 3}{ }_{i=1} a_{4 j} \Delta \text { fdi }_{t-i}+\sum{ }^{q 4}{ }_{i=1} a_{5 j} \Delta \text { human }_{t-i}+\varepsilon_{2 t}\end{aligned}$ (9)
$\begin{aligned} \Delta \mathrm{fdi}_{\mathrm{t}}= & \mathrm{a}_{30}+\beta_{13} \Delta \mathrm{Y}_{\mathrm{t}-\mathrm{i}}+\beta_{23} \Delta \text { inq }_{\mathrm{t}-\mathrm{i}}+\beta_{33} \Delta \text { gfcf }_{\mathrm{t}-\mathrm{i}}+\beta_{43} \Delta \text { fdi }_{\mathrm{t}-\mathrm{i}}+\beta_{53} \Delta \text { human }_{\mathrm{t}-\mathrm{i}}+\sum^{\mathrm{p}}{ }_{\mathrm{i}=1} \mathrm{a}_{1 j} \Delta \mathrm{fdi}_{\mathrm{t}-\mathrm{i}}+\sum{ }^{\mathrm{q} 1}{ }_{\mathrm{i}=1} \mathrm{a}_{2 \mathrm{j}} \Delta \mathrm{Y}_{\mathrm{t}-\mathrm{i}} \\ & +\sum{ }^{\mathrm{q} 2}{ }_{\mathrm{i}=1} \mathrm{a}_{3 \mathrm{j}} \Delta \text { inq }_{\mathrm{t}-\mathrm{i}}+\sum{ }^{\mathrm{q} 3}{ }_{\mathrm{i}=1} \mathrm{a}_{4 \mathrm{j}} \Delta \text { ggfcf }_{\mathrm{t}-\mathrm{i}}+\sum{ }^{\mathrm{q} 4}{ }_{\mathrm{i}=1} \mathrm{a}_5 \Delta \text { human }_{\mathrm{t}-\mathrm{i}}+\varepsilon_{3 \mathrm{t}}\end{aligned}$ (10)
$\begin{aligned}\Delta \text { human }_t= & \alpha_{40}+\beta_{14} \Delta Y_{t-i}+\beta_{24} \Delta \text { inq }_{t-i}+\beta_{34} \Delta \text { gfcf }_{t-i}+\beta_{44} \Delta \text { fdi }_{t-i}+\beta_{54} \Delta \text { human }_{t-i}+\sum \mathrm{p}_{i=1} \alpha_{1 j} \Delta \text { human }_{t-i} \sum{ }^{q 1}{ }_{i=1} \alpha_{2 j} \Delta Y_{t-i} \\& +\sum{ }^{q 2}{ }_{i=1} \alpha_{3 j} \Delta \text { inq }_{t-i}+\sum{ }^{q 3}{ }_{i=1} \alpha_{4 j} \Delta \text { ggfcf }_{t-i}+\sum{ }^{q 4}{ }_{i=1} \alpha_{5 j} \Delta \text { fdi }_{t-i}+\varepsilon_{4 t}\end{aligned}$ (11)
Appendix B:
The models (we need to be clear about which is s long run or a short-run model)
Error Correction Models
$\Delta Y_t=a_{10}+\sum p_{i=1} a_{1 i} \Delta Y_{t-i}+\sum q_{i=1} a_{2 i} \Delta$ inq $_{t-i}+\sum q_{i=1} a_{3 i} \Delta$ gfcf $_{t-i}+\sum q_{i=1} a_{4 i} \Delta f d i_{t-i}+\sum q_{i=1} a_{5 i} \Delta$ human $_{t-i}+$$\lambda E C T_{t-i}+\varepsilon_{1 t}$ (12)
$\Delta i n q_t=a_{20}+\sum p_{i=1} a_{1 j} \Delta i n q_{t-i}+\sum q_{i=1} a_{2 j} \Delta Y_{t-i}+\sum q_{i=1} a_{3 i} \Delta g f c f_{t-i} \sum q_{i=1} a_{4 j} \Delta f d i_{t-i}+\sum q_{i=1} a_{5 j} \Delta$ human $_{t-i}$$+\lambda E C T_{t-i}+\varepsilon_{2 t}$ (13)
$\Delta f d i_t=a_{30}+\sum{ }_{i=1}^p a_{1 j} \Delta f d i_{t-i}+\sum q_{i=1} a_{2 j} \Delta Y_{t-i-i}+\sum q_{i=1} a_{3 j} \Delta i n q_{t-i}+\sum q_{i=1} a_{4 i} \Delta g f c f_{t-i}+\sum q_{i=1} a_{5 j} \Delta$ human $_{t-i}$$+\lambda E C T_{t-i}+\varepsilon_{3 t}$ (14)
$\Delta$ human $_t=a_{40}+\sum q_{i=1} a_{1 j} \Delta$ human $_{t-i} \sum{ }^p{ }_{i=1} a_{2 j} \Delta Y_{t-i}+\sum{ }_{i=1} a_{3 j} \Delta$ inq $_{t-i}+\sum{ }_{i=1} a_{4 i} \Delta$ gfcf $_{t-i}+\sum{ }_{i=1} a_{5 j} \Delta$ fdi $_{t-i}$$+\lambda E C T_{t-i}+\varepsilon_{4 t}$ (15)
ARDL model
$\Delta \operatorname{gfcf}_t=a_{10}+\sum p_{i=1} a_{1 j} \Delta \operatorname{ggfcf}_{t-i}+\sum q_{i=1} a_{2 j} \Delta Y_{t-i}+\sum q_{i=1} a_{3 j} \Delta \operatorname{inq}_{t-i}+\sum q_{i=1} a_{4 j} \Delta \operatorname{fdi}_{t-i}+$$\sum{ }^q{ }_{i=1} a_{5 j} \Delta$ human $_{t-i}+\boldsymbol{\varepsilon}_{4 t}$ (16)
Views: 4